Onyxl's picture
Upload 2661 files
8c763fb verified
// Copyright (c) Microsoft Corporation.
// Licensed under the MIT License.
using System.Collections;
using System.Collections.Generic;
using System.Collections.ObjectModel;
using System.Diagnostics.Tracing;
using System.Globalization;
using System.IO;
using System.Linq;
using System.Management.Automation.Runspaces;
using System.Management.Automation.Security;
using System.Management.Automation.Subsystem;
using System.Management.Automation.Subsystem.DSC;
using System.Reflection;
using System.Runtime.CompilerServices;
using System.Threading.Tasks;
using Dsc = Microsoft.PowerShell.DesiredStateConfiguration.Internal;
namespace System.Management.Automation.Language
{
using KeyValuePair = Tuple<ExpressionAst, StatementAst>;
using IfClause = Tuple<PipelineBaseAst, StatementBlockAst>;
using SwitchClause = Tuple<ExpressionAst, StatementBlockAst>;
internal enum ParseMode
{
Default = 0,
ModuleAnalysis = 1
}
/// <summary>
/// The parser that parses PowerShell script and returns a <see cref="ScriptBlockAst"/>, tokens, and error messages
/// if the script cannot be parsed successfully.
/// </summary>
public sealed class Parser
{
private readonly Tokenizer _tokenizer;
internal Token _ungotToken;
private bool _disableCommaOperator;
private bool _savingTokens;
private bool _inConfiguration;
private ParseMode _parseMode;
internal string _fileName;
internal bool ProduceV2Tokens { get; set; }
internal const string VERBATIM_ARGUMENT = "--%";
internal const string VERBATIM_PARAMETERNAME = "-%"; // Same as VERBATIM_ARGUMENT w/o the first '-'.
internal Parser()
{
_tokenizer = new Tokenizer(this);
ErrorList = new List<ParseError>();
_fileName = null;
}
/// <summary>
/// Parse input from the specified file.
/// </summary>
/// <param name="fileName">The name of the file to parse.</param>
/// <param name="tokens">Returns the tokens from parsing the script.</param>
/// <param name="errors">Returns errors, if any, discovered while parsing the script.</param>
/// <returns>The <see cref="ScriptBlockAst"/> that represents the input script file.</returns>
public static ScriptBlockAst ParseFile(string fileName, out Token[] tokens, out ParseError[] errors)
{
const string scriptSchemaExtension = ".schema.psm1";
var parseDscResource = false;
// If the file has the 'schema.psm1' extension, then it is a 'DSC module file' however we don't actually load the
// module at parse time so we can't use the normal mechanisms to bind the module name for configuration commands.
// As an alternative, we extract the base name of the module file and use that as the module name for any keywords exported by this file.
var parser = new Parser();
if (!string.IsNullOrEmpty(fileName) && fileName.Length > scriptSchemaExtension.Length && fileName.EndsWith(scriptSchemaExtension, StringComparison.OrdinalIgnoreCase))
{
parser._keywordModuleName = Path.GetFileName(fileName.AsSpan(0, fileName.Length - scriptSchemaExtension.Length)).ToString();
parseDscResource = true;
}
string scriptContents;
try
{
var esi = new ExternalScriptInfo(fileName, fileName);
scriptContents = esi.ScriptContents;
}
catch (Exception e)
{
var emptyExtent = new EmptyScriptExtent();
var errorMsg = string.Format(CultureInfo.CurrentCulture, ParserStrings.FileReadError, e.Message);
errors = new[] { new ParseError(emptyExtent, "FileReadError", errorMsg) };
tokens = Array.Empty<Token>();
return new ScriptBlockAst(emptyExtent, null, new StatementBlockAst(emptyExtent, null, null), false);
}
var tokenList = new List<Token>();
ScriptBlockAst result;
try
{
if (!parseDscResource)
{
DynamicKeyword.Push();
}
result = parser.Parse(fileName, scriptContents, tokenList, out errors, ParseMode.Default);
}
catch (Exception e)
{
throw new ParseException(ParserStrings.UnrecoverableParserError, e);
}
finally
{
if (!parseDscResource)
{
DynamicKeyword.Pop();
}
}
tokens = tokenList.ToArray();
return result;
}
private string _keywordModuleName;
/// <summary>
/// Parse input that does not come from a file.
/// </summary>
/// <param name="input">The input to parse.</param>
/// <param name="tokens">Returns the tokens from parsing the script.</param>
/// <param name="errors">Returns errors, if any, discovered while parsing the script.</param>
/// <returns>The <see cref="ScriptBlockAst"/> that represents the input script file.</returns>
public static ScriptBlockAst ParseInput(string input, out Token[] tokens, out ParseError[] errors)
{
return ParseInput(input, null /* fileName */, out tokens, out errors);
}
/// <summary>
/// Parse input that does not come from a file.
/// </summary>
/// <param name="input">The input to parse.</param>
/// <param name="fileName">The fileName if present or null.</param>
/// <param name="tokens">Returns the tokens from parsing the script.</param>
/// <param name="errors">Returns errors, if any, discovered while parsing the script.</param>
/// <returns>The <see cref="ScriptBlockAst"/> that represents the input script file.</returns>
public static ScriptBlockAst ParseInput(string input, string fileName, out Token[] tokens, out ParseError[] errors)
{
ArgumentNullException.ThrowIfNull(input);
Parser parser = new Parser();
List<Token> tokenList = new List<Token>();
ScriptBlockAst result;
try
{
result = parser.Parse(fileName, input, tokenList, out errors, ParseMode.Default);
}
catch (Exception e)
{
throw new ParseException(ParserStrings.UnrecoverableParserError, e);
}
tokens = tokenList.ToArray();
return result;
}
internal ScriptBlockAst Parse(string fileName, string input, List<Token> tokenList, out ParseError[] errors, ParseMode parseMode)
{
try
{
return ParseTask(fileName, input, tokenList, false, parseMode);
}
finally
{
errors = ErrorList.ToArray();
}
}
private ScriptBlockAst ParseTask(string fileName, string input, List<Token> tokenList, bool recursed, ParseMode parseMode)
{
bool etwEnabled = ParserEventSource.Log.IsEnabled();
if (etwEnabled) ParserEventSource.Log.ParseStart(ParserEventSource.GetFileOrScript(fileName, input), input.Length);
ScriptBlockAst ast = null;
_parseMode = parseMode;
_fileName = fileName;
_tokenizer.Initialize(fileName, input, tokenList);
_savingTokens = (tokenList != null);
ErrorList.Clear();
try
{
ast = ScriptBlockRule(null, isFilter: false);
#if DEBUG
ast.InternalVisit(new CheckAllParentsSet(ast));
#endif
ast.ScriptRequirements = _tokenizer.GetScriptRequirements();
if (parseMode == ParseMode.Default)
{
ast.PerformPostParseChecks(this);
}
#if DEBUG
ast.InternalVisit(new CheckTypeBuilder());
#endif
}
catch (InsufficientExecutionStackException)
{
if (!recursed)
{
// We'll try parsing once more, this time on a new thread. The assumption here is
// that the stack was close to overflowing before we tried to parse, and that won't
// be a problem on a new thread.
var task = new Task<ScriptBlockAst>(() => ParseTask(fileName, input, tokenList, true, parseMode));
task.Start();
task.Wait();
ast = task.Result;
}
else
{
ReportError(_tokenizer.CurrentExtent(),
nameof(ParserStrings.ScriptTooComplicated),
ParserStrings.ScriptTooComplicated);
}
}
if (etwEnabled) ParserEventSource.Log.ParseStop();
return ast;
}
// This helper routine is used from the runtime to convert a string to a number.
internal static object ScanNumber(string str, Type toType, bool shouldTryCoercion = true)
{
str = str.Trim();
if (str.Length == 0)
{
// For backwards compatibility, we treat the empty string (w/ or w/o whitespace) as the
// integer constant 0. This is a slight change in semantics, the empty string was previously
// only considered a number if we committed to a numeric operation, e.g.:
// 5 - "" - this results in 5
// Returning 0 here means that
// "" - "" - will result in 0, previously was an error.
// This is mostly of interest for the operators -, /, %, -band, -bor, -bxor as these operators
// allow strings as the lval and still do a numeric operation, whereas + and * treat string
// lvals differently.
return 0;
}
var tokenizer = (new Parser())._tokenizer;
tokenizer.Initialize(null, str, null);
tokenizer.AllowSignedNumbers = true;
var token = tokenizer.NextToken() as NumberToken;
if (token == null || !tokenizer.IsAtEndOfScript(token.Extent))
{
if (shouldTryCoercion)
{
// We call ConvertTo, primarily because we expect it will throw an exception,
// but it's possible it could succeed, e.g. if the string had commas, our lexer
// will fail, but Convert.ChangeType could succeed.
return LanguagePrimitives.ConvertTo(str, toType, CultureInfo.InvariantCulture);
}
else
{
throw new ParseException();
}
}
return token.Value;
}
internal static ITypeName ScanType(string typename, bool ignoreErrors)
{
typename = typename.Trim();
if (typename.Length == 0)
{
return null;
}
var parser = new Parser();
var tokenizer = parser._tokenizer;
tokenizer.Initialize(null, typename, null);
var result = parser.TypeNameRule(allowAssemblyQualifiedNames: true, firstTypeNameToken: out _);
SemanticChecks.CheckArrayTypeNameDepth(result, PositionUtilities.EmptyExtent, parser);
if (!ignoreErrors && result is not null && (parser.ErrorList.Count > 0 || !result.Extent.Text.Equals(typename, StringComparison.OrdinalIgnoreCase)))
{
result = null;
}
return result;
}
internal static ExpressionAst ScanString(string str)
{
str = str.Replace("\"", "\"\"");
var parser = new Parser();
parser._tokenizer.Initialize(null, '"' + str + '"', null);
var strToken = (StringExpandableToken)parser._tokenizer.NextToken();
var ast = parser.ExpandableStringRule(strToken);
if (parser.ErrorList.Count > 0)
{
throw new ParseException(parser.ErrorList.ToArray());
}
return ast;
}
private string _previousFirstTokenText;
private string _previousLastTokenText;
private static bool IgnoreTokenWhenUpdatingPreviousFirstLast(Token token)
{
return (token.Kind == TokenKind.Variable || token.Kind == TokenKind.Generic) &&
(token.Text.Equals("$^", StringComparison.OrdinalIgnoreCase) ||
token.Text.Equals("$$", StringComparison.OrdinalIgnoreCase));
}
internal void SetPreviousFirstLastToken(ExecutionContext context)
{
var firstToken = _tokenizer.FirstToken;
if (firstToken != null)
{
context.SetVariable(SpecialVariables.FirstTokenVarPath, _previousFirstTokenText);
if (!IgnoreTokenWhenUpdatingPreviousFirstLast(firstToken))
{
var stringToken = firstToken as StringToken;
_previousFirstTokenText = stringToken != null
? stringToken.Value
: firstToken.Text;
}
context.SetVariable(SpecialVariables.LastTokenVarPath, _previousLastTokenText);
var lastToken = _tokenizer.LastToken;
if (!IgnoreTokenWhenUpdatingPreviousFirstLast(lastToken))
{
var stringToken = lastToken as StringToken;
_previousLastTokenText = stringToken != null
? stringToken.Value
: lastToken.Text;
}
}
}
internal List<ParseError> ErrorList { get; }
#region Utilities
private void SkipNewlines()
{
if (_ungotToken == null || _ungotToken.Kind == TokenKind.NewLine)
{
_ungotToken = null;
_tokenizer.SkipNewlines(skipSemis: false);
}
}
private void SkipNewlinesAndSemicolons()
{
if (_ungotToken == null || _ungotToken.Kind == TokenKind.NewLine || _ungotToken.Kind == TokenKind.Semi)
{
_ungotToken = null;
_tokenizer.SkipNewlines(skipSemis: true);
}
}
private void SyncOnError(bool consumeClosingToken, params TokenKind[] syncTokens)
{
int parens = syncTokens.Contains(TokenKind.RParen) ? 1 : 0;
int curlies = syncTokens.Contains(TokenKind.RCurly) ? 1 : 0;
int braces = syncTokens.Contains(TokenKind.RBracket) ? 1 : 0;
while (true)
{
Token token = NextToken();
switch (token.Kind)
{
case TokenKind.LParen: ++parens; break;
case TokenKind.RParen:
--parens;
if (parens == 0 && syncTokens.Contains(TokenKind.RParen))
{
if (!consumeClosingToken)
{
UngetToken(token);
}
return;
}
break;
case TokenKind.LCurly: ++curlies; break;
case TokenKind.RCurly:
--curlies;
if (curlies == 0 && syncTokens.Contains(TokenKind.RCurly))
{
if (!consumeClosingToken)
{
UngetToken(token);
}
return;
}
break;
case TokenKind.LBracket: ++braces; break;
case TokenKind.RBracket:
--braces;
if (braces == 0 && syncTokens.Contains(TokenKind.RBracket))
{
if (!consumeClosingToken)
{
UngetToken(token);
}
return;
}
break;
case TokenKind.EndOfInput:
// Never consume <EOF>, but return it to caller
UngetToken(token);
return;
}
if (syncTokens.Contains(token.Kind) && parens == 0 && curlies == 0 && braces == 0)
{
break;
}
}
}
private Token NextToken()
{
Token token = _ungotToken ?? _tokenizer.NextToken();
_ungotToken = null;
return token;
}
private Token PeekToken()
{
Token token = _ungotToken ?? _tokenizer.NextToken();
_ungotToken ??= token;
return token;
}
private Token NextMemberAccessToken(bool allowLBracket)
{
// If _ungotToken is not null, we're in some sort of error state, don't return the token.
if (_ungotToken != null)
return null;
return _tokenizer.GetMemberAccessOperator(allowLBracket);
}
private Token NextInvokeMemberToken()
{
// If _ungotToken is not null, we're in some sort of error state, don't return the token.
if (_ungotToken != null)
return null;
return _tokenizer.GetInvokeMemberOpenParen();
}
private Token NextLBracket()
{
if (_ungotToken != null)
{
if (_ungotToken.Kind == TokenKind.LBracket) return NextToken();
// If _ungotToken is not null, we're in some sort of error state, don't return the token.
return null;
}
return _tokenizer.GetLBracket();
}
private StringToken GetVerbatimCommandArgumentToken()
{
if (_ungotToken == null || _ungotToken.Kind == TokenKind.Parameter)
{
_ungotToken = null;
return _tokenizer.GetVerbatimCommandArgument();
}
return null;
}
private void SkipToken()
{
Diagnostics.Assert(_ungotToken != null, "Don't skip a token you didn't unget");
_ungotToken = null;
}
private void UngetToken(Token token)
{
Diagnostics.Assert(_ungotToken == null, "Only 1 token lookahead is supported");
_ungotToken = token;
}
private void SetTokenizerMode(TokenizerMode mode)
{
if (mode != _tokenizer.Mode && _ungotToken != null)
{
// Only rescan tokens that differ b/w command and expression modes.
if (!_ungotToken.Kind.HasTrait(TokenFlags.ParseModeInvariant))
{
Resync(_ungotToken);
}
#if DEBUG
else if (_ungotToken.Kind != TokenKind.EndOfInput)
{
// Verify the comment above.
Token ungotToken = _ungotToken;
var oldTokenList = _tokenizer.TokenList;
_tokenizer.TokenList = null;
Resync(_ungotToken);
_tokenizer.Mode = mode;
Token rescan = _tokenizer.NextToken();
Diagnostics.Assert(ungotToken.Kind == rescan.Kind, "Rescan failed to return same kind");
Diagnostics.Assert(ungotToken.Text == rescan.Text, "Rescan failed to return same text");
IScriptPosition pos1 = ungotToken.Extent.StartScriptPosition;
IScriptPosition pos2 = rescan.Extent.StartScriptPosition;
Diagnostics.Assert(pos1.ColumnNumber == pos2.ColumnNumber, "Rescan failed to return same start column");
Diagnostics.Assert(pos1.LineNumber == pos2.LineNumber, "Rescan failed to return same start line#");
pos1 = ungotToken.Extent.EndScriptPosition;
pos2 = rescan.Extent.EndScriptPosition;
Diagnostics.Assert(pos1.ColumnNumber == pos2.ColumnNumber, "Rescan failed to return same end column");
Diagnostics.Assert(pos1.LineNumber == pos2.LineNumber, "Rescan failed to return same end line#");
// Make sure we leave things as they were - Resync clears _ungotToken.
_ungotToken = ungotToken;
_tokenizer.TokenList = oldTokenList;
}
#endif
}
_tokenizer.Mode = mode;
}
private void Resync(Token token)
{
_ungotToken = null;
_tokenizer.Resync(token);
}
private void Resync(int restorePoint)
{
_ungotToken = null;
_tokenizer.Resync(restorePoint);
}
private static bool IsSpecificParameter(Token token, string parameter)
{
Diagnostics.Assert(token.Kind == TokenKind.Parameter, "Token must be a ParameterToken");
var paramToken = (ParameterToken)token;
return parameter.StartsWith(paramToken.ParameterName, StringComparison.OrdinalIgnoreCase);
}
internal void RequireStatementTerminator()
{
var terminatorToken = PeekToken();
if (terminatorToken.Kind == TokenKind.NewLine || terminatorToken.Kind == TokenKind.Semi)
{
SkipToken();
}
else if (terminatorToken.Kind != TokenKind.EndOfInput)
{
ReportIncompleteInput(terminatorToken.Extent,
nameof(ParserStrings.MissingStatementTerminator),
ParserStrings.MissingStatementTerminator);
}
}
internal static IScriptExtent ExtentOf(IScriptExtent first, IScriptExtent last)
{
if (first is EmptyScriptExtent) return last;
if (last is EmptyScriptExtent) return first;
Diagnostics.Assert(first is InternalScriptExtent && last is InternalScriptExtent,
"Private method expects internal position representation");
InternalScriptExtent l = (InternalScriptExtent)first;
InternalScriptExtent r = (InternalScriptExtent)last;
Diagnostics.Assert(l.PositionHelper == r.PositionHelper, "Can't get the extent across files");
return new InternalScriptExtent(l.PositionHelper, l.StartOffset, r.EndOffset);
}
internal static IScriptExtent Before(IScriptExtent extent)
{
Diagnostics.Assert(extent is InternalScriptExtent, "Private method expects internal position representation");
InternalScriptExtent scriptExtent = (InternalScriptExtent)extent;
int offset = scriptExtent.StartOffset - 1;
if (offset < 0) offset = 0;
return new InternalScriptExtent(scriptExtent.PositionHelper, offset, offset);
}
internal static IScriptExtent After(IScriptExtent extent)
{
Diagnostics.Assert(extent is InternalScriptExtent, "Private method expects internal position representation");
InternalScriptExtent scriptExtent = (InternalScriptExtent)extent;
int offset = scriptExtent.EndOffset;
return new InternalScriptExtent(scriptExtent.PositionHelper, offset, offset);
}
internal static IScriptExtent LastCharacterOf(IScriptExtent extent)
{
Diagnostics.Assert(extent is InternalScriptExtent, "Private method expects internal position representation");
InternalScriptExtent scriptExtent = (InternalScriptExtent)extent;
int offset = scriptExtent.EndOffset - 1;
if (offset < 0)
{
offset = 0;
}
return new InternalScriptExtent(scriptExtent.PositionHelper, offset, offset);
}
internal static IScriptExtent ExtentFromFirstOf(params object[] objs)
{
foreach (object obj in objs)
{
if (obj != null)
{
var token = obj as Token;
if (token != null)
{
return token.Extent;
}
var ast = obj as Ast;
if (ast != null)
{
return ast.Extent;
}
var typename = obj as ITypeName;
if (typename != null)
{
return typename.Extent;
}
Diagnostics.Assert(obj is IScriptExtent, "Only accepts tokens, asts, and IScriptExtents");
return (IScriptExtent)obj;
}
}
Diagnostics.Assert(false, "One of the objects must not be null");
return PositionUtilities.EmptyExtent;
}
internal static IScriptExtent ExtentOf(Token first, Token last) { return ExtentOf(first.Extent, last.Extent); }
internal static IScriptExtent ExtentOf(Ast first, Ast last) { return ExtentOf(first.Extent, last.Extent); }
internal static IScriptExtent ExtentOf(Ast first, Token last) { return ExtentOf(first.Extent, last.Extent); }
internal static IScriptExtent ExtentOf(Token first, Ast last) { return ExtentOf(first.Extent, last.Extent); }
internal static IScriptExtent ExtentOf(IScriptExtent first, Ast last) { return ExtentOf(first, last.Extent); }
internal static IScriptExtent ExtentOf(IScriptExtent first, Token last) { return ExtentOf(first, last.Extent); }
internal static IScriptExtent ExtentOf(Ast first, IScriptExtent last) { return ExtentOf(first.Extent, last); }
internal static IScriptExtent ExtentOf(Token first, IScriptExtent last) { return ExtentOf(first.Extent, last); }
// private static IScriptExtent Before(Ast ast) { return Before(ast.Extent); }
internal static IScriptExtent Before(Token token) { return Before(token.Extent); }
internal static IScriptExtent After(Ast ast) { return After(ast.Extent); }
internal static IScriptExtent After(Token token) { return After(token.Extent); }
private static IEnumerable<Ast> GetNestedErrorAsts(params object[] asts)
{
foreach (var obj in asts)
{
if (obj != null)
{
Ast ast = obj as Ast;
if (ast != null)
{
yield return ast;
}
else
{
var enumerable = obj as IEnumerable<Ast>;
if (enumerable != null)
{
foreach (var ast2 in enumerable)
{
if (ast2 != null)
{
yield return ast2;
}
}
}
else
{
Diagnostics.Assert(false, "Caller to pass only asts, or IEnumerable<Ast>");
}
}
}
}
}
/// <summary>
/// Parses the specified constant hashtable string into a Hashtable object.
/// </summary>
/// <param name="input">The Hashtable string.</param>
/// <param name="result">The Hashtable object.</param>
/// <returns></returns>
internal static bool TryParseAsConstantHashtable(string input, out Hashtable result)
{
result = null;
if (string.IsNullOrWhiteSpace(input))
{
return false;
}
Token[] throwAwayTokens;
ParseError[] parseErrors;
var ast = Parser.ParseInput(input, out throwAwayTokens, out parseErrors);
if (ast == null
|| parseErrors.Length > 0
|| ast.BeginBlock != null
|| ast.ProcessBlock != null
|| ast.CleanBlock != null
|| ast.DynamicParamBlock != null
|| ast.EndBlock.Traps != null)
{
return false;
}
var statements = ast.EndBlock.Statements;
if (statements.Count != 1)
{
return false;
}
if (!(statements[0] is PipelineAst pipelineAst))
{
return false;
}
var expr = pipelineAst.GetPureExpression();
if (expr == null)
{
return false;
}
if (!(expr is HashtableAst hashTableAst))
{
return false;
}
object hashtable;
if (!IsConstantValueVisitor.IsConstant(hashTableAst, out hashtable, forRequires: true))
{
return false;
}
var data = hashtable as Hashtable;
Diagnostics.Assert((data != null), "IsConstantValueVisitor.IsConstant() should return false when the specified HashtableAst is not a avalid Hashtable");
result = data;
return true;
}
#endregion Utilities
#region Statements
private ScriptBlockAst ScriptBlockRule(Token lCurly, bool isFilter)
{
return ScriptBlockRule(lCurly, isFilter, null);
}
private ScriptBlockAst ScriptBlockRule(Token lCurly, bool isFilter, StatementAst predefinedStatementAst)
{
// G script-block:
// G using-statements:opt param-block:opt statement-terminators:opt script-block-body:opt
// G
// G using-statements:
// G using-statement
// G using-statements using-statement
// We could set the mode here, but we can avoid rescanning keywords if the caller
// sets the mode before skipping newlines.
Diagnostics.Assert(_tokenizer.Mode == TokenizerMode.Command,
"Caller to make sure the mode is correct.");
// Skipping newlines here saves a more expensive resync if there is no parameter block.
SkipNewlines();
var usingStatements = lCurly == null ? UsingStatementsRule() : null;
var restorePoint = _tokenizer.GetRestorePoint();
ParamBlockAst paramBlock = ParamBlockRule();
if (paramBlock == null)
{
// In case we scanned some attributes or type constraints but they didn't
// belong to a param statement, we need to reparse them (because they will
// mean something different, such as a type literal expression, or a cast.)
Resync(restorePoint);
}
SkipNewlinesAndSemicolons();
return ScriptBlockBodyRule(lCurly, usingStatements, paramBlock, isFilter, predefinedStatementAst);
}
private List<UsingStatementAst> UsingStatementsRule()
{
List<UsingStatementAst> result = null;
while (true)
{
var token = PeekToken();
if (token.Kind == TokenKind.Using)
{
SkipToken();
var statement = UsingStatementRule(token);
SkipNewlinesAndSemicolons();
result ??= new List<UsingStatementAst>();
var usingStatement = statement as UsingStatementAst;
// otherwise returned statement is ErrorStatementAst.
// We ignore it here, because error already reported to the parser.
if (usingStatement != null)
{
result.Add(usingStatement);
}
continue;
}
// Normally we don't need to resync, but our caller may speculatively scan much more than
// one token, so it needs the restore point to include this token.
Resync(token);
break;
}
return result;
}
private ParamBlockAst ParamBlockRule()
{
// G param-block:
// G new-lines:opt attribute-list:opt new-lines:opt 'param' new-lines:opt
// G '(' parameter-list:opt new-lines:opt ')'
SkipNewlines();
List<AttributeBaseAst> candidateAttributes = AttributeListRule(false);
SkipNewlines();
Token paramToken = PeekToken();
if (paramToken.Kind != TokenKind.Param)
{
return null;
}
SkipToken();
SkipNewlines();
Token lparen = NextToken();
if (lparen.Kind != TokenKind.LParen)
{
UngetToken(lparen);
// This is not an error, we'll end up trying to invoke a command named 'param'.
return null;
}
List<ParameterAst> parameters = ParameterListRule();
SkipNewlines();
Token rParen = NextToken();
var endExtent = rParen.Extent;
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: assume we saw the closing paren and continue like normal.
UngetToken(rParen);
endExtent = Before(rParen);
ReportIncompleteInput(After(parameters != null && parameters.Count > 0 ? parameters.Last().Extent : lparen.Extent),
nameof(ParserStrings.MissingEndParenthesisInFunctionParameterList),
ParserStrings.MissingEndParenthesisInFunctionParameterList);
}
List<AttributeAst> attributes = new List<AttributeAst>();
if (candidateAttributes != null)
{
foreach (AttributeBaseAst attr in candidateAttributes)
{
AttributeAst attribute = attr as AttributeAst;
if (attribute != null)
{
attributes.Add(attribute);
}
else
{
// ErrorRecovery: nothing to do, this is a semantic error that is caught in the parser
// because the ast only allows attributes, no type constraints.
ReportError(attr.Extent,
nameof(ParserStrings.TypeNotAllowedBeforeParam),
ParserStrings.TypeNotAllowedBeforeParam,
attr.TypeName.FullName);
}
}
}
return new ParamBlockAst(ExtentOf(paramToken, endExtent), attributes, parameters);
}
private List<ParameterAst> ParameterListRule()
{
// G parameter-list:
// G script-parameter
// G parameter-list new-lines:opt ',' script-parameter
List<ParameterAst> parameters = new List<ParameterAst>();
Token commaToken = null;
while (true)
{
ParameterAst parameter = ParameterRule();
if (parameter == null)
{
if (commaToken != null)
{
// ErrorRecovery: ??
ReportIncompleteInput(After(commaToken),
nameof(ParserStrings.MissingExpressionAfterToken),
ParserStrings.MissingExpressionAfterToken,
commaToken.Kind.Text());
}
break;
}
parameters.Add(parameter);
SkipNewlines();
commaToken = PeekToken();
if (commaToken.Kind != TokenKind.Comma)
{
break;
}
SkipToken();
}
return parameters;
}
private ParameterAst ParameterRule()
{
// G script-parameter:
// G new-lines:opt attribute-list:opt new-lines:opt variable script-parameter-default:opt
// G script-parameter-default:
// G new-lines:opt '=' new-lines:opt expression
List<AttributeBaseAst> attributes;
VariableToken variableToken;
ExpressionAst defaultValue = null;
bool oldDisableCommaOperator = _disableCommaOperator;
var oldTokenizerMode = _tokenizer.Mode;
try
{
_disableCommaOperator = true;
SetTokenizerMode(TokenizerMode.Expression);
SkipNewlines();
attributes = AttributeListRule(false);
SkipNewlines();
Token token = NextToken();
if (token.Kind != TokenKind.Variable && token.Kind != TokenKind.SplattedVariable)
{
UngetToken(token);
if (attributes != null)
{
// ErrorRecovery: skip to closing paren because returning null signals the last parameter.
ReportIncompleteInput(After(attributes.Last()),
nameof(ParserStrings.InvalidFunctionParameter),
ParserStrings.InvalidFunctionParameter);
SyncOnError(true, TokenKind.RParen);
// Even though we don't have a complete parameter, we do have attributes. Intellisense
// might want to complete something in the attributes, so we need to return something useful.
var extent = ExtentOf(attributes[0].Extent, attributes[attributes.Count - 1].Extent);
return new ParameterAst(extent, new VariableExpressionAst(extent, "__error__", false), attributes, null);
}
return null;
}
variableToken = ((VariableToken)token);
SkipNewlines();
Token equalsToken = PeekToken();
if (equalsToken.Kind == TokenKind.Equals)
{
SkipToken();
SkipNewlines();
defaultValue = ExpressionRule();
if (defaultValue == null)
{
ReportIncompleteInput(After(equalsToken),
nameof(ParserStrings.MissingExpressionAfterToken),
ParserStrings.MissingExpressionAfterToken,
equalsToken.Kind.Text());
}
}
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
SetTokenizerMode(oldTokenizerMode);
}
IScriptExtent startExtent = (attributes == null) ? variableToken.Extent : attributes[0].Extent;
IScriptExtent endExtent = (defaultValue == null) ? variableToken.Extent : defaultValue.Extent;
return new ParameterAst(ExtentOf(startExtent, endExtent),
new VariableExpressionAst(variableToken), attributes, defaultValue);
}
private List<AttributeBaseAst> AttributeListRule(bool inExpressionMode)
{
// G attribute-list:
// G attribute
// G attribute-list attribute
List<AttributeBaseAst> attributes = new List<AttributeBaseAst>();
AttributeBaseAst attribute = AttributeRule();
while (attribute != null)
{
attributes.Add(attribute);
if (!inExpressionMode || attribute is AttributeAst)
{
SkipNewlines();
}
attribute = AttributeRule();
}
if (attributes.Count == 0)
{
return null;
}
return attributes;
}
private AttributeBaseAst AttributeRule()
{
// G attribute:
// G '[' attribute-name '(' attribute-arguments ')' ']'
// G attribute-name:
// G type-spec
var lBracket = NextLBracket();
if (lBracket == null)
{
return null;
}
SkipNewlines();
Token firstTypeNameToken;
ITypeName typeName = TypeNameRule(allowAssemblyQualifiedNames: true, firstTypeNameToken: out firstTypeNameToken);
if (typeName == null)
{
// ErrorRecovery: Return null so we stop looking for attributes.
Resync(lBracket); // TypeNameRule might have consumed some tokens
ReportIncompleteInput(After(lBracket),
nameof(ParserStrings.MissingTypename),
ParserStrings.MissingTypename);
return null;
}
Token lParenOrRBracket = NextToken();
if (lParenOrRBracket.Kind == TokenKind.LParen)
{
SkipNewlines();
List<ExpressionAst> positionalArguments = new List<ExpressionAst>();
List<NamedAttributeArgumentAst> namedArguments = new List<NamedAttributeArgumentAst>();
IScriptExtent lastItemExtent = lParenOrRBracket.Extent;
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
AttributeArgumentsRule(positionalArguments, namedArguments, ref lastItemExtent);
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
SkipNewlines();
Token rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: pretend we saw a ')', attempt to find an ']'.
UngetToken(rParen);
rParen = null;
ReportIncompleteInput(After(lastItemExtent),
nameof(ParserStrings.MissingEndParenthesisInExpression),
ParserStrings.MissingEndParenthesisInExpression);
}
SkipNewlines();
Token rBracket = NextToken();
if (rBracket.Kind != TokenKind.RBracket)
{
// ErrorRecovery: pretend we saw a ']', return our result.
UngetToken(rBracket);
rBracket = null;
// Don't bother reporting a missing ']' if we reported a missing ')'.
if (rParen != null)
{
ReportIncompleteInput(After(rParen),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfAttribute),
ParserStrings.EndSquareBracketExpectedAtEndOfAttribute);
}
}
firstTypeNameToken.TokenFlags |= TokenFlags.AttributeName;
return new AttributeAst(ExtentOf(lBracket, ExtentFromFirstOf(rBracket, rParen, lastItemExtent)), typeName, positionalArguments, namedArguments);
}
if (ProduceV2Tokens)
{
var typeToken = new Token((InternalScriptExtent)ExtentOf(lBracket, lParenOrRBracket),
TokenKind.Identifier, TokenFlags.TypeName);
_tokenizer.ReplaceSavedTokens(lBracket, lParenOrRBracket, typeToken);
}
if (lParenOrRBracket.Kind != TokenKind.RBracket)
{
UngetToken(lParenOrRBracket);
ReportError(Before(lParenOrRBracket),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfAttribute),
ParserStrings.EndSquareBracketExpectedAtEndOfAttribute);
lParenOrRBracket = null;
}
return new TypeConstraintAst(ExtentOf(lBracket, ExtentFromFirstOf(lParenOrRBracket, typeName.Extent)), typeName);
}
private void AttributeArgumentsRule(ICollection<ExpressionAst> positionalArguments,
ICollection<NamedAttributeArgumentAst> namedArguments,
ref IScriptExtent lastItemExtent)
{
// G attribute-arguments:
// G attribute-argument
// G attribute-argument new-lines:opt ',' attribute-arguments
// G attribute-argument:
// G new-lines:opt expression
// G new-lines:opt property-name '=' new-lines:opt expression
bool oldDisableCommaOperator = _disableCommaOperator;
Token commaToken = null;
HashSet<string> keysSeen = new HashSet<string>();
try
{
_disableCommaOperator = true;
while (true)
{
SkipNewlines();
StringConstantExpressionAst name = SimpleNameRule();
ExpressionAst expr;
bool expressionOmitted = false;
if (name != null)
{
Token token = PeekToken();
if (token.Kind == TokenKind.Equals)
{
token = NextToken();
SkipNewlines();
expr = ExpressionRule();
if (expr == null)
{
// ErrorRecovery: ?
IScriptExtent errorPosition = After(token);
ReportIncompleteInput(
errorPosition,
nameof(ParserStrings.MissingExpressionInNamedArgument),
ParserStrings.MissingExpressionInNamedArgument);
expr = new ErrorExpressionAst(errorPosition);
SyncOnError(true, TokenKind.Comma, TokenKind.RParen, TokenKind.RBracket, TokenKind.NewLine);
}
lastItemExtent = expr.Extent;
}
else
{
// If the expression is missing, assume the value true
// and record that it was defaulted for better error messages.
expr = new ConstantExpressionAst(name.Extent, true);
expressionOmitted = true;
}
}
else
{
expr = ExpressionRule();
}
if (name != null)
{
if (keysSeen.Contains(name.Value))
{
// ErrorRecovery: this is a semantic error, so just keep parsing.
ReportError(name.Extent,
nameof(ParserStrings.DuplicateNamedArgument),
ParserStrings.DuplicateNamedArgument,
name.Value);
}
else
{
namedArguments.Add(new NamedAttributeArgumentAst(ExtentOf(name, expr),
name.Value, expr, expressionOmitted));
}
}
else if (expr != null)
{
positionalArguments.Add(expr);
lastItemExtent = expr.Extent;
}
else if (commaToken != null)
{
// ErrorRecovery: Pretend we saw the argument and keep going.
IScriptExtent errorExtent = After(commaToken);
ReportIncompleteInput(
errorExtent,
nameof(ParserStrings.MissingExpressionAfterToken),
ParserStrings.MissingExpressionAfterToken,
commaToken.Kind.Text());
positionalArguments.Add(new ErrorExpressionAst(errorExtent));
lastItemExtent = errorExtent;
}
SkipNewlines();
commaToken = PeekToken();
if (commaToken.Kind != TokenKind.Comma)
{
break;
}
lastItemExtent = commaToken.Extent;
SkipToken();
}
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
}
}
private ITypeName TypeNameRule(bool allowAssemblyQualifiedNames, out Token firstTypeNameToken)
{
// G type-spec:
// G array-type-name dimension:opt ']'
// G generic-type-name generic-type-arguments ']'
// G type-name
// G array-type-name:
// G type-name '['
// G generic-type-name:
// G type-name '['
// G dimension:
// G ','
// G dimension ','
// G generic-type-arguments:
// G type-spec
// G generic-type-arguments ',' type-spec
// G type-name:
// G namespace-type-name
// G namespace-type-name ',' assembly-name-spec
// G namespace-type-name:
// G nested-type-name
// G namespace-spec '.' nested-type-name
// G nested-type-name:
// G type-name-identifier
// G nested-type-name '+' type-name-identifier
// G namespace-spec:
// G type-name-identifier
// G type-name-identifier '.' type-name-identifier
// G type-name-identifier:
// G type-name-identifier-char
// G type-name-identifier type-name-identifier-char
// G type-name-identifier-char:
// G any letter, digit, '.', '`', or '_'.
// The above grammar is specified by the CLR. We don't attempt to implement the above grammar precisely, e.g.
// we would permit 'a+b.c'. The above grammar would disallow this.
var oldTokenizerMode = _tokenizer.Mode;
try
{
// Switch to typename mode to avoid aggressive argument tokenization.
SetTokenizerMode(TokenizerMode.TypeName);
Token typeName = NextToken();
if (typeName.Kind != TokenKind.Identifier)
{
UngetToken(typeName);
firstTypeNameToken = null;
return null;
}
// firstTypeNameToken is returned so we can mark it as an attribute if it turns out
// that we've got a attribute instead of a type literal.
firstTypeNameToken = typeName;
return FinishTypeNameRule(typeName, allowAssemblyQualifiedNames: allowAssemblyQualifiedNames);
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private ITypeName FinishTypeNameRule(Token typeName, bool unBracketedGenericArg = false, bool allowAssemblyQualifiedNames = true)
{
Diagnostics.Assert(typeName.Kind == TokenKind.Identifier, "Caller must verify the argument.");
Token token = PeekToken();
if (token.Kind == TokenKind.LBracket)
{
var lbracket = token;
// Array or generic
SkipToken();
SkipNewlines();
token = NextToken();
switch (token.Kind)
{
case TokenKind.RBracket:
case TokenKind.Comma:
var elementType = new TypeName(typeName.Extent, typeName.Text);
return CompleteArrayTypeName(elementType, elementType, token, unBracketedGenericArg);
case TokenKind.LBracket:
case TokenKind.Identifier:
return GenericTypeNameRule(typeName, token, unBracketedGenericArg);
default:
// ErrorRecovery: sync to ']', and return non-null to avoid cascading errors.
if (token.Kind != TokenKind.EndOfInput)
{
ReportError(token.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
token.Text);
SyncOnError(true, TokenKind.RBracket);
}
else
{
UngetToken(token);
ReportIncompleteInput(After(lbracket),
nameof(ParserStrings.MissingTypename),
ParserStrings.MissingTypename);
}
return new TypeName(typeName.Extent, typeName.Text);
}
}
if (token.Kind == TokenKind.Comma && allowAssemblyQualifiedNames && !unBracketedGenericArg)
{
SkipToken();
string assemblyNameSpec = _tokenizer.GetAssemblyNameSpec();
if (string.IsNullOrWhiteSpace(assemblyNameSpec))
{
ReportError(After(token),
nameof(ParserStrings.MissingAssemblyNameSpecification),
ParserStrings.MissingAssemblyNameSpecification);
return new TypeName(typeName.Extent, typeName.Text);
}
return new TypeName(ExtentOf(typeName.Extent, _tokenizer.CurrentExtent()), typeName.Text, assemblyNameSpec);
}
return new TypeName(typeName.Extent, typeName.Text);
}
private ITypeName GetSingleGenericArgument(Token firstToken)
{
if (firstToken.Kind == TokenKind.Identifier)
{
return FinishTypeNameRule(firstToken, unBracketedGenericArg: true);
}
Diagnostics.Assert(firstToken.Kind == TokenKind.LBracket, "caller to verify correct token kind");
Token token = NextToken();
if (token.Kind != TokenKind.Identifier)
{
UngetToken(token);
ReportIncompleteInput(After(token),
nameof(ParserStrings.MissingTypename),
ParserStrings.MissingTypename);
return new TypeName(firstToken.Extent, ":ErrorTypeName:");
}
ITypeName typeName = FinishTypeNameRule(token);
if (typeName != null)
{
Token rBracket = NextToken();
if (rBracket.Kind != TokenKind.RBracket)
{
// ErrorRecovery: pretend we saw the closing bracket.
UngetToken(rBracket);
ReportIncompleteInput(Before(rBracket),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfType),
ParserStrings.EndSquareBracketExpectedAtEndOfType);
}
}
return typeName;
}
private List<ITypeName> GenericTypeArgumentsRule(Token firstToken, out Token lastToken)
{
Diagnostics.Assert(firstToken.Kind == TokenKind.Identifier || firstToken.Kind == TokenKind.LBracket, "unexpected first token");
RuntimeHelpers.EnsureSufficientExecutionStack();
var genericArguments = new List<ITypeName>();
ITypeName typeName = GetSingleGenericArgument(firstToken);
genericArguments.Add(typeName);
while (true)
{
SkipNewlines();
lastToken = NextToken();
if (lastToken.Kind != TokenKind.Comma)
{
break;
}
SkipNewlines();
Token token = PeekToken();
if (token.Kind == TokenKind.Identifier || token.Kind == TokenKind.LBracket)
{
SkipToken();
typeName = GetSingleGenericArgument(token);
}
else
{
ReportIncompleteInput(
After(lastToken),
nameof(ParserStrings.MissingTypename),
ParserStrings.MissingTypename);
typeName = new TypeName(lastToken.Extent, ":ErrorTypeName:");
}
genericArguments.Add(typeName);
}
return genericArguments;
}
private ITypeName GenericTypeNameRule(Token genericTypeName, Token firstToken, bool unbracketedGenericArg)
{
List<ITypeName> genericArguments = GenericTypeArgumentsRule(firstToken, out Token rBracketToken);
if (rBracketToken.Kind != TokenKind.RBracket)
{
// ErrorRecovery: pretend we had the closing bracket and just continue on.
UngetToken(rBracketToken);
ReportIncompleteInput(
Before(rBracketToken),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfAttribute),
ParserStrings.EndSquareBracketExpectedAtEndOfAttribute);
rBracketToken = null;
}
var openGenericType = new TypeName(genericTypeName.Extent, genericTypeName.Text, genericArguments.Count);
var result = new GenericTypeName(
ExtentOf(genericTypeName.Extent, ExtentFromFirstOf(rBracketToken, genericArguments.LastOrDefault(), firstToken)),
openGenericType,
genericArguments);
Token token = PeekToken();
if (token.Kind == TokenKind.LBracket)
{
SkipToken();
return CompleteArrayTypeName(result, openGenericType, NextToken(), unbracketedGenericArg);
}
if (token.Kind == TokenKind.Comma && !unbracketedGenericArg)
{
SkipToken();
string assemblyNameSpec = _tokenizer.GetAssemblyNameSpec();
if (string.IsNullOrEmpty(assemblyNameSpec))
{
ReportError(
After(token),
nameof(ParserStrings.MissingAssemblyNameSpecification),
ParserStrings.MissingAssemblyNameSpecification);
}
else
{
openGenericType.AssemblyName = assemblyNameSpec;
}
}
return result;
}
private ITypeName CompleteArrayTypeName(ITypeName elementType, TypeName typeForAssemblyQualification, Token firstTokenAfterLBracket, bool unBracketedGenericArg)
{
while (true)
{
Token token;
switch (firstTokenAfterLBracket.Kind)
{
case TokenKind.Comma:
int dim = 1;
token = firstTokenAfterLBracket;
Token lastComma;
do
{
lastComma = token;
dim += 1;
token = NextToken();
} while (token.Kind == TokenKind.Comma);
// The dimensions for an array must be less than or equal to 32.
// Search the doc for 'Type.MakeArrayType(int rank)' for more details.
if (dim > 32)
{
// If the next token is right bracket, we swallow it to make it easier to parse the rest of script.
// Otherwise, we unget the token for the subsequent parsing to consume.
if (token.Kind != TokenKind.RBracket)
{
UngetToken(token);
}
ReportError(
ExtentOf(firstTokenAfterLBracket, lastComma),
nameof(ParserStrings.ArrayHasTooManyDimensions),
ParserStrings.ArrayHasTooManyDimensions,
arg: dim);
break;
}
if (token.Kind != TokenKind.RBracket)
{
// ErrorRecovery: just pretend we saw a ']'.
UngetToken(token);
ReportError(After(lastComma),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfAttribute),
ParserStrings.EndSquareBracketExpectedAtEndOfAttribute);
}
elementType = new ArrayTypeName(ExtentOf(elementType.Extent, token.Extent), elementType, dim);
break;
case TokenKind.RBracket:
elementType = new ArrayTypeName(ExtentOf(elementType.Extent, firstTokenAfterLBracket.Extent), elementType, 1);
break;
case TokenKind.EndOfInput:
UngetToken(firstTokenAfterLBracket);
ReportError(Before(firstTokenAfterLBracket),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfAttribute),
ParserStrings.EndSquareBracketExpectedAtEndOfAttribute);
break;
default:
// ErrorRecovery: sync to ']', and return null to avoid cascading errors.
ReportError(firstTokenAfterLBracket.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
firstTokenAfterLBracket.Text);
SyncOnError(true, TokenKind.RBracket);
break;
}
token = PeekToken();
// An array declared inside an unbracketed generic type argument cannot be assembly qualified
if (!unBracketedGenericArg && token.Kind == TokenKind.Comma)
{
SkipToken();
var assemblyName = _tokenizer.GetAssemblyNameSpec();
if (string.IsNullOrEmpty(assemblyName))
{
ReportError(After(token),
nameof(ParserStrings.MissingAssemblyNameSpecification),
ParserStrings.MissingAssemblyNameSpecification);
}
else
{
typeForAssemblyQualification.AssemblyName = assemblyName;
}
break;
}
if (token.Kind != TokenKind.LBracket)
{
break;
}
// Jagged array, skip the '[' and keep parsing.
SkipToken();
firstTokenAfterLBracket = NextToken();
}
return elementType;
}
private bool CompleteScriptBlockBody(Token lCurly, ref IScriptExtent bodyExtent, out IScriptExtent fullBodyExtent)
{
// If the caller passed in the open curly, then they expect us to consume the closing curly
// and include that in the extent.
if (lCurly != null)
{
Token rCurly = NextToken();
IScriptExtent endScriptBlock;
if (rCurly.Kind != TokenKind.RCurly)
{
// ErrorRecovery: pretend we saw the closing curly.
UngetToken(rCurly);
endScriptBlock = bodyExtent ?? lCurly.Extent;
ReportIncompleteInput(lCurly.Extent,
rCurly.Extent,
nameof(ParserStrings.MissingEndCurlyBrace),
ParserStrings.MissingEndCurlyBrace);
}
else
{
endScriptBlock = rCurly.Extent;
// If body was empty, use the extent b/w (but not including) the curlies, but only
// if that region has at least 1 character.
if (bodyExtent == null && (lCurly.Extent.EndColumnNumber != rCurly.Extent.StartColumnNumber))
{
bodyExtent = ExtentOf(After(lCurly), Before(rCurly));
}
}
fullBodyExtent = ExtentOf(lCurly, endScriptBlock);
}
else
{
fullBodyExtent = _tokenizer.GetScriptExtent();
Token token = NextToken();
if (token.Kind != TokenKind.EndOfInput)
{
// ErrorRecovery: eat the unexpected token, and continue parsing to find more
// of the script.
ReportError(token.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
token.Text);
return false;
}
}
return true;
}
private ScriptBlockAst ScriptBlockBodyRule(Token lCurly, List<UsingStatementAst> usingStatements, ParamBlockAst paramBlockAst, bool isFilter, StatementAst predefinedStatementAst)
{
// G script-block-body:
// G named-block-list
// G statement-list
Token token = PeekToken();
if ((token.TokenFlags & TokenFlags.ScriptBlockBlockName) == TokenFlags.ScriptBlockBlockName)
{
return NamedBlockListRule(lCurly, usingStatements, paramBlockAst);
}
List<TrapStatementAst> traps = new List<TrapStatementAst>();
List<StatementAst> statements = new List<StatementAst>();
if (predefinedStatementAst != null)
{
statements.Add(predefinedStatementAst);
}
IScriptExtent statementListExtent = paramBlockAst?.Extent;
IScriptExtent scriptBlockExtent;
while (true)
{
IScriptExtent extent = StatementListRule(statements, traps);
if (statementListExtent == null)
{
statementListExtent = extent;
}
else if (extent != null)
{
statementListExtent = ExtentOf(statementListExtent, extent);
}
if (CompleteScriptBlockBody(lCurly, ref statementListExtent, out scriptBlockExtent))
{
break;
}
}
return new ScriptBlockAst(scriptBlockExtent, usingStatements, paramBlockAst,
new StatementBlockAst(statementListExtent ?? PositionUtilities.EmptyExtent, statements, traps), isFilter);
}
private ScriptBlockAst NamedBlockListRule(Token lCurly, List<UsingStatementAst> usingStatements, ParamBlockAst paramBlockAst)
{
// G named-block-list:
// G named-block
// G named-block-list named-block
// G named-block:
// G statement-terminators:opt block-name statement-block
// G block-name: one of
// G 'dynamicparam' 'begin' 'process' 'end'
NamedBlockAst dynamicParamBlock = null;
NamedBlockAst beginBlock = null;
NamedBlockAst processBlock = null;
NamedBlockAst endBlock = null;
NamedBlockAst cleanBlock = null;
IScriptExtent startExtent = lCurly?.Extent ?? paramBlockAst?.Extent;
IScriptExtent endExtent = null;
IScriptExtent extent = null;
IScriptExtent scriptBlockExtent = null;
while (true)
{
Token blockNameToken = NextToken();
switch (blockNameToken.Kind)
{
default:
// Next token is unexpected.
// ErrorRecovery: if 'lCurly' is present, pretend we saw a closing curly; otherwise, eat the unexpected token.
if (lCurly != null)
{
UngetToken(blockNameToken);
scriptBlockExtent = ExtentOf(startExtent, endExtent);
}
else
{
// If "lCurly == null", then it's a ps1/psm1 file, and thus the extent is the whole file.
scriptBlockExtent = _tokenizer.GetScriptExtent();
}
// Report error about the unexpected token.
ReportError(blockNameToken.Extent,
nameof(ParserStrings.MissingNamedBlocks),
ParserStrings.MissingNamedBlocks,
blockNameToken.Text);
goto return_script_block_ast;
case TokenKind.RCurly:
case TokenKind.EndOfInput:
// If the next token is RCurly or <eof>, handle it in 'CompleteScriptBlockBody'.
UngetToken(blockNameToken);
extent = ExtentOf(startExtent, endExtent);
goto finished_named_block_list;
case TokenKind.Dynamicparam:
case TokenKind.Begin:
case TokenKind.Process:
case TokenKind.End:
case TokenKind.Clean:
break;
}
startExtent ??= blockNameToken.Extent;
endExtent = blockNameToken.Extent;
StatementBlockAst statementBlock = StatementBlockRule();
if (statementBlock == null)
{
// ErrorRecovery: Eat the block name and keep going, there might be a valid block next.
ReportIncompleteInput(After(blockNameToken.Extent),
nameof(ParserStrings.MissingNamedStatementBlock),
ParserStrings.MissingNamedStatementBlock,
blockNameToken.Kind.Text());
statementBlock = new StatementBlockAst(blockNameToken.Extent, Array.Empty<StatementAst>(), null);
}
else
{
endExtent = statementBlock.Extent;
}
extent = ExtentOf(blockNameToken, endExtent);
if (blockNameToken.Kind == TokenKind.Begin && beginBlock == null)
{
beginBlock = new NamedBlockAst(extent, TokenKind.Begin, statementBlock, false);
}
else if (blockNameToken.Kind == TokenKind.Process && processBlock == null)
{
processBlock = new NamedBlockAst(extent, TokenKind.Process, statementBlock, false);
}
else if (blockNameToken.Kind == TokenKind.End && endBlock == null)
{
endBlock = new NamedBlockAst(extent, TokenKind.End, statementBlock, false);
}
else if (blockNameToken.Kind == TokenKind.Clean && cleanBlock == null)
{
cleanBlock = new NamedBlockAst(extent, TokenKind.Clean, statementBlock, false);
}
else if (blockNameToken.Kind == TokenKind.Dynamicparam && dynamicParamBlock == null)
{
dynamicParamBlock = new NamedBlockAst(extent, TokenKind.Dynamicparam, statementBlock, false);
}
else
{
// ErrorRecovery: this is a semantic error, we can keep parsing w/o trouble.
ReportError(extent,
nameof(ParserStrings.DuplicateScriptCommandClause),
ParserStrings.DuplicateScriptCommandClause,
blockNameToken.Kind.Text());
}
SkipNewlinesAndSemicolons();
}
finished_named_block_list:
CompleteScriptBlockBody(lCurly, ref extent, out scriptBlockExtent);
return_script_block_ast:
return new ScriptBlockAst(
scriptBlockExtent,
usingStatements,
paramBlockAst,
beginBlock,
processBlock,
endBlock,
cleanBlock,
dynamicParamBlock);
}
private StatementBlockAst StatementBlockRule()
{
// G statement-block:
// G new-lines:opt '{' statement-list:opt new-lines:opt '}'
SkipNewlines();
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
UngetToken(lCurly);
return null;
}
List<TrapStatementAst> traps = new List<TrapStatementAst>();
List<StatementAst> statements = new List<StatementAst>();
IScriptExtent statementListExtent = StatementListRule(statements, traps);
Token rCurly = NextToken();
IScriptExtent endBlock;
if (rCurly.Kind != TokenKind.RCurly)
{
// ErrorRecovery: Pretend we saw the missing curly and keep parsing.
UngetToken(rCurly);
endBlock = statementListExtent ?? lCurly.Extent;
ReportIncompleteInput(lCurly.Extent,
rCurly.Extent,
nameof(ParserStrings.MissingEndCurlyBrace),
ParserStrings.MissingEndCurlyBrace);
}
else
{
endBlock = rCurly.Extent;
}
return new StatementBlockAst(ExtentOf(lCurly, endBlock), statements, traps);
}
private IScriptExtent StatementListRule(List<StatementAst> statements, List<TrapStatementAst> traps)
{
// G statement-list:
// G statement
// G statement-list statement
StatementAst firstStatement = null;
StatementAst lastStatement = null;
SkipNewlinesAndSemicolons();
while (true)
{
StatementAst statement = StatementRule();
if (statement == null)
{
break;
}
_tokenizer.CheckAstIsBeforeSignature(statement);
var trapStatementAst = statement as TrapStatementAst;
if (trapStatementAst != null)
{
traps.Add(trapStatementAst);
}
else
{
statements.Add(statement);
}
// Track the last statement inside our loop so we don't use the EmptyPipeline
// as our last statement. The last statement is used to track the extent of
// this statement list.
firstStatement ??= statement;
lastStatement = statement;
SkipNewlinesAndSemicolons();
Token token = PeekToken();
if (token.Kind == TokenKind.RParen || token.Kind == TokenKind.RCurly)
{
break;
}
}
return (firstStatement == null) ? null : ExtentOf(firstStatement, lastStatement);
}
/// <summary>
/// Parse a single statement.
/// </summary>
/// <returns>A statement ast. Never returns null, always returns PipelineAst.EmptyPipeline if there was no statement.</returns>
private StatementAst StatementRule()
{
// G statement:
// G if-statement
// G label:opt labeled-statement
// G function-statement
// G flow-control-statement statement-terminator
// G trap-statement
// G try-statement
// G data-statement
// G pipeline-chain statement-terminator
// G
// G labeled-statement:
// G switch-statement
// G foreach-statement
// G for-statement
// G while-statement
// G do-statement
// G
// G flow-control-statement:
// G 'break' label-expression:opt
// G 'continue' label-expression:opt
// G 'throw' pipeline:opt
// G 'return' pipeline:opt
// G 'exit' pipeline:opt
// G
// G statement-terminator:
// G ';'
// G new-line-character
RuntimeHelpers.EnsureSufficientExecutionStack();
int restorePoint = 0;
StatementAst statement;
Token token = NextToken();
List<AttributeBaseAst> attributes = null;
if (token.Kind == TokenKind.Generic && token.Text[0] == '[')
{
restorePoint = token.Extent.StartOffset;
Resync(token);
attributes = AttributeListRule(false);
token = NextToken();
if (attributes != null
&& attributes.Count > 0)
{
if ((token.TokenFlags & TokenFlags.StatementDoesntSupportAttributes) != 0)
{
if (attributes.OfType<TypeConstraintAst>().Any())
{
Resync(restorePoint);
token = NextToken();
}
else
{
ReportError(attributes[0].Extent,
nameof(ParserStrings.UnexpectedAttribute),
ParserStrings.UnexpectedAttribute,
attributes[0].TypeName.FullName);
}
}
else if ((token.TokenFlags & TokenFlags.Keyword) != 0)
{
foreach (var attr in attributes.Where(static attr => attr is not AttributeAst))
{
ReportError(attr.Extent,
nameof(ParserStrings.TypeNotAllowedBeforeStatement),
ParserStrings.TypeNotAllowedBeforeStatement,
attr.TypeName.FullName);
break;
}
}
else
{
Resync(restorePoint);
token = NextToken();
}
}
}
switch (token.Kind)
{
case TokenKind.If:
statement = IfStatementRule(token);
break;
case TokenKind.Switch:
statement = SwitchStatementRule(null, token);
break;
case TokenKind.Foreach:
statement = ForeachStatementRule(null, token);
break;
case TokenKind.For:
statement = ForStatementRule(null, token);
break;
case TokenKind.While:
statement = WhileStatementRule(null, token);
break;
case TokenKind.Do:
statement = DoWhileStatementRule(null, token);
break;
case TokenKind.Function:
case TokenKind.Filter:
case TokenKind.Workflow:
statement = FunctionDeclarationRule(token);
break;
case TokenKind.Return:
statement = ReturnStatementRule(token);
break;
case TokenKind.Throw:
statement = ThrowStatementRule(token);
break;
case TokenKind.Exit:
statement = ExitStatementRule(token);
break;
case TokenKind.Break:
statement = BreakStatementRule(token);
break;
case TokenKind.Continue:
statement = ContinueStatementRule(token);
break;
case TokenKind.Trap:
statement = TrapStatementRule(token);
break;
case TokenKind.Try:
statement = TryStatementRule(token);
break;
case TokenKind.Data:
statement = DataStatementRule(token);
break;
case TokenKind.Parallel:
case TokenKind.Sequence:
statement = BlockStatementRule(token);
break;
case TokenKind.Configuration:
statement = ConfigurationStatementRule(attributes?.OfType<AttributeAst>(), token);
break;
case TokenKind.From:
case TokenKind.Define:
case TokenKind.Var:
ReportError(token.Extent,
nameof(ParserStrings.ReservedKeywordNotAllowed),
ParserStrings.ReservedKeywordNotAllowed,
token.Kind.Text());
statement = new ErrorStatementAst(token.Extent);
break;
case TokenKind.Label:
SkipNewlines();
statement = LabeledStatementRule((LabelToken)token);
break;
case TokenKind.EndOfInput:
if (attributes != null)
{
Resync(restorePoint);
statement = PipelineChainRule();
}
else
{
UngetToken(token);
statement = null;
}
break;
case TokenKind.Else:
case TokenKind.ElseIf:
case TokenKind.Catch:
case TokenKind.Until:
if (ErrorList.Count > 0)
{
// If we have already seen an error, just eat these tokens. By eating the token, we won't
// generate an odd pipeline with the keyword as a command name, which should provider a better
// user experience (e.g., better syntax coloring.)
SkipNewlines();
return StatementRule();
}
goto default;
case TokenKind.DynamicKeyword:
DynamicKeyword keywordData = DynamicKeyword.GetKeyword(token.Text);
statement = DynamicKeywordStatementRule(token, keywordData);
break;
case TokenKind.Class:
statement = ClassDefinitionRule(attributes, token);
break;
case TokenKind.Enum:
statement = EnumDefinitionRule(attributes, token);
break;
case TokenKind.Using:
statement = UsingStatementRule(token);
// Report an error - usings must appear before anything else in the script, but parse it anyway
ReportError(statement.Extent,
nameof(ParserStrings.UsingMustBeAtStartOfScript),
ParserStrings.UsingMustBeAtStartOfScript);
break;
default:
if (attributes != null)
{
Resync(restorePoint);
}
else
{
UngetToken(token);
}
statement = PipelineChainRule();
break;
}
return statement;
}
private StringConstantExpressionAst SimpleNameRule()
{
Token token;
return SimpleNameRule(out token);
}
private StringConstantExpressionAst SimpleNameRule(out Token token)
{
try
{
_tokenizer.WantSimpleName = true;
token = PeekToken();
}
finally
{
_tokenizer.WantSimpleName = false;
}
if (token.Kind == TokenKind.Identifier)
{
// We mark the token as a member name even though it may be a loop name or data variable name.
// V2 did this, so it's at least backwards compatible.
token.TokenFlags |= TokenFlags.MemberName;
SkipToken();
return new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord);
}
return null;
}
private ExpressionAst LabelOrKeyRule()
{
// G label-expression:
// G simple-name
// G unary-expression
// G key-expression:
// G simple-name
// G unary-expression
var simpleName = SimpleNameRule();
if (simpleName != null)
{
return simpleName;
}
Token token = PeekToken();
if (token.Kind != TokenKind.NewLine && token.Kind != TokenKind.Semi)
{
ExpressionAst labelExpr;
bool disableCommaOperator = _disableCommaOperator;
try
{
_disableCommaOperator = true;
labelExpr = UnaryExpressionRule();
}
finally
{
_disableCommaOperator = disableCommaOperator;
}
if (labelExpr != null)
{
return labelExpr;
}
// If this is a label, then any token other than EOF here is effectively unreachable code. E.g.
// break & "cmd"
// Gets parsed as:
// break ; & "cmd"
}
return null;
}
private BreakStatementAst BreakStatementRule(Token breakToken)
{
// G flow-control-statement:
// G 'break' label-expression:opt
ExpressionAst labelExpr = LabelOrKeyRule();
IScriptExtent extent = (labelExpr != null)
? ExtentOf(breakToken, labelExpr)
: breakToken.Extent;
return new BreakStatementAst(extent, labelExpr);
}
private ContinueStatementAst ContinueStatementRule(Token continueToken)
{
// G flow-control-statement:
// G 'continue' label-expression:opt
ExpressionAst labelExpr = LabelOrKeyRule();
IScriptExtent extent = (labelExpr != null)
? ExtentOf(continueToken, labelExpr)
: continueToken.Extent;
return new ContinueStatementAst(extent, labelExpr);
}
private ReturnStatementAst ReturnStatementRule(Token token)
{
// G flow-control-statement:
// G 'return' pipeline-chain:opt
PipelineBaseAst pipeline = PipelineChainRule();
IScriptExtent extent = (pipeline != null)
? ExtentOf(token, pipeline)
: token.Extent;
return new ReturnStatementAst(extent, pipeline);
}
private ExitStatementAst ExitStatementRule(Token token)
{
// G flow-control-statement:
// G 'exit' pipeline-chain:opt
PipelineBaseAst pipeline = PipelineChainRule();
IScriptExtent extent = (pipeline != null)
? ExtentOf(token, pipeline)
: token.Extent;
return new ExitStatementAst(extent, pipeline);
}
private ThrowStatementAst ThrowStatementRule(Token token)
{
// G flow-control-statement:
// G 'throw' pipeline:opt
PipelineBaseAst pipeline = PipelineChainRule();
IScriptExtent extent = (pipeline != null)
? ExtentOf(token, pipeline)
: token.Extent;
return new ThrowStatementAst(extent, pipeline);
}
private StatementAst LabeledStatementRule(LabelToken label)
{
// G statement:
// G label:opt labeled-statement
// G
// G labeled-statement:
// G switch-statement
// G foreach-statement
// G for-statement
// G while-statement
// G do-statement
// G pipeline-chain
StatementAst statement;
Token token = NextToken();
switch (token.Kind)
{
case TokenKind.Switch:
statement = SwitchStatementRule(label, token);
break;
case TokenKind.Foreach:
statement = ForeachStatementRule(label, token);
break;
case TokenKind.For:
statement = ForStatementRule(label, token);
break;
case TokenKind.While:
statement = WhileStatementRule(label, token);
break;
case TokenKind.Do:
statement = DoWhileStatementRule(label, token);
break;
default:
// We can only unget 1 token, but have 2 to unget, so resync on the label.
Resync(label);
statement = PipelineChainRule();
break;
}
return statement;
}
private StatementAst BlockStatementRule(Token kindToken)
{
// G block-statement
// G keyword statement-block
StatementBlockAst body = StatementBlockRule();
// ErrorRecovery: nothing more to look for, so just return the error statement.
if (body == null)
{
ReportIncompleteInput(After(kindToken.Extent),
nameof(ParserStrings.MissingStatementAfterKeyword),
ParserStrings.MissingStatementAfterKeyword,
kindToken.Text);
return new ErrorStatementAst(ExtentOf(kindToken, kindToken));
}
return new BlockStatementAst(ExtentOf(kindToken, body), kindToken, body);
}
/// <summary>
/// Handle the InlineScript syntax in the script workflow.
/// </summary>
/// <param name="inlineScriptToken"></param>
/// <param name="elements"></param>
/// <returns>
/// true -- InlineScript parsing successful
/// false -- InlineScript parsing unsuccessful
/// </returns>
private bool InlineScriptRule(Token inlineScriptToken, List<CommandElementAst> elements)
{
// G Command
// G InlineScript scriptblock-expression
Diagnostics.Assert(elements != null && elements.Count == 0, "The CommandElement list should be empty");
var commandName = new StringConstantExpressionAst(inlineScriptToken.Extent, inlineScriptToken.Text, StringConstantType.BareWord);
inlineScriptToken.TokenFlags |= TokenFlags.CommandName;
elements.Add(commandName);
SkipNewlines();
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: If there is no opening curly, assume it hasn't been entered yet and don't consume anything.
UngetToken(lCurly);
ReportIncompleteInput(After(inlineScriptToken),
nameof(ParserStrings.MissingStatementAfterKeyword),
ParserStrings.MissingStatementAfterKeyword,
inlineScriptToken.Text);
return false;
}
var expr = ScriptBlockExpressionRule(lCurly);
elements.Add(expr);
return true;
}
private StatementAst IfStatementRule(Token ifToken)
{
// G if-statement:
// G 'if' new-lines:opt '(' pipeline-chain ')' statement-block elseif-clauses:opt else-clause:opt
// G elseif-clauses:
// G elseif-clause
// G elseif-clauses elseif-clause
// G elseif-clause:
// G 'elseif' new-lines:opt '(' pipeline-chain ')' statement-block
// G else-clause:
// G 'else' statement-block
List<IfClause> clauses = new List<IfClause>();
List<Ast> componentAsts = new List<Ast>();
StatementBlockAst elseClause = null;
Token keyword = ifToken;
while (true)
{
SkipNewlines();
Token lParen = NextToken();
if (lParen.Kind != TokenKind.LParen)
{
// ErrorRecovery: assume user just typed 'if' or 'elseif' and hadn't started typing anything
// else yet. Next token is likely a newline, so just put it back and keep parsing.
UngetToken(lParen);
ReportIncompleteInput(After(keyword),
nameof(ParserStrings.MissingOpenParenthesisInIfStatement),
ParserStrings.MissingOpenParenthesisInIfStatement,
keyword.Text);
return new ErrorStatementAst(ExtentOf(ifToken, keyword), componentAsts);
}
SkipNewlines();
PipelineBaseAst condition = PipelineChainRule();
if (condition == null)
{
// ErrorRecovery: assume pipeline just hasn't been entered yet, continue hoping
// to find a close paren and statement block.
IScriptExtent errorPosition = After(lParen);
ReportIncompleteInput(
errorPosition,
nameof(ParserStrings.IfStatementMissingCondition),
ParserStrings.IfStatementMissingCondition,
keyword.Text);
condition = new ErrorStatementAst(errorPosition);
}
else
{
componentAsts.Add(condition);
}
SkipNewlines();
Token rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: assume the next token is a newline or part of something else,
// so stop parsing the statement and try parsing something else if possible.
UngetToken(rParen);
// Don't bother reporting this error if we already reported an empty condition error.
if (condition is not ErrorStatementAst)
{
ReportIncompleteInput(rParen.Extent,
nameof(ParserStrings.MissingEndParenthesisAfterStatement),
ParserStrings.MissingEndParenthesisAfterStatement,
keyword.Text);
}
return new ErrorStatementAst(ExtentOf(ifToken, Before(rParen)), componentAsts);
}
SkipNewlines();
StatementBlockAst body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: assume the next token is a newline or part of something else,
// so stop parsing the statement and try parsing something else.
ReportIncompleteInput(rParen.Extent,
nameof(ParserStrings.MissingStatementBlock),
ParserStrings.MissingStatementBlock,
keyword.Text);
return new ErrorStatementAst(ExtentOf(ifToken, rParen), componentAsts);
}
componentAsts.Add(body);
clauses.Add(new IfClause(condition, body));
// Save a restore point here. In case there is no 'elseif' or 'else' following,
// we should resync back here to preserve the possible new lines. The new lines
// could be important for the following parsing. For example, in case we are in
// a HashExpression, a new line might be needed for parsing the key-value that
// is following the if statement:
// @{
// a = if (1) {}
// b = 10
// }
int restorePoint = _ungotToken == null ? _tokenizer.GetRestorePoint() : _ungotToken.Extent.StartOffset;
SkipNewlines();
keyword = PeekToken();
if (keyword.Kind == TokenKind.ElseIf)
{
SkipToken();
continue;
}
else if (keyword.Kind == TokenKind.Else)
{
SkipToken();
SkipNewlines();
elseClause = StatementBlockRule();
if (elseClause == null)
{
// ErrorRecovery: assume the next token is a newline or part of something else,
// so stop parsing the statement and try parsing something else.
ReportIncompleteInput(After(keyword),
nameof(ParserStrings.MissingStatementBlockAfterElse),
ParserStrings.MissingStatementBlockAfterElse);
return new ErrorStatementAst(ExtentOf(ifToken, keyword), componentAsts);
}
}
else
{
// There is no 'elseif' or 'else' following, so resync back to the possible new lines.
Resync(restorePoint);
}
break;
}
IScriptExtent endExtent = (elseClause != null)
? elseClause.Extent
: clauses[clauses.Count - 1].Item2.Extent;
IScriptExtent extent = ExtentOf(ifToken, endExtent);
return new IfStatementAst(extent, clauses, elseClause);
}
private StatementAst SwitchStatementRule(LabelToken labelToken, Token switchToken)
{
// G switch-statement:
// G 'switch' new-lines:opt switch-parameters:opt switch-condition switch-body
// G switch-parameters:
// G switch-parameter
// G switch-parameters switch-parameter
// G switch-parameter:
// G '-regex'
// G '-wildcard'
// G '-exact'
// G '-casesensitive'
// G '-parallel'
// G switch-condition:
// G '(' new-lines:opt pipeline new-lines:opt ')'
// G -file new-lines:opt switch-filename
// G switch-filename:
// G command-argument
// G primary-expression
// G switch-body:
// G new-lines:opt '{' new-lines:opt switch-clauses '}'
// G switch-clauses:
// G switch-clause
// G switch-clauses switch-clause
// G switch-clause:
// G switch-clause-condition statement-block statement-terminators:opt
// G switch-clause-condition:
// G command-argument
// G primary-expression
IScriptExtent startExtent = (labelToken ?? switchToken).Extent;
IScriptExtent endErrorStatement = startExtent;
bool isError = false;
bool isIncompleteError = false;
SkipNewlines();
bool needErrorCondition = false; // Only used to track if we need to include (condition) ast for the error statement.
PipelineBaseAst condition = null;
Dictionary<string, Tuple<Token, Ast>> specifiedFlags = null; // Only used to track all flags specified for the error ast
Token switchParameterToken = PeekToken();
SwitchFlags flags = SwitchFlags.None;
while (switchParameterToken.Kind == TokenKind.Parameter)
{
SkipToken();
endErrorStatement = switchParameterToken.Extent;
specifiedFlags ??= new Dictionary<string, Tuple<Token, Ast>>();
if (IsSpecificParameter(switchParameterToken, "regex"))
{
flags |= SwitchFlags.Regex;
flags &= ~SwitchFlags.Wildcard;
if (!specifiedFlags.ContainsKey("regex"))
{
specifiedFlags.Add("regex", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else if (IsSpecificParameter(switchParameterToken, "wildcard"))
{
flags |= SwitchFlags.Wildcard;
flags &= ~SwitchFlags.Regex;
if (!specifiedFlags.ContainsKey("wildcard"))
{
specifiedFlags.Add("wildcard", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else if (IsSpecificParameter(switchParameterToken, "exact"))
{
flags &= ~SwitchFlags.Regex;
flags &= ~SwitchFlags.Wildcard;
if (!specifiedFlags.ContainsKey("exact"))
{
specifiedFlags.Add("exact", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else if (IsSpecificParameter(switchParameterToken, "casesensitive"))
{
flags |= SwitchFlags.CaseSensitive;
if (!specifiedFlags.ContainsKey("casesensitive"))
{
specifiedFlags.Add("casesensitive", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else if (IsSpecificParameter(switchParameterToken, "parallel"))
{
flags |= SwitchFlags.Parallel;
if (!specifiedFlags.ContainsKey("parallel"))
{
specifiedFlags.Add("parallel", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else if (IsSpecificParameter(switchParameterToken, "file"))
{
flags |= SwitchFlags.File;
SkipNewlines();
ExpressionAst fileNameExpr = GetSingleCommandArgument(CommandArgumentContext.FileName);
if (fileNameExpr == null)
{
// ErrorRecovery: pretend we saw the filename and continue.
isError = true;
isIncompleteError = ReportIncompleteInput(After(switchParameterToken),
nameof(ParserStrings.MissingFilenameOption),
ParserStrings.MissingFilenameOption);
if (!specifiedFlags.ContainsKey("file"))
{
specifiedFlags.Add("file", new Tuple<Token, Ast>(switchParameterToken, null));
}
}
else
{
endErrorStatement = fileNameExpr.Extent;
condition = new PipelineAst(fileNameExpr.Extent,
new CommandExpressionAst(fileNameExpr.Extent, fileNameExpr, null), background: false);
if (!specifiedFlags.ContainsKey("file"))
{
specifiedFlags.Add("file", new Tuple<Token, Ast>(switchParameterToken, condition));
}
}
}
else
{
// ErrorRecovery: just ignore the token, continue parsing.
isError = true;
ReportError(switchParameterToken.Extent,
nameof(ParserStrings.InvalidSwitchFlag),
ParserStrings.InvalidSwitchFlag,
((ParameterToken)switchParameterToken).ParameterName);
}
switchParameterToken = PeekToken();
}
if (switchParameterToken.Kind == TokenKind.Minus)
{
specifiedFlags ??= new Dictionary<string, Tuple<Token, Ast>>();
specifiedFlags.Add(VERBATIM_ARGUMENT, new Tuple<Token, Ast>(switchParameterToken, null));
}
Token lParen = PeekToken();
if (lParen.Kind == TokenKind.LParen)
{
endErrorStatement = lParen.Extent;
SkipToken();
if ((flags & SwitchFlags.File) == SwitchFlags.File)
{
// ErrorRecovery: nothing special this is a semantic error.
isError = true;
ReportError(lParen.Extent,
nameof(ParserStrings.PipelineValueRequired),
ParserStrings.PipelineValueRequired);
}
needErrorCondition = true; // need to add condition ast to the error statement if the parsing fails
SkipNewlines();
condition = PipelineChainRule();
if (condition == null)
{
// ErrorRecovery: pretend we saw the condition and keep parsing.
isError = true;
isIncompleteError = ReportIncompleteInput(After(lParen),
nameof(ParserStrings.PipelineValueRequired),
ParserStrings.PipelineValueRequired);
}
else
{
endErrorStatement = condition.Extent;
}
SkipNewlines();
Token rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: Try to parse the switch body, if we don't find a body, then bail.
UngetToken(rParen);
if (!isIncompleteError)
{
isError = true;
isIncompleteError =
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingEndParenthesisInSwitchStatement),
ParserStrings.MissingEndParenthesisInSwitchStatement);
}
}
else
{
endErrorStatement = rParen.Extent;
}
}
else if (condition == null)
{
if ((flags & SwitchFlags.File) == 0)
{
isError = true;
isIncompleteError = ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.PipelineValueRequired),
ParserStrings.PipelineValueRequired);
}
else
{
Diagnostics.Assert(isError, "An error should already have been issued");
}
}
SkipNewlines();
Token lCurly = NextToken();
StatementBlockAst @default = null;
List<SwitchClause> clauses = new List<SwitchClause>();
List<Ast> errorAsts = new List<Ast>(); // in case there is an error, we want the asts parsed up to the error.
Token rCurly = null;
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: Assume we don't have any switch body to parse.
UngetToken(lCurly);
if (!isIncompleteError)
{
isError = true;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingCurlyBraceInSwitchStatement),
ParserStrings.MissingCurlyBraceInSwitchStatement);
}
}
else
{
endErrorStatement = lCurly.Extent;
SkipNewlines();
while (true)
{
Token token = PeekToken();
bool isDefaultClause = token.Kind == TokenKind.Default;
ExpressionAst clauseCondition = null;
if (isDefaultClause)
{
// Consume the 'default' token.
SkipToken();
clauseCondition = new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord);
}
else
{
clauseCondition = GetSingleCommandArgument(CommandArgumentContext.SwitchCondition);
if (clauseCondition == null)
{
// ErrorRecovery: if we don't have anything that looks like a condition, we won't
// find a body (because a body is just a script block, which works as a condition.)
// So don't look for a body, hope we find the '}' next.
isError = true;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingSwitchConditionExpression),
ParserStrings.MissingSwitchConditionExpression);
// Consume a closing curly, if there is one, to avoid an extra error
if (PeekToken().Kind == TokenKind.RCurly)
{
SkipToken();
}
break;
}
}
errorAsts.Add(clauseCondition);
endErrorStatement = clauseCondition.Extent;
StatementBlockAst clauseBody = StatementBlockRule();
if (clauseBody == null)
{
// ErrorRecovery: We might find another condition/body pair, so keep going.
isError = true;
isIncompleteError = ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingSwitchStatementClause),
ParserStrings.MissingSwitchStatementClause);
}
else
{
errorAsts.Add(clauseBody);
endErrorStatement = clauseBody.Extent;
if (isDefaultClause)
{
if (@default != null)
{
// ErrorRecovery: just report the error and continue, forget the previous default clause.
isError = true;
ReportError(clauseCondition.Extent,
nameof(ParserStrings.MultipleSwitchDefaultClauses),
ParserStrings.MultipleSwitchDefaultClauses);
}
@default = clauseBody;
}
else
{
clauses.Add(new SwitchClause(clauseCondition, clauseBody));
}
}
SkipNewlinesAndSemicolons();
token = PeekToken();
if (token.Kind == TokenKind.RCurly)
{
rCurly = token;
SkipToken();
break;
}
if (token.Kind == TokenKind.EndOfInput)
{
if (!isIncompleteError)
{
isError = true;
ReportIncompleteInput(lCurly.Extent,
token.Extent,
nameof(ParserStrings.MissingEndCurlyBrace),
ParserStrings.MissingEndCurlyBrace);
}
break;
}
}
}
if (isError)
{
return new ErrorStatementAst(ExtentOf(startExtent, endErrorStatement), switchToken, specifiedFlags,
needErrorCondition ? GetNestedErrorAsts(condition) : null, GetNestedErrorAsts(errorAsts));
}
return new SwitchStatementAst(ExtentOf(labelToken ?? switchToken, rCurly),
labelToken?.LabelText, condition, flags, clauses, @default);
}
private StatementAst ConfigurationStatementRule(IEnumerable<AttributeAst> customAttributes, Token configurationToken)
{
// G configuration-statement:
// G 'configuration' new-lines:opt singleNameExpression new-lines:opt statement-block
// G singleNameExpression:
// G command-argument
// G primary-expression
IScriptExtent startExtent = configurationToken.Extent;
IScriptExtent endErrorStatement = startExtent;
bool isError = false;
// The expression that returns the configuration name.
ExpressionAst configurationName;
string simpleConfigurationNameValue = null;
SkipNewlines();
Token configurationNameToken = NextToken();
Token configurationKeywordToken = configurationNameToken;
if (configurationNameToken.Kind == TokenKind.LCurly)
{
ReportError(After(startExtent),
nameof(ParserStrings.MissingConfigurationName),
ParserStrings.MissingConfigurationName);
// Try reading the configuration body - this should keep the parse in sync - but we won't return it
ScriptBlockExpressionRule(configurationNameToken);
return null;
}
if (configurationNameToken.Kind is TokenKind.EndOfInput or TokenKind.Comma)
{
UngetToken(configurationNameToken);
ReportIncompleteInput(After(configurationNameToken.Extent),
nameof(ParserStrings.MissingConfigurationName),
ParserStrings.MissingConfigurationName);
return null;
}
// Unget the configuration token so it can possibly be re-read as part of an expression
UngetToken(configurationNameToken);
// Finally read the name for this configuration
configurationName = GetWordOrExpression(configurationNameToken);
if (configurationName == null)
{
isError = true;
ReportIncompleteInput(configurationNameToken.Extent,
nameof(ParserStrings.MissingConfigurationName),
ParserStrings.MissingConfigurationName);
}
else
{
object outValue;
if (IsConstantValueVisitor.IsConstant(configurationName, out outValue))
{
simpleConfigurationNameValue = outValue as string;
if (simpleConfigurationNameValue == null ||
!System.Text.RegularExpressions.Regex.IsMatch(simpleConfigurationNameValue, "^[A-Za-z][A-Za-z0-9_./-]*$"))
{
// This is actually a semantics check, the syntax is fine at this point.
// Continue parsing to get as much information as possible
isError = true;
ReportError(configurationName.Extent,
nameof(ParserStrings.InvalidConfigurationName),
ParserStrings.InvalidConfigurationName,
simpleConfigurationNameValue ?? string.Empty);
}
}
}
SkipNewlines();
//
// Load the system classes and import them as keywords
//
Runspaces.Runspace localRunspace = null;
bool topLevel = false;
try
{
// At this point, we'll need a runspace to use to hold the metadata for the parse. If there is no
// current runspace to use, we create one and set it to be the default for this thread...
if (Runspaces.Runspace.DefaultRunspace == null)
{
localRunspace =
Runspaces.RunspaceFactory.CreateRunspace(Runspaces.InitialSessionState.CreateDefault2());
localRunspace.ThreadOptions = Runspaces.PSThreadOptions.UseCurrentThread;
localRunspace.Open();
Runspaces.Runspace.DefaultRunspace = localRunspace;
}
// Configuration is not supported in ConstrainedLanguage
if (Runspace.DefaultRunspace?.ExecutionContext?.LanguageMode == PSLanguageMode.ConstrainedLanguage)
{
if (SystemPolicy.GetSystemLockdownPolicy() != SystemEnforcementMode.Audit)
{
ReportError(configurationToken.Extent,
nameof(ParserStrings.ConfigurationNotAllowedInConstrainedLanguage),
ParserStrings.ConfigurationNotAllowedInConstrainedLanguage,
configurationToken.Kind.Text());
return null;
}
SystemPolicy.LogWDACAuditMessage(
context: Runspace.DefaultRunspace?.ExecutionContext,
title: ParserStrings.WDACParserConfigKeywordLogTitle,
message: ParserStrings.WDACParserConfigKeywordLogMessage,
fqid: "ConfigurationLanguageKeywordNotAllowed",
dropIntoDebugger: true);
}
// Configuration is not supported for ARM or ARM64 process architecture.
if (PsUtils.IsRunningOnProcessArchitectureARM())
{
ReportError(
configurationToken.Extent,
nameof(ParserStrings.ConfigurationNotAllowedOnArm64),
ParserStrings.ConfigurationNotAllowedOnArm64,
configurationToken.Kind.Text());
return null;
}
// Configuration is not supported on WinPE
if (Utils.IsWinPEHost())
{
ReportError(configurationToken.Extent,
nameof(ParserStrings.ConfigurationNotAllowedOnWinPE),
ParserStrings.ConfigurationNotAllowedOnWinPE,
configurationToken.Kind.Text());
return null;
}
ExpressionAst configurationBodyScriptBlock = null;
PowerShell p = null;
// Save the parser we're using so we can resume the current parse when we're done.
var currentParser = Runspaces.Runspace.DefaultRunspace.ExecutionContext.Engine.EngineParser;
Runspaces.Runspace.DefaultRunspace.ExecutionContext.Engine.EngineParser = new Parser();
try
{
if (localRunspace != null)
{
p = PowerShell.Create();
p.Runspace = localRunspace;
}
else
{
p = PowerShell.Create(RunspaceMode.CurrentRunspace);
}
try
{
// See of the default CIM keywords are already loaded. If they haven't been
// then this is the top level. Record that information and then load the defaults
// keywords.
if (DynamicKeyword.GetKeyword("OMI_ConfigurationDocument") == null)
{
// Load the default CIM keywords
Collection<Exception> CIMKeywordErrors = new Collection<Exception>();
// DscSubsystem is auto-registered when PSDesiredStateConfiguration v3 module is loaded
// so if DscSubsystem is registered that means user intention to use v3 APIs.
ICrossPlatformDsc dscSubsystem = SubsystemManager.GetSubsystem<ICrossPlatformDsc>();
if (dscSubsystem != null)
{
dscSubsystem.LoadDefaultKeywords(CIMKeywordErrors);
}
else
{
Dsc.DscClassCache.LoadDefaultCimKeywords(CIMKeywordErrors);
}
// Report any errors encountered while loading CIM dynamic keywords.
if (CIMKeywordErrors.Count > 0)
{
ReportErrorsAsWarnings(CIMKeywordErrors);
}
// Load any keywords that have been defined earlier in the script.
if (_configurationKeywordsDefinedInThisFile != null)
{
foreach (var kw in _configurationKeywordsDefinedInThisFile.Values)
{
if (!DynamicKeyword.ContainsKeyword(kw.Keyword))
{
DynamicKeyword.AddKeyword(kw);
}
}
}
topLevel = true;
}
}
catch (Exception e)
{
// This shouldn't happen - the system classes should always be good, but just in case,
// we'll catch the exception and report it as an error.
ReportError(configurationKeywordToken.Extent,
nameof(ParserStrings.ParserError),
ParserStrings.ParserError,
e.ToString());
return null;
}
}
finally
{
p?.Dispose();
//
// Put the parser back...
//
Runspaces.Runspace.DefaultRunspace.ExecutionContext.Engine.EngineParser = currentParser;
}
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
ReportIncompleteInput(After(lCurly.Extent),
nameof(ParserStrings.MissingCurlyInConfigurationStatement),
ParserStrings.MissingCurlyInConfigurationStatement);
isError = true;
UngetToken(lCurly);
}
else
{
var oldInConfiguration = _inConfiguration;
try
{
_inConfiguration = true;
configurationBodyScriptBlock = ScriptBlockExpressionRule(lCurly);
}
finally
{
_inConfiguration = oldInConfiguration;
}
if (configurationBodyScriptBlock == null)
{
ReportError(After(lCurly.Extent),
nameof(ParserStrings.ConfigurationBodyEmpty),
ParserStrings.ConfigurationBodyEmpty);
return null;
}
}
if (isError)
{
return new ErrorStatementAst(ExtentOf(startExtent, endErrorStatement), configurationToken);
}
#region "Add Configuration Keywords"
// If the configuration name is a constant string, then
// if we're not at the top level, we'll add it to the list of configuration resource keywords.
// If we are at the top level, then we'll add it to the list of keywords defined in this
// parse so it can be used as a resource in subsequent config statements.
var scAst = configurationName as StringConstantExpressionAst;
if (scAst != null)
{
var keywordToAddForThisConfigurationStatement = new System.Management.Automation.Language.DynamicKeyword
{
BodyMode = DynamicKeywordBodyMode.Hashtable,
ImplementingModule = _keywordModuleName,
Keyword = scAst.Value,
NameMode = DynamicKeywordNameMode.NameRequired,
DirectCall = true,
};
// Add the DependsOn property.
var dependsOnProp = new DynamicKeywordProperty
{
Mandatory = true,
Name = "DependsOn",
};
keywordToAddForThisConfigurationStatement.Properties.Add(dependsOnProp.Name, dependsOnProp);
// Add the PsDscRunAsCredential property.
var RunAsProp = new DynamicKeywordProperty
{
Mandatory = true,
Name = "PsDscRunAsCredential",
};
keywordToAddForThisConfigurationStatement.Properties.Add(RunAsProp.Name, RunAsProp);
// Extract the parameters, if any and them to the keyword definition.
var sbeAst = configurationBodyScriptBlock as ScriptBlockExpressionAst;
if (sbeAst != null)
{
var pList = sbeAst.ScriptBlock.ParamBlock;
if (pList != null)
{
foreach (var parm in pList.Parameters)
{
var keywordProp = new DynamicKeywordProperty();
keywordProp.Name = parm.Name.VariablePath.UserPath;
if (parm.Attributes != null)
{
foreach (var attr in parm.Attributes)
{
var typeConstraint = attr as TypeConstraintAst;
if (typeConstraint != null)
{
keywordProp.TypeConstraint = typeConstraint.TypeName.Name;
continue;
}
var aAst = attr as AttributeAst;
if (aAst != null)
{
if (string.Equals(aAst.TypeName.Name, "Parameter", StringComparison.OrdinalIgnoreCase))
{
if (aAst.NamedArguments != null)
{
foreach (var na in aAst.NamedArguments)
{
if (string.Equals(na.ArgumentName, "Mandatory", StringComparison.OrdinalIgnoreCase))
{
if (na.ExpressionOmitted)
{
keywordProp.Mandatory = true;
}
else if (na.Argument != null)
{
ConstantExpressionAst ceAst = na.Argument as ConstantExpressionAst;
if (ceAst != null)
{
keywordProp.Mandatory = System.Management.Automation.LanguagePrimitives.IsTrue(ceAst.Value);
}
}
}
}
}
}
}
}
}
keywordToAddForThisConfigurationStatement.Properties.Add(keywordProp.Name, keywordProp);
}
}
}
if (topLevel)
{
_configurationKeywordsDefinedInThisFile ??= new Dictionary<string, DynamicKeyword>();
_configurationKeywordsDefinedInThisFile[keywordToAddForThisConfigurationStatement.Keyword] = keywordToAddForThisConfigurationStatement;
}
else
{
System.Management.Automation.Language.DynamicKeyword.AddKeyword(keywordToAddForThisConfigurationStatement);
}
}
#endregion
// End of dynamic keyword definition for this function...
//############################################################
//############################################################
bool isMetaConfiguration = false;
if (customAttributes != null)
{
isMetaConfiguration = customAttributes.Any(
attribute => (attribute.TypeName.GetReflectionAttributeType() != null) &&
(attribute.TypeName.GetReflectionAttributeType() == typeof(DscLocalConfigurationManagerAttribute)));
}
ScriptBlockExpressionAst bodyAst = configurationBodyScriptBlock as ScriptBlockExpressionAst;
IScriptExtent configurationExtent = ExtentOf(startExtent, bodyAst);
return new ConfigurationDefinitionAst(configurationExtent,
bodyAst,
isMetaConfiguration ? ConfigurationType.Meta : ConfigurationType.Resource,
configurationName)
{
LCurlyToken = lCurly,
ConfigurationToken = configurationToken,
CustomAttributes = customAttributes,
DefinedKeywords = DynamicKeyword.GetKeyword()
};
}
catch (Exception e)
{
// In theory this should never happen so if it does, we'll report the actual exception rather than introducing a new message
ReportError(configurationKeywordToken.Extent,
nameof(ParserStrings.ParserError),
ParserStrings.ParserError,
"ConfigurationStatementToken: " + e);
return null;
}
finally
{
// If we had to allocate a runspace for the parser, we'll free it now.
if (localRunspace != null)
{
// If a runspace was created for this operation, close it and reset the default runspace to null.
localRunspace.Close();
Runspaces.Runspace.DefaultRunspace = null;
}
if (topLevel)
{
//
// Clear out all of the cached classes and keywords.
// They will need to be reloaded when the generated function is actually run.
//
ICrossPlatformDsc dscSubsystem = SubsystemManager.GetSubsystem<ICrossPlatformDsc>();
if (dscSubsystem != null)
{
dscSubsystem.ClearCache();
}
else
{
Dsc.DscClassCache.ClearCache();
}
System.Management.Automation.Language.DynamicKeyword.Reset();
}
// Finally resync the tokenizer at the current position
// this will flush any cached dynamic keyword tokens.
var restorePoint = _tokenizer.GetRestorePoint();
Resync(restorePoint);
}
}
private Dictionary<string, DynamicKeyword> _configurationKeywordsDefinedInThisFile;
/// <summary>
/// Reads an argument expression for a keyword or keyword parameter.
/// This can be either a bare word or an expression.
/// </summary>
/// <param name="keywordToken">The token of the associated keyword.</param>
private ExpressionAst GetWordOrExpression(Token keywordToken)
{
Token nameToken = NextToken();
if (nameToken.Kind == TokenKind.EndOfInput)
{
// ErrorRecovery: report incomplete statement and return
UngetToken(nameToken);
ReportIncompleteInput(After(keywordToken),
nameof(ParserStrings.RequiredNameOrExpressionMissing),
ParserStrings.RequiredNameOrExpressionMissing);
return null;
}
ExpressionAst argument = GetCommandArgument(CommandArgumentContext.CommandArgument, nameToken);
if (argument == null)
{
var extent = keywordToken.Extent;
ReportError(After(extent),
nameof(ParserStrings.ParameterRequiresArgument),
ParserStrings.ParameterRequiresArgument,
keywordToken.Text);
}
return argument;
}
private StatementAst ForeachStatementRule(LabelToken labelToken, Token forEachToken)
{
// G foreach-statement:
// G 'foreach' new-lines:opt foreach-parameters:opt new-lines:opt '('
// G new-lines:opt variable new-lines:opt 'in' new-lines:opt pipeline
// G new-lines:opt ')' statement-block
// G foreach-parameters:
// G foreach-parameter
// G foreach-parameters foreach-parameter
// G foreach-parameter:
// G '-parallel'
// G '-throttlelimit' new-lines:opt foreach-throttlelimit
// G foreach-throttlelimit:
// G command-argument
// G primary-expression
IScriptExtent startOfStatement = labelToken != null ? labelToken.Extent : forEachToken.Extent;
IScriptExtent endErrorStatement = null;
SkipNewlines();
// Process parameters on foreach
Token foreachParameterToken = PeekToken();
ForEachFlags flags = ForEachFlags.None;
ExpressionAst throttleLimit = null;
while (foreachParameterToken.Kind == TokenKind.Parameter)
{
SkipToken();
if (IsSpecificParameter(foreachParameterToken, "parallel"))
{
flags |= ForEachFlags.Parallel;
}
else if (IsSpecificParameter(foreachParameterToken, "throttlelimit"))
{
SkipNewlines();
throttleLimit = GetSingleCommandArgument(CommandArgumentContext.CommandArgument);
if (throttleLimit == null)
{
// ErrorRecovery: pretend we saw the throttle limit and continue.
ReportIncompleteInput(After(foreachParameterToken),
nameof(ParserStrings.MissingThrottleLimit),
ParserStrings.MissingThrottleLimit);
}
}
else
{
// ErrorRecovery: just ignore the token, continue parsing.
endErrorStatement = foreachParameterToken.Extent;
ReportError(foreachParameterToken.Extent,
nameof(ParserStrings.InvalidForeachFlag),
ParserStrings.InvalidForeachFlag,
((ParameterToken)foreachParameterToken).ParameterName);
}
SkipNewlines();
foreachParameterToken = PeekToken();
}
Token lParen = NextToken();
if (lParen.Kind != TokenKind.LParen)
{
// ErrorRecovery: assume the rest of the statement is missing.
UngetToken(lParen);
endErrorStatement = forEachToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingOpenParenthesisAfterKeyword),
ParserStrings.MissingOpenParenthesisAfterKeyword,
forEachToken.Kind.Text());
return new ErrorStatementAst(ExtentOf(startOfStatement, endErrorStatement));
}
SkipNewlines();
Token token = NextToken();
if (token.Kind != TokenKind.Variable && token.Kind != TokenKind.SplattedVariable)
{
// ErrorRecovery: assume the rest of the statement is missing.
UngetToken(token);
endErrorStatement = lParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingVariableNameAfterForeach),
ParserStrings.MissingVariableNameAfterForeach);
return new ErrorStatementAst(ExtentOf(startOfStatement, endErrorStatement));
}
var variableAst = new VariableExpressionAst((VariableToken)token);
SkipNewlines();
PipelineBaseAst pipeline = null;
StatementBlockAst body = null;
Token inToken = NextToken();
if (inToken.Kind != TokenKind.In)
{
// ErrorRecovery: assume the rest of the statement is missing.
UngetToken(inToken);
endErrorStatement = variableAst.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingInInForeach),
ParserStrings.MissingInInForeach);
}
else
{
SkipNewlines();
pipeline = PipelineChainRule();
if (pipeline == null)
{
// ErrorRecovery: assume the rest of the statement is missing.
endErrorStatement = inToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingForeachExpression),
ParserStrings.MissingForeachExpression);
}
else
{
SkipNewlines();
Token rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: assume the rest of the statement is missing.
UngetToken(rParen);
endErrorStatement = pipeline.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingEndParenthesisAfterForeach),
ParserStrings.MissingEndParenthesisAfterForeach);
}
else
{
body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: nothing more to look for, so just return the error statement.
endErrorStatement = rParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingForeachStatement),
ParserStrings.MissingForeachStatement);
}
}
}
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(startOfStatement, endErrorStatement),
GetNestedErrorAsts(variableAst, pipeline, body));
}
return new ForEachStatementAst(ExtentOf(startOfStatement, body),
labelToken?.LabelText,
flags,
throttleLimit, variableAst, pipeline, body);
}
private StatementAst ForStatementRule(LabelToken labelToken, Token forToken)
{
// G for-statement:
// G 'for' new-lines:opt '('
// G new-lines:opt for-initializer:opt statement-terminator
// G new-lines:opt for-condition:opt statement-terminator
// G new-lines:opt for-iterator:opt
// G new-lines:opt ')' statement-block
// G 'for' new-lines:opt '('
// G new-lines:opt for-initializer:opt statement-terminator
// G new-lines:opt for-condition:opt
// G new-lines:opt ')' statement-block
// G 'for' new-lines:opt '('
// G new-lines:opt for-initializer:opt
// G new-lines:opt ')' statement-block
// G for-initializer:
// G pipeline-chain
// G for-condition:
// G pipeline-chain
// G for-iterator:
// G pipeline-chain
IScriptExtent endErrorStatement = null;
SkipNewlines();
Token lParen = NextToken();
if (lParen.Kind != TokenKind.LParen)
{
// ErrorRecovery: don't continue parsing the for statement.
UngetToken(lParen);
endErrorStatement = forToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingOpenParenthesisAfterKeyword),
ParserStrings.MissingOpenParenthesisAfterKeyword,
forToken.Kind.Text());
return new ErrorStatementAst(ExtentOf(labelToken ?? forToken, endErrorStatement));
}
SkipNewlines();
PipelineBaseAst initializer = PipelineChainRule();
if (initializer != null)
{
endErrorStatement = initializer.Extent;
}
if (PeekToken().Kind == TokenKind.Semi)
{
endErrorStatement = NextToken().Extent;
}
SkipNewlines();
PipelineBaseAst condition = PipelineChainRule();
if (condition != null)
{
endErrorStatement = condition.Extent;
}
if (PeekToken().Kind == TokenKind.Semi)
{
endErrorStatement = NextToken().Extent;
}
SkipNewlines();
PipelineBaseAst iterator = PipelineChainRule();
if (iterator != null)
{
endErrorStatement = iterator.Extent;
}
SkipNewlines();
Token rParen = NextToken();
StatementBlockAst body = null;
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: don't continue parsing the for statement.
UngetToken(rParen);
endErrorStatement ??= lParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingEndParenthesisAfterStatement),
ParserStrings.MissingEndParenthesisAfterStatement,
forToken.Kind.Text());
}
else
{
body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: return an error statement.
endErrorStatement = rParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingLoopStatement),
ParserStrings.MissingLoopStatement,
forToken.Kind.Text());
}
}
if (body == null)
{
return new ErrorStatementAst(ExtentOf(labelToken ?? forToken, endErrorStatement),
GetNestedErrorAsts(initializer, condition, iterator));
}
return new ForStatementAst(ExtentOf(labelToken ?? forToken, body),
labelToken?.LabelText, initializer, condition, iterator, body);
}
private StatementAst WhileStatementRule(LabelToken labelToken, Token whileToken)
{
// G while-statement:
// G 'while ' new-lines:opt '(' new-lines:opt while-condition new-lines:opt ')' statement-block
// G
// G while-condition:
// G pipeline-chain
SkipNewlines();
Token lParen = NextToken();
if (lParen.Kind != TokenKind.LParen)
{
// ErrorRecovery: assume user just typed 'while' and hadn't started typing anything
// else yet. Next token is likely a newline, so just put it back and keep parsing.
UngetToken(lParen);
ReportIncompleteInput(After(whileToken),
nameof(ParserStrings.MissingOpenParenthesisAfterKeyword),
ParserStrings.MissingOpenParenthesisAfterKeyword,
whileToken.Text);
return new ErrorStatementAst(ExtentOf(labelToken ?? whileToken, whileToken));
}
SkipNewlines();
PipelineBaseAst condition = PipelineChainRule();
PipelineBaseAst errorCondition = null;
if (condition == null)
{
// ErrorRecovery: assume pipeline just hasn't been entered yet, continue hoping
// to find a close paren and statement block.
IScriptExtent errorPosition = After(lParen);
ReportIncompleteInput(
errorPosition,
nameof(ParserStrings.MissingExpressionAfterKeyword),
ParserStrings.MissingExpressionAfterKeyword,
whileToken.Kind.Text());
condition = new ErrorStatementAst(errorPosition);
}
else
{
errorCondition = condition;
}
SkipNewlines();
Token rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: assume the next token is a newline or part of something else,
// so stop parsing the statement and try parsing something else if possible.
UngetToken(rParen);
if (condition is not ErrorStatementAst)
{
ReportIncompleteInput(After(condition),
nameof(ParserStrings.MissingEndParenthesisAfterStatement),
ParserStrings.MissingEndParenthesisAfterStatement,
whileToken.Kind.Text());
}
return new ErrorStatementAst(ExtentOf(labelToken ?? whileToken, condition), GetNestedErrorAsts(errorCondition));
}
SkipNewlines();
StatementBlockAst body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: assume the next token is a newline or part of something else.
ReportIncompleteInput(After(rParen),
nameof(ParserStrings.MissingLoopStatement),
ParserStrings.MissingLoopStatement,
whileToken.Kind.Text());
return new ErrorStatementAst(ExtentOf(labelToken ?? whileToken, rParen), GetNestedErrorAsts(errorCondition));
}
return new WhileStatementAst(ExtentOf(labelToken ?? whileToken, body),
labelToken?.LabelText, condition, body);
}
/// <summary>
/// Parse a dynamic keyword statement which will be either of the form
/// keyword [parameters] [name] { a=1; b=2; } # constructor with properties
/// or
/// keyword [parameters] [name] { ... } # constructor with a simple body.
/// or keywordcommand parameters
/// This custom keyword does not introduce a new AST node type. Instead it generates a
/// CommandAst that calls a PowerShell command to implement the keyword's logic.
/// This command has one of two signatures:
/// keywordImplCommand.
/// </summary>
/// <param name="functionName">The name of the function to invoke.</param>
/// <param name="keywordData">The data for this keyword definition.</param>
/// <returns></returns>
private StatementAst DynamicKeywordStatementRule(Token functionName, DynamicKeyword keywordData)
{
//////////////////////////////////////////////////////////////////////////////////
// If a custom action was provided. then invoke it
//////////////////////////////////////////////////////////////////////////////////
if (keywordData.PreParse != null)
{
try
{
ParseError[] errors = keywordData.PreParse(keywordData);
if (errors != null && errors.Length > 0)
{
foreach (var e in errors)
{
ReportError(e);
}
}
}
catch (Exception e)
{
ReportError(functionName.Extent,
nameof(ParserStrings.DynamicKeywordPreParseException),
ParserStrings.DynamicKeywordPreParseException,
keywordData.ResourceName, e.ToString());
return null;
}
}
if (keywordData.IsReservedKeyword)
{
// ErrorRecovery: eat the token
ReportError(functionName.Extent,
nameof(ParserStrings.UnsupportedReservedKeyword),
ParserStrings.UnsupportedReservedKeyword,
keywordData.Keyword);
return null;
}
if (keywordData.HasReservedProperties)
{
// ErrorRecovery: eat the token
ReportError(functionName.Extent,
nameof(ParserStrings.UnsupportedReservedProperty),
ParserStrings.UnsupportedReservedProperty,
"'Require', 'Trigger', 'Notify', 'Before', 'After' and 'Subscribe'");
return null;
}
string elementName = string.Empty;
DynamicKeywordStatementAst dynamicKeywordAst;
if (keywordData.BodyMode == DynamicKeywordBodyMode.Command)
{
UngetToken(functionName);
dynamicKeywordAst = (DynamicKeywordStatementAst)CommandRule(forDynamicKeyword: true);
dynamicKeywordAst.Keyword = keywordData;
dynamicKeywordAst.FunctionName = functionName;
}
else
{
SkipNewlines();
// The expression that returns the resource name or names.
ExpressionAst instanceName = null;
Token nameToken = NextToken();
if (nameToken.Kind == TokenKind.EndOfInput)
{
UngetToken(nameToken);
if (keywordData.NameMode == DynamicKeywordNameMode.NameRequired || keywordData.NameMode == DynamicKeywordNameMode.SimpleNameRequired)
{
ReportIncompleteInput(After(functionName),
nameof(ParserStrings.RequiredNameOrExpressionMissing),
ParserStrings.RequiredNameOrExpressionMissing);
}
else
{
// Name not required so report missing brace
ReportIncompleteInput(After(functionName.Extent),
nameof(ParserStrings.MissingBraceInObjectDefinition),
ParserStrings.MissingBraceInObjectDefinition);
}
return null;
}
// If it's an lcurly, then no name was provided, and we skip to the body processing
Token lCurly = null;
if (nameToken.Kind == TokenKind.LCurly)
{
lCurly = nameToken;
if (keywordData.NameMode == DynamicKeywordNameMode.NameRequired || keywordData.NameMode == DynamicKeywordNameMode.SimpleNameRequired)
{
ReportError(After(functionName),
nameof(ParserStrings.RequiredNameOrExpressionMissing),
ParserStrings.RequiredNameOrExpressionMissing);
UngetToken(nameToken);
return null;
}
}
else if (nameToken.Kind == TokenKind.Identifier || nameToken.Kind == TokenKind.DynamicKeyword)
{
if (keywordData.NameMode == DynamicKeywordNameMode.NoName)
{
ReportError(After(functionName),
nameof(ParserStrings.UnexpectedNameForType),
ParserStrings.UnexpectedNameForType,
functionName.Text,
nameToken.Text);
UngetToken(nameToken);
return null;
}
// If it's an identifier then this is the name for the data object
elementName = nameToken.Text;
// If only a simple name is allowed, then the string must be non-null.
if ((keywordData.NameMode == DynamicKeywordNameMode.SimpleNameRequired || keywordData.NameMode == DynamicKeywordNameMode.SimpleOptionalName) && string.IsNullOrEmpty(elementName))
{
ReportIncompleteInput(After(functionName),
nameof(ParserStrings.RequiredNameOrExpressionMissing),
ParserStrings.RequiredNameOrExpressionMissing);
UngetToken(nameToken);
return null;
}
}
else
{
// see if an expression was provided instead of a bare word...
UngetToken(nameToken);
instanceName = GetSingleCommandArgument(CommandArgumentContext.CommandName);
if (instanceName == null)
{
if (keywordData.NameMode == DynamicKeywordNameMode.SimpleNameRequired || keywordData.NameMode == DynamicKeywordNameMode.SimpleOptionalName)
{
ReportError(After(functionName),
nameof(ParserStrings.RequiredNameOrExpressionMissing),
ParserStrings.RequiredNameOrExpressionMissing);
}
else
{
// It wasn't an '{' and it wasn't a name expression so it's a unexpected token.
ReportError(After(functionName),
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
nameToken.Text);
}
return null;
}
// Ok, we got a name expression, but we're expecting no name, so it's and error.
if (keywordData.NameMode == DynamicKeywordNameMode.NoName)
{
ReportError(After(functionName),
nameof(ParserStrings.UnexpectedNameForType),
ParserStrings.UnexpectedNameForType,
functionName.Text,
instanceName.ToString());
return null;
}
// We were expecting a simple name so report an error
if (keywordData.NameMode == DynamicKeywordNameMode.SimpleNameRequired || keywordData.NameMode == DynamicKeywordNameMode.SimpleOptionalName)
{
// If no match, then this is an incomplete token BUGBUG fix message
ReportError(nameToken.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
nameToken.Text);
return null;
}
}
// If we didn't get a resource expression AST, then we need to build one out of the
// name that was specified. It may be the case that we don't have
// a resource name in which case it will be the empty string. Even in the cases were
// we aren't expecting a name, we still do this so that the signature of the implementing function remains
// the same.
ExpressionAst originalInstanceName = instanceName;
instanceName ??= new StringConstantExpressionAst(nameToken.Extent, elementName, StringConstantType.BareWord);
SkipNewlines();
//
// Now look for the body of the data statement.
//
if (lCurly == null)
{
lCurly = NextToken();
if (lCurly.Kind == TokenKind.EndOfInput)
{
UngetToken(lCurly);
ReportIncompleteInput(After(functionName.Extent),
nameof(ParserStrings.MissingBraceInObjectDefinition),
ParserStrings.MissingBraceInObjectDefinition);
// Preserve the name expression for tab completion
return originalInstanceName == null
? null
: new ErrorStatementAst(ExtentOf(functionName, originalInstanceName),
GetNestedErrorAsts(originalInstanceName));
}
if (lCurly.Kind != TokenKind.LCurly)
{
// We need to generate a reasonable error message for this case:
//
// Configuration C {
// node $AllNode.NodeName{ # There is no space before curly, and we are converting scriptblock to and argument to call 'NodeName'
// ...
// }
// } # we don't want to simple report an unexpected token here, it would be super-confusing.
InvokeMemberExpressionAst instanceInvokeMemberExpressionAst = instanceName as InvokeMemberExpressionAst;
if (instanceInvokeMemberExpressionAst != null &&
instanceInvokeMemberExpressionAst.Arguments.Count == 1 &&
instanceInvokeMemberExpressionAst.Arguments[0] is ScriptBlockExpressionAst &&
// the last condition checks that there is no space between "method" name and '{'
instanceInvokeMemberExpressionAst.Member.Extent.EndOffset == instanceInvokeMemberExpressionAst.Arguments[0].Extent.StartOffset)
{
ReportError(LastCharacterOf(instanceInvokeMemberExpressionAst.Member.Extent),
nameof(ParserStrings.UnexpectedTokenInDynamicKeyword),
ParserStrings.UnexpectedTokenInDynamicKeyword,
functionName.Text);
}
else
{
ReportError(lCurly.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
lCurly.Text);
}
if (lCurly.Kind == TokenKind.Dot && originalInstanceName != null && lCurly.Extent.StartOffset == originalInstanceName.Extent.EndOffset)
{
// Generate more useful ast for tab-completing extension methods on special DSC collection variables
// e.g. configuration foo { node $AllNodes.<tab>
IScriptExtent errorExtent = ExtentOf(originalInstanceName, lCurly);
var errorExpr = new ErrorExpressionAst(errorExtent);
var memberExpr = new MemberExpressionAst(originalInstanceName.Extent, originalInstanceName, errorExpr, @static: false);
return new ErrorStatementAst(errorExtent, new[] { memberExpr });
}
UngetToken(lCurly);
// Preserve the name expression for tab completion
return originalInstanceName == null
? null
: new ErrorStatementAst(ExtentOf(functionName, originalInstanceName),
GetNestedErrorAsts(originalInstanceName));
}
}
//
// The keyword data is used to see
// if a scriptblock or a hashtable is expected.
//
ExpressionAst body = null;
if (keywordData.BodyMode == DynamicKeywordBodyMode.ScriptBlock)
{
var oldInConfiguration = _inConfiguration;
try
{
_inConfiguration = false;
body = ScriptBlockExpressionRule(lCurly);
}
finally
{
_inConfiguration = oldInConfiguration;
}
}
else if (keywordData.BodyMode == DynamicKeywordBodyMode.Hashtable)
{
// Resource property value could be set to nested DSC resources except Script resource
bool isScriptResource = string.Equals(functionName.Text, @"Script", StringComparison.OrdinalIgnoreCase);
try
{
if (isScriptResource)
DynamicKeyword.Push();
body = HashExpressionRule(lCurly, true /* parsingSchemaElement */);
}
finally
{
if (isScriptResource)
DynamicKeyword.Pop();
}
}
// commandast
// elements: instancename/dynamickeyword/hashtable or scripblockexpress
if (body == null)
{
// Failed to read the statement body
ReportIncompleteInput(After(lCurly),
nameof(ParserStrings.MissingStatementAfterKeyword),
ParserStrings.MissingStatementAfterKeyword,
keywordData.Keyword);
// Preserve the name expression for tab completion
return originalInstanceName == null
? null
: new ErrorStatementAst(ExtentOf(functionName, originalInstanceName),
GetNestedErrorAsts(originalInstanceName));
}
//////////////////////////////////////////////////////////////////////////
// The statement is now fully parsed
//////////////////////////////////////////////////////////////////////////
//
// Create DynamicKeywordStatementAst
//
Collection<CommandElementAst> commandElements = new Collection<CommandElementAst>
{
new StringConstantExpressionAst(functionName.Extent, functionName.Text, StringConstantType.BareWord),
(ExpressionAst)instanceName.Copy(),
(ExpressionAst)body.Copy()
};
Token nextToken = NextToken();
IScriptExtent dynamicKeywordExtent = ExtentOf(functionName, Before(nextToken));
UngetToken(nextToken);
dynamicKeywordAst = new DynamicKeywordStatementAst(dynamicKeywordExtent, commandElements)
{
Keyword = keywordData,
LCurly = lCurly,
FunctionName = functionName,
InstanceName = instanceName,
OriginalInstanceName = originalInstanceName,
BodyExpression = body,
ElementName = elementName,
};
}
//////////////////////////////////////////////////////////////////////////////////
// If a custom action was provided. then invoke it
//////////////////////////////////////////////////////////////////////////////////
if (keywordData.PostParse != null)
{
try
{
ParseError[] errors = keywordData.PostParse(dynamicKeywordAst);
if (errors != null && errors.Length > 0)
{
foreach (var e in errors)
{
ReportError(e);
}
}
}
catch (Exception e)
{
ReportError(functionName.Extent,
nameof(ParserStrings.DynamicKeywordPostParseException),
ParserStrings.DynamicKeywordPostParseException,
keywordData.Keyword,
e.ToString());
return null;
}
}
return dynamicKeywordAst;
}
internal StatementAst CreateErrorStatementAst(
Token functionName,
ExpressionAst instanceName,
ExpressionAst bodyExpression)
{
return new ErrorStatementAst(ExtentOf(functionName, bodyExpression), GetNestedErrorAsts(instanceName, bodyExpression));
}
private StatementAst DoWhileStatementRule(LabelToken labelToken, Token doToken)
{
// G do-statement:
// G 'do' statement-block new-lines:opt 'while' new-lines:opt '(' while-condition new-lines:opt ')'
// G 'do' statement-block new-lines:opt 'until' new-lines:opt '(' while-condition new-lines:opt ')'
// G
// G while-condition:
// G new-lines:opt pipeline
IScriptExtent startExtent = (labelToken ?? doToken).Extent;
IScriptExtent endErrorStatement = null;
Token rParen = null;
Token whileOrUntilToken = null;
PipelineBaseAst condition = null;
StatementBlockAst body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: Skip the keyword and stop trying to parse this statement, continue on whatever
// comes next.
endErrorStatement = doToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingLoopStatement),
ParserStrings.MissingLoopStatement,
TokenKind.Do.Text());
}
else
{
SkipNewlines();
whileOrUntilToken = NextToken();
if (whileOrUntilToken.Kind != TokenKind.While && whileOrUntilToken.Kind != TokenKind.Until)
{
// ErrorRecovery: Skip looking for a condition, continue on whatever comes next.
UngetToken(whileOrUntilToken);
endErrorStatement = body.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingWhileOrUntilInDoWhile),
ParserStrings.MissingWhileOrUntilInDoWhile);
}
else
{
SkipNewlines();
Token lParen = NextToken();
if (lParen.Kind != TokenKind.LParen)
{
// ErrorRecovery: Skip looking for the condition, return an error statement.
UngetToken(lParen);
endErrorStatement = whileOrUntilToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingOpenParenthesisAfterKeyword),
ParserStrings.MissingOpenParenthesisAfterKeyword,
whileOrUntilToken.Kind.Text());
}
else
{
SkipNewlines();
condition = PipelineChainRule();
if (condition == null)
{
// ErrorRecovery: try to get the matching close paren, then return an error statement.
endErrorStatement = lParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingExpressionAfterKeyword),
ParserStrings.MissingExpressionAfterKeyword,
whileOrUntilToken.Kind.Text());
}
SkipNewlines();
rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: this is it, so just pretend we saw the paren, return an error statement.
UngetToken(rParen);
// If condition == null, we issue an error message already, don't bother with this one.
if (condition != null)
{
endErrorStatement = condition.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingEndParenthesisAfterStatement),
ParserStrings.MissingEndParenthesisAfterStatement,
whileOrUntilToken.Kind.Text());
}
}
}
}
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(startExtent, endErrorStatement), GetNestedErrorAsts(body, condition));
}
IScriptExtent extent = ExtentOf(startExtent, rParen);
string label = labelToken?.LabelText;
if (whileOrUntilToken.Kind == TokenKind.Until)
{
return new DoUntilStatementAst(extent, label, condition, body);
}
return new DoWhileStatementAst(extent, label, condition, body);
}
private StatementAst ClassDefinitionRule(List<AttributeBaseAst> customAttributes, Token classToken)
{
// G class-statement:
// G attribute-list:opt 'class' new-lines:opt class-name new-lines:opt '{' class-member-list '}'
// G attribute-list:opt 'class' new-lines:opt class-name new-lines:opt ':' base-type-list '{' new-lines:opt class-member-list:opt '}'
// G
// G class-name:
// G simple-name
// G
// G base-type-list:
// G new-lines:opt type-name new-lines:opt
// G base-class-list ',' new-lines:opt type-name new-lines:opt
// G
// G class-member-list:
// G class-member new-lines:opt
// G class-member-list class-member
// PowerShell classes are not supported in ConstrainedLanguage
if (Runspace.DefaultRunspace?.ExecutionContext?.LanguageMode == PSLanguageMode.ConstrainedLanguage)
{
if (SystemPolicy.GetSystemLockdownPolicy() != SystemEnforcementMode.Audit)
{
ReportError(classToken.Extent,
nameof(ParserStrings.ClassesNotAllowedInConstrainedLanguage),
ParserStrings.ClassesNotAllowedInConstrainedLanguage,
classToken.Kind.Text());
return null;
}
SystemPolicy.LogWDACAuditMessage(
context: Runspace.DefaultRunspace?.ExecutionContext,
title: ParserStrings.WDACParserClassKeywordLogTitle,
message: ParserStrings.WDACParserClassKeywordLogMessage,
fqid: "ClassLanguageKeywordNotAllowed",
dropIntoDebugger: true);
}
SkipNewlines();
Token classNameToken;
var name = SimpleNameRule(out classNameToken);
if (name == null)
{
ReportIncompleteInput(After(classToken),
nameof(ParserStrings.MissingNameAfterKeyword),
ParserStrings.MissingNameAfterKeyword,
classToken.Text);
return new ErrorStatementAst(classToken.Extent);
}
// Class name token represents a name of type, not a member. Highlight it as a type name.
classNameToken.TokenFlags &= ~TokenFlags.MemberName;
classNameToken.TokenFlags |= TokenFlags.TypeName;
SkipNewlines();
// handle inheritance constraint.
var oldTokenizerMode = _tokenizer.Mode;
var superClassesList = new List<TypeConstraintAst>();
try
{
SetTokenizerMode(TokenizerMode.Signature);
Token colonToken = PeekToken();
if (colonToken.Kind == TokenKind.Colon)
{
this.SkipToken();
SkipNewlines();
ITypeName superClass;
Token commaToken = null;
while (true)
{
superClass = this.TypeNameRule(allowAssemblyQualifiedNames: false, firstTypeNameToken: out _);
if (superClass == null)
{
ReportIncompleteInput(After(ExtentFromFirstOf(commaToken, colonToken)),
nameof(ParserStrings.TypeNameExpected),
ParserStrings.TypeNameExpected);
break;
}
superClassesList.Add(new TypeConstraintAst(superClass.Extent, superClass));
SkipNewlines();
commaToken = this.PeekToken();
if (commaToken.Kind == TokenKind.Comma)
{
this.SkipToken();
this.SkipNewlines();
}
else
{
break;
}
}
}
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: If there is no opening curly, assume it hasn't been entered yet and don't consume anything.
UngetToken(lCurly);
var lastElement = (superClassesList.Count > 0) ? (Ast)superClassesList[superClassesList.Count - 1] : name;
ReportIncompleteInput(After(lastElement),
nameof(ParserStrings.MissingTypeBody),
ParserStrings.MissingTypeBody,
classToken.Kind.Text());
return new ErrorStatementAst(ExtentOf(classToken, lastElement), superClassesList);
}
IScriptExtent lastExtent = lCurly.Extent;
List<Ast> nestedAsts = null;
MemberAst member;
List<MemberAst> members = new List<MemberAst>();
List<Ast> astsOnError = null;
while ((member = ClassMemberRule(name.Value, out astsOnError)) != null || astsOnError != null)
{
if (member != null)
{
members.Add(member);
lastExtent = member.Extent;
}
if (astsOnError != null && astsOnError.Count > 0)
{
nestedAsts ??= new List<Ast>();
nestedAsts.AddRange(astsOnError);
lastExtent = astsOnError.Last().Extent;
}
}
var rCurly = NextToken();
if (rCurly.Kind != TokenKind.RCurly)
{
UngetToken(rCurly);
ReportIncompleteInput(
After(lCurly),
rCurly.Extent,
nameof(ParserStrings.MissingEndCurlyBrace),
ParserStrings.MissingEndCurlyBrace);
}
else
{
lastExtent = rCurly.Extent;
}
var startExtent = customAttributes != null && customAttributes.Count > 0
? customAttributes[0].Extent
: classToken.Extent;
var extent = ExtentOf(startExtent, lastExtent);
var classDefn = new TypeDefinitionAst(extent, name.Value, customAttributes?.OfType<AttributeAst>(), members, TypeAttributes.Class, superClassesList);
if (customAttributes != null && customAttributes.OfType<TypeConstraintAst>().Any())
{
nestedAsts ??= new List<Ast>();
// no need to report error since the error is reported in method StatementRule
nestedAsts.AddRange(customAttributes.OfType<TypeConstraintAst>());
nestedAsts.Add(classDefn);
}
if (nestedAsts != null && nestedAsts.Count > 0)
{
return new ErrorStatementAst(extent, nestedAsts);
}
return classDefn;
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private MemberAst ClassMemberRule(string className, out List<Ast> astsOnError)
{
// G class-member:
// G method-member
// G property-member
// G
// G method-member:
// G member-attribute-list:opt function-statement
// G
// G property-member:
// G member-attribute-list:opt variable
// G member-attribute-list:opt variable '=' expression
// G
// G member-attribute-list:
// G member-attribute
// G member-attribute-list member-attribute
// G
// G member-attribute:
// G attribute
// G 'static'
// G 'hidden'
IScriptExtent startExtent = null;
var attributeList = new List<AttributeAst>();
TypeConstraintAst typeConstraint = null;
bool scanningAttributes = true;
Token staticToken = null;
Token hiddenToken = null;
Token token = null;
object lastAttribute = null;
astsOnError = null;
#if SUPPORT_PUBLIC_PRIVATE
// Public/private not yet supported
Token publicToken = null;
Token privateToken = null;
#endif
while (scanningAttributes)
{
SkipNewlines();
var attribute = AttributeRule();
if (attribute != null)
{
lastAttribute = attribute;
startExtent ??= attribute.Extent;
var attributeAst = attribute as AttributeAst;
if (attributeAst != null)
{
attributeList.Add(attributeAst);
}
else if (typeConstraint == null)
{
typeConstraint = (TypeConstraintAst)attribute;
}
else
{
ReportError(attribute.Extent, nameof(ParserStrings.TooManyTypes), ParserStrings.TooManyTypes);
}
continue;
}
token = PeekToken();
startExtent ??= token.Extent;
switch (token.Kind)
{
#if SUPPORT_PUBLIC_PRIVATE
case TokenKind.Public:
if (publicToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.DuplicateQualifier),
ParserStrings.DuplicateQualifier,
token.Text);
}
if (privateToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.ModifiersCannotBeCombined),
ParserStrings.ModifiersCannotBeCombined,
privateToken.Text,
token.Text);
}
publicToken = token;
SkipToken();
break;
case TokenKind.Private:
if (privateToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.DuplicateQualifier),
ParserStrings.DuplicateQualifier,
token.Text);
}
if (publicToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.ModifiersCannotBeCombined),
ParserStrings.ModifiersCannotBeCombined,
publicToken.Text,
token.Text);
}
privateToken = token;
SkipToken();
break;
#endif
case TokenKind.Hidden:
if (hiddenToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.DuplicateQualifier),
ParserStrings.DuplicateQualifier,
token.Text);
}
hiddenToken = token;
lastAttribute = token;
SkipToken();
break;
case TokenKind.Static:
if (staticToken != null)
{
ReportError(token.Extent,
nameof(ParserStrings.DuplicateQualifier),
ParserStrings.DuplicateQualifier,
token.Text);
}
staticToken = token;
lastAttribute = token;
SkipToken();
break;
default:
scanningAttributes = false;
break;
}
}
if (token.Kind == TokenKind.Variable)
{
SkipToken();
var varToken = token as VariableToken;
ExpressionAst initialValueAst = null;
var assignToken = PeekToken();
if (assignToken.Kind == TokenKind.Equals)
{
SkipToken();
SkipNewlines();
initialValueAst = ExpressionRule();
}
#if SUPPORT_PUBLIC_PRIVATE
PropertyAttributes attributes = privateToken != null ? PropertyAttributes.Private : PropertyAttributes.Public;
#else
PropertyAttributes attributes = PropertyAttributes.Public;
#endif
if (staticToken != null)
{
attributes |= PropertyAttributes.Static;
}
if (hiddenToken != null)
{
attributes |= PropertyAttributes.Hidden;
}
var endExtent = initialValueAst != null ? initialValueAst.Extent : varToken.Extent;
Token terminatorToken = PeekToken();
if (terminatorToken.Kind != TokenKind.NewLine && terminatorToken.Kind != TokenKind.Semi && terminatorToken.Kind != TokenKind.RCurly)
{
ReportIncompleteInput(After(endExtent),
nameof(ParserStrings.MissingPropertyTerminator),
ParserStrings.MissingPropertyTerminator);
}
SkipNewlinesAndSemicolons();
// Include the semicolon in the extent but not newline or rcurly as that will look weird, e.g. if an error is reported on the full extent
if (terminatorToken.Kind == TokenKind.Semi)
{
endExtent = terminatorToken.Extent;
}
if (!string.IsNullOrEmpty(varToken.Name))
{
return new PropertyMemberAst(ExtentOf(startExtent, endExtent), varToken.Name,
typeConstraint, attributeList, attributes, initialValueAst);
}
else
{
// Incompleted input like:
// class foo { $private: }
// Error message already emitted by tokenizer ScanVariable
RecordErrorAsts(attributeList, ref astsOnError);
RecordErrorAsts(typeConstraint, ref astsOnError);
RecordErrorAsts(initialValueAst, ref astsOnError);
return null;
}
}
if (TryUseTokenAsSimpleName(token))
{
SkipToken();
var functionDefinition = MethodDeclarationRule(token, className, staticToken != null) as FunctionDefinitionAst;
if (functionDefinition == null)
{
// TODO: better error recovery - shouldn't assume this was the last class member
Diagnostics.Assert(ErrorList.Count > 0, "Should be an error if we don't have a function");
SyncOnError(false, TokenKind.RCurly);
RecordErrorAsts(attributeList, ref astsOnError);
RecordErrorAsts(typeConstraint, ref astsOnError);
return null;
}
#if SUPPORT_PUBLIC_PRIVATE
MethodAttributes attributes = privateToken != null ? MethodAttributes.Private : MethodAttributes.Public;
#else
MethodAttributes attributes = MethodAttributes.Public;
#endif
if (staticToken != null)
{
attributes |= MethodAttributes.Static;
}
if (hiddenToken != null)
{
attributes |= MethodAttributes.Hidden;
}
return new FunctionMemberAst(ExtentOf(startExtent, functionDefinition), functionDefinition, typeConstraint, attributeList, attributes);
}
if (lastAttribute != null)
{
// We have the start of a member, but didn't see a variable or 'def'.
ReportIncompleteInput(After(ExtentFromFirstOf(lastAttribute)),
nameof(ParserStrings.IncompleteMemberDefinition),
ParserStrings.IncompleteMemberDefinition);
RecordErrorAsts(attributeList, ref astsOnError);
RecordErrorAsts(typeConstraint, ref astsOnError);
}
return null;
}
private static bool TryUseTokenAsSimpleName(Token token)
{
if (token.Kind == TokenKind.Identifier
|| token.Kind == TokenKind.DynamicKeyword
|| token.TokenFlags.HasFlag(TokenFlags.Keyword))
{
token.TokenFlags = TokenFlags.None;
return true;
}
return false;
}
private static void RecordErrorAsts(Ast errAst, ref List<Ast> astsOnError)
{
if (errAst == null)
{
return;
}
astsOnError ??= new List<Ast>();
astsOnError.Add(errAst);
}
private static void RecordErrorAsts(IEnumerable<Ast> errAsts, ref List<Ast> astsOnError)
{
if (errAsts == null || !errAsts.Any())
{
return;
}
astsOnError ??= new List<Ast>();
astsOnError.AddRange(errAsts);
}
private Token NextTypeIdentifierToken()
{
var oldTokenizerMode = _tokenizer.Mode;
try
{
// Switch to typename mode to avoid aggressive argument tokenization.
SetTokenizerMode(TokenizerMode.TypeName);
Token typeName = NextToken();
if (typeName.Kind != TokenKind.Identifier)
{
UngetToken(typeName);
return null;
}
return typeName;
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private StatementAst EnumDefinitionRule(List<AttributeBaseAst> customAttributes, Token enumToken)
{
// G enum-statement:
// G 'enum' new-lines:opt enum-name '{' enum-member-list '}'
// G 'enum' new-lines:opt enum-name ':' enum-underlying-type '{' enum-member-list '}'
// G
// G enum-name:
// G simple-name
// G
// G enum-underlying-type:
// G new-lines:opt valid-type-name new-lines:opt
// G
// G enum-member-list:
// G enum-member new-lines:opt
// G enum-member-list enum-member
const TypeCode ValidUnderlyingTypeCodes = TypeCode.Byte | TypeCode.Int16 | TypeCode.Int32 | TypeCode.Int64 | TypeCode.SByte | TypeCode.UInt16 | TypeCode.UInt32 | TypeCode.UInt64;
SkipNewlines();
var name = SimpleNameRule();
if (name == null)
{
ReportIncompleteInput(
After(enumToken),
nameof(ParserStrings.MissingNameAfterKeyword),
ParserStrings.MissingNameAfterKeyword,
enumToken.Text);
return new ErrorStatementAst(enumToken.Extent);
}
TypeConstraintAst underlyingTypeConstraint = null;
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Signature);
Token colonToken = PeekToken();
if (colonToken.Kind == TokenKind.Colon)
{
this.SkipToken();
SkipNewlines();
ITypeName underlyingType;
underlyingType = this.TypeNameRule(allowAssemblyQualifiedNames: false, firstTypeNameToken: out _);
if (underlyingType == null)
{
ReportIncompleteInput(
After(colonToken),
nameof(ParserStrings.TypeNameExpected),
ParserStrings.TypeNameExpected);
}
else
{
var resolvedType = underlyingType.GetReflectionType();
if (resolvedType == null || !ValidUnderlyingTypeCodes.HasFlag(resolvedType.GetTypeCode()))
{
ReportError(
underlyingType.Extent,
nameof(ParserStrings.InvalidUnderlyingType),
ParserStrings.InvalidUnderlyingType,
underlyingType.Name);
}
underlyingTypeConstraint = new TypeConstraintAst(underlyingType.Extent, underlyingType);
}
}
SkipNewlines();
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: If there is no opening curly, assume it hasn't been entered yet and don't consume anything.
UngetToken(lCurly);
ReportIncompleteInput(
After(name),
nameof(ParserStrings.MissingTypeBody),
ParserStrings.MissingTypeBody,
enumToken.Kind.Text());
return new ErrorStatementAst(ExtentOf(enumToken, name));
}
IScriptExtent lastExtent = lCurly.Extent;
MemberAst member;
List<MemberAst> members = new List<MemberAst>();
while ((member = EnumMemberRule()) != null)
{
members.Add(member);
lastExtent = member.Extent;
}
var rCurly = NextToken();
if (rCurly.Kind != TokenKind.RCurly)
{
UngetToken(rCurly);
ReportIncompleteInput(
After(lCurly),
rCurly.Extent,
nameof(ParserStrings.MissingEndCurlyBrace),
ParserStrings.MissingEndCurlyBrace);
}
var startExtent = customAttributes != null && customAttributes.Count > 0
? customAttributes[0].Extent
: enumToken.Extent;
var extent = ExtentOf(startExtent, rCurly);
var enumDefn = new TypeDefinitionAst(extent, name.Value, customAttributes?.OfType<AttributeAst>(), members, TypeAttributes.Enum, underlyingTypeConstraint == null ? null : new[] { underlyingTypeConstraint });
if (customAttributes != null && customAttributes.OfType<TypeConstraintAst>().Any())
{
// No need to report error since there is error reported in method StatementRule
List<Ast> nestedAsts = new List<Ast>();
nestedAsts.AddRange(customAttributes.OfType<TypeConstraintAst>());
nestedAsts.Add(enumDefn);
return new ErrorStatementAst(startExtent, nestedAsts);
}
return enumDefn;
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private MemberAst EnumMemberRule()
{
// G enum-member:
// G enum-member-name
// G enum-member-name '=' expression
// G
// G enum-member-name:
// G simple-name
const PropertyAttributes enumMemberAttributes =
PropertyAttributes.Public | PropertyAttributes.Static | PropertyAttributes.Literal;
SkipNewlines();
var enumeratorName = SimpleNameRule();
if (enumeratorName == null)
{
return null;
}
IScriptExtent endExtent = enumeratorName.Extent;
ExpressionAst initialValueAst = null;
var missingInitializer = false;
var oldTokenizerMode = _tokenizer.Mode;
Token assignToken = null;
try
{
SetTokenizerMode(TokenizerMode.Expression);
assignToken = PeekToken();
if (assignToken.Kind == TokenKind.Equals)
{
SkipToken();
initialValueAst = ExpressionRule();
if (initialValueAst == null)
{
ReportError(After(assignToken),
nameof(ParserStrings.ExpectedValueExpression),
ParserStrings.ExpectedValueExpression,
assignToken.Kind.Text());
endExtent = assignToken.Extent;
missingInitializer = true;
}
else
{
endExtent = initialValueAst.Extent;
}
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
Token terminatorToken = PeekToken();
if (terminatorToken.Kind != TokenKind.NewLine && terminatorToken.Kind != TokenKind.Semi && terminatorToken.Kind != TokenKind.RCurly)
{
// If the initializer is missing, no sense in reporting another error about a missing terminator
if (!missingInitializer)
{
ReportIncompleteInput(After(endExtent),
nameof(ParserStrings.MissingPropertyTerminator),
ParserStrings.MissingPropertyTerminator);
}
}
SkipNewlinesAndSemicolons();
// Include the semicolon in the extent but not newline or rcurly as that will look weird, e.g. if an error is reported on the full extent
if (terminatorToken.Kind == TokenKind.Semi)
{
endExtent = terminatorToken.Extent;
}
return new PropertyMemberAst(ExtentOf(enumeratorName, endExtent), enumeratorName.Value, null, null, enumMemberAttributes, initialValueAst);
}
private StatementAst UsingStatementRule(Token usingToken)
{
// G using-statement:
// G 'using' 'namespace' identifier
// G 'using' 'namespace' identifier '=' identifier
// G 'using' 'module' identifier
// G 'using' 'module' identifier '=' identifier
// G 'using' 'module' hashtable
// G 'using' 'module' identifier '=' hashtable
// G 'using' 'type' identifier '=' identifier
// G 'using' 'assembly' identifier
// G 'using' 'command' identifier '=' identifier
// in case of aliasing (=), first identifier is alias name, second identifier is alias value (the real name).
var directiveToken = NextToken();
UsingStatementKind kind;
bool aliasAllowed = false;
bool aliasRequired = false;
switch (directiveToken.Kind)
{
case TokenKind.Namespace:
kind = UsingStatementKind.Namespace;
aliasAllowed = true;
break;
case TokenKind.Type:
kind = UsingStatementKind.Type;
aliasRequired = true;
break;
case TokenKind.Module:
kind = UsingStatementKind.Module;
aliasAllowed = true;
break;
case TokenKind.Command:
kind = UsingStatementKind.Command;
aliasRequired = true;
break;
case TokenKind.Assembly:
kind = UsingStatementKind.Assembly;
break;
default:
UngetToken(directiveToken);
ReportIncompleteInput(After(usingToken),
nameof(ParserStrings.MissingUsingStatementDirective),
ParserStrings.MissingUsingStatementDirective);
return new ErrorStatementAst(usingToken.Extent);
}
var itemToken = NextToken();
switch (itemToken.Kind)
{
case TokenKind.EndOfInput:
case TokenKind.NewLine:
// Example: 'using module ,exampleModuleName'
// GetCommandArgument will successfully return an argument for a unary array argument
// but we don't want to allow that syntax with a using statement.
case TokenKind.Comma:
case TokenKind.Semi:
{
ReportIncompleteInput(After(directiveToken),
nameof(ParserStrings.MissingUsingItemName),
ParserStrings.MissingUsingItemName);
return new ErrorStatementAst(ExtentOf(usingToken, directiveToken));
}
}
var itemAst = GetCommandArgument(CommandArgumentContext.CommandArgument, itemToken);
if (itemAst == null)
{
ReportError(itemToken.Extent,
nameof(ParserStrings.InvalidValueForUsingItemName),
ParserStrings.InvalidValueForUsingItemName,
itemToken.Text);
// ErrorRecovery: If there is no identifier, skip whole 'using' line
SyncOnError(true, TokenKind.Semi, TokenKind.NewLine);
return new ErrorStatementAst(ExtentOf(usingToken, itemToken.Extent));
}
if (itemAst is not StringConstantExpressionAst
&& (kind != UsingStatementKind.Module || itemAst is not HashtableAst))
{
ReportError(ExtentFromFirstOf(itemAst, itemToken),
nameof(ParserStrings.InvalidValueForUsingItemName),
ParserStrings.InvalidValueForUsingItemName,
itemAst.Extent.Text);
return new ErrorStatementAst(ExtentOf(usingToken, ExtentFromFirstOf(itemAst, itemToken)));
}
HashtableAst htAst = null;
switch (kind)
{
case UsingStatementKind.Assembly:
itemAst = ResolveUsingAssembly((StringConstantExpressionAst)itemAst);
break;
case UsingStatementKind.Module:
htAst = itemAst as HashtableAst;
break;
}
// if htAst is not null, then we don't expect alias
if ((aliasAllowed || aliasRequired) && (htAst == null))
{
var equalsToken = PeekToken();
if (equalsToken.Kind == TokenKind.Equals)
{
SkipToken();
var aliasToken = NextToken();
if (aliasToken.Kind is TokenKind.EndOfInput or TokenKind.NewLine or TokenKind.Semi)
{
UngetToken(aliasToken);
ReportIncompleteInput(After(equalsToken),
nameof(ParserStrings.MissingNamespaceAlias),
ParserStrings.MissingNamespaceAlias);
return new ErrorStatementAst(ExtentOf(usingToken, equalsToken));
}
if (aliasToken.Kind == TokenKind.Comma)
{
ReportError(aliasToken.Extent, nameof(ParserStrings.UnexpectedUnaryOperator), ParserStrings.UnexpectedUnaryOperator, aliasToken.Text);
return new ErrorStatementAst(ExtentOf(usingToken, aliasToken));
}
var aliasAst = GetCommandArgument(CommandArgumentContext.CommandArgument, aliasToken);
if (kind == UsingStatementKind.Module && aliasAst is HashtableAst)
{
htAst = (HashtableAst)aliasAst;
}
else if (aliasAst is not StringConstantExpressionAst)
{
var errorExtent = ExtentFromFirstOf(aliasAst, aliasToken);
Ast[] nestedAsts;
if (aliasAst is null)
{
nestedAsts = new Ast[] { itemAst };
}
else
{
nestedAsts = new Ast[] { itemAst, aliasAst };
}
ReportError(errorExtent, nameof(ParserStrings.InvalidValueForUsingItemName), ParserStrings.InvalidValueForUsingItemName, errorExtent.Text);
return new ErrorStatementAst(ExtentOf(usingToken, errorExtent), nestedAsts);
}
RequireStatementTerminator();
if (htAst == null)
{
return new UsingStatementAst(
ExtentOf(usingToken, aliasToken),
kind,
(StringConstantExpressionAst)itemAst,
(StringConstantExpressionAst)aliasAst);
}
else
{
return new UsingStatementAst(
ExtentOf(usingToken, aliasToken),
(StringConstantExpressionAst)itemAst,
htAst);
}
}
if (aliasRequired)
{
ReportIncompleteInput(After(itemToken),
nameof(ParserStrings.MissingEqualsInUsingAlias),
ParserStrings.MissingEqualsInUsingAlias);
return new ErrorStatementAst(ExtentOf(usingToken, itemAst), new Ast[] { itemAst });
}
}
RequireStatementTerminator();
if (htAst == null)
{
return new UsingStatementAst(ExtentOf(usingToken, itemAst), kind, (StringConstantExpressionAst)itemAst);
}
else
{
return new UsingStatementAst(ExtentOf(usingToken, itemAst), htAst);
}
}
private StringConstantExpressionAst ResolveUsingAssembly(StringConstantExpressionAst name)
{
var assemblyName = name.Value;
Uri uri;
// assemblyName can be invalid, i.e. during typing.
// We only use uri to check that we don't allow UNC paths.
if (Uri.TryCreate(assemblyName, UriKind.Absolute, out uri))
{
if (uri.IsUnc)
{
ReportError(name.Extent,
nameof(ParserStrings.CannotLoadAssemblyFromUncPath),
ParserStrings.CannotLoadAssemblyFromUncPath,
assemblyName);
}
// don't allow things like 'using assembly http://microsoft.com'
if (uri.Scheme != "file")
{
ReportError(name.Extent,
nameof(ParserStrings.CannotLoadAssemblyWithUriSchema),
ParserStrings.CannotLoadAssemblyWithUriSchema,
uri.Scheme);
}
}
else
{
string assemblyFileName = assemblyName;
try
{
var scriptFileName = name.Extent.File;
if (!Path.IsPathRooted(assemblyFileName))
{
string workingDirectory;
if (string.IsNullOrEmpty(scriptFileName))
{
// We are in REPL, or Invoke-Expression, or ScriptBlock.Create, etc.
// It's legal to use '.\foo.dll', and we should do lookup in $pwd.
// We resolving it in parse time to avoid difference in script behavior,
// when script execution is delayed from parsing, i.e.
// $script = [scriptblock]::Create('using assembly .\foo.dll; [Foo].foo()')
// # using assembly behavior should be the same in these two cases:
// cd A; $script.Invoke()
// cd B; $script.Invoke()
if (Runspaces.Runspace.DefaultRunspace != null)
{
workingDirectory = Runspaces.Runspace.DefaultRunspace.ExecutionContext.EngineIntrinsics.SessionState.Path.CurrentLocation.Path;
}
else
{
// In this case, our best guess is process working directory.
workingDirectory = Directory.GetCurrentDirectory();
}
}
else
{
workingDirectory = Path.GetDirectoryName(scriptFileName);
}
assemblyFileName = Path.Combine(workingDirectory, assemblyFileName);
}
}
catch
{
}
if (assemblyFileName == null || !File.Exists(assemblyFileName))
{
ReportError(name.Extent,
nameof(ParserStrings.ErrorLoadingAssembly),
ParserStrings.ErrorLoadingAssembly,
assemblyName);
}
else
{
return new StringConstantExpressionAst(name.Extent, assemblyFileName, name.StringConstantType);
}
}
return name;
}
private StatementAst MethodDeclarationRule(Token functionNameToken, string className, bool isStaticMethod)
{
// G method-statement:
// G new-lines:opt function-name function-parameter-declaration base-ctor-call:opt '{' script-block '}'
// G
// G base-ctor-call: // can be present only if function-name == className
// G ':' new-lines:opt 'base' new-lines:opt parenthesized-expression new-lines:opt
// G
// G function-name:
// G command-argument
var functionName = functionNameToken.Text;
List<ParameterAst> parameters;
Token lParen = this.PeekToken();
IScriptExtent endErrorStatement = null;
Token rParen = null;
if (lParen.Kind == TokenKind.LParen)
{
parameters = this.FunctionParameterDeclarationRule(out endErrorStatement, out rParen);
}
else
{
this.ReportIncompleteInput(After(functionNameToken),
nameof(ParserStrings.MissingMethodParameterList),
ParserStrings.MissingMethodParameterList);
parameters = new List<ParameterAst>();
}
bool isCtor = functionName.Equals(className, StringComparison.OrdinalIgnoreCase);
List<ExpressionAst> baseCtorCallParams = null;
Token baseToken = null;
IScriptExtent baseCallLastExtent = null;
TokenizerMode oldTokenizerMode;
if (isCtor && !isStaticMethod)
{
this.SkipNewlines();
oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Signature);
Token colonToken = PeekToken();
if (colonToken.Kind == TokenKind.Colon)
{
SkipToken();
this.SkipNewlines();
baseToken = PeekToken();
if (baseToken.Kind == TokenKind.Base)
{
SkipToken();
this.SkipNewlines();
lParen = PeekToken();
if (lParen.Kind == TokenKind.LParen)
{
SkipToken();
// we don't allow syntax
// : base{ script }
// as a short for
// : base( { script } )
baseCtorCallParams = InvokeParamParenListRule(lParen, out baseCallLastExtent);
this.SkipNewlines();
}
else
{
endErrorStatement = baseToken.Extent;
ReportIncompleteInput(After(baseToken),
nameof(ParserStrings.MissingMethodParameterList),
ParserStrings.MissingMethodParameterList);
}
}
else
{
endErrorStatement = colonToken.Extent;
ReportIncompleteInput(After(colonToken),
nameof(ParserStrings.MissingBaseCtorCall),
ParserStrings.MissingBaseCtorCall);
}
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
baseCtorCallParams ??=
// Assuming implicit default ctor
new List<ExpressionAst>();
}
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: If there is no opening curly, assume it hasn't been entered yet and don't consume anything.
UngetToken(lCurly);
if (endErrorStatement == null)
{
endErrorStatement = ExtentFromFirstOf(rParen, functionNameToken);
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingFunctionBody),
ParserStrings.MissingFunctionBody);
}
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(functionNameToken, endErrorStatement), parameters);
}
StatementAst baseCtorCallStatement = null;
if (isCtor && !isStaticMethod)
{
IScriptExtent baseCallExtent;
IScriptExtent baseKeywordExtent;
if (baseToken != null)
{
baseCallExtent = ExtentOf(baseToken, baseCallLastExtent);
baseKeywordExtent = baseToken.Extent;
}
else
{
baseCallExtent = PositionUtilities.EmptyExtent;
baseKeywordExtent = PositionUtilities.EmptyExtent;
}
var invokeMemberAst = new BaseCtorInvokeMemberExpressionAst(baseKeywordExtent, baseCallExtent, baseCtorCallParams);
baseCtorCallStatement = new CommandExpressionAst(invokeMemberAst.Extent, invokeMemberAst, null);
}
oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Command);
ScriptBlockAst scriptBlock = ScriptBlockRule(lCurly, false, baseCtorCallStatement);
var result = new FunctionDefinitionAst(ExtentOf(functionNameToken, scriptBlock),
/*isFilter:*/false, /*isWorkflow:*/false, functionNameToken, parameters, scriptBlock);
return result;
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private StatementAst FunctionDeclarationRule(Token functionToken)
{
// G function-statement:
// G 'function' new-lines:opt function-name function-parameter-declaration:opt '{' script-block '}'
// G 'filter' new-lines:opt function-name function-parameter-declaration:opt '{' script-block '}'
// G 'workflow' new-lines:opt function-name function-parameter-declaration:opt '{' script-block '}'
// G
// G function-name:
// G command-argument
SkipNewlines();
Token functionNameToken = NextToken();
switch (functionNameToken.Kind)
{
// The following are not allowed as function names, anything else is allowed:
// The general rule is - punctuators and unary operators are not allowed
// We want to allow as much as reasonably possible, e.g. if it could be a
// valid command name, we want to allow it with very limited exceptions.
case TokenKind.Pipe:
case TokenKind.LParen:
case TokenKind.LCurly:
case TokenKind.AtParen:
case TokenKind.AtCurly:
case TokenKind.RCurly:
case TokenKind.RParen:
case TokenKind.EndOfInput:
case TokenKind.Semi:
case TokenKind.Redirection:
case TokenKind.RedirectInStd:
case TokenKind.AndAnd:
case TokenKind.OrOr:
case TokenKind.Ampersand:
case TokenKind.Variable:
case TokenKind.SplattedVariable:
case TokenKind.HereStringExpandable:
case TokenKind.HereStringLiteral:
case TokenKind.StringLiteral:
case TokenKind.StringExpandable:
// ErrorRecovery: Don't continue, assume the function isn't there yet.
UngetToken(functionNameToken);
ReportIncompleteInput(After(functionToken),
nameof(ParserStrings.MissingNameAfterKeyword),
ParserStrings.MissingNameAfterKeyword,
functionToken.Text);
return new ErrorStatementAst(functionToken.Extent);
}
// A function name that matches a keyword isn't really a keyword, so don't color it that way
functionNameToken.TokenFlags &= ~TokenFlags.Keyword;
IScriptExtent endErrorStatement;
Token rParen;
var parameters = this.FunctionParameterDeclarationRule(out endErrorStatement, out rParen);
Token lCurly = NextToken();
if (lCurly.Kind != TokenKind.LCurly)
{
// ErrorRecovery: If there is no opening curly, assume it hasn't been entered yet and don't consume anything.
UngetToken(lCurly);
if (endErrorStatement == null)
{
endErrorStatement = ExtentFromFirstOf(rParen, functionNameToken);
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingFunctionBody),
ParserStrings.MissingFunctionBody);
}
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(functionToken, endErrorStatement), parameters);
}
bool isFilter = functionToken.Kind == TokenKind.Filter;
bool isWorkflow = functionToken.Kind == TokenKind.Workflow;
bool oldTokenizerWorkflowContext = _tokenizer.InWorkflowContext;
try
{
_tokenizer.InWorkflowContext = isWorkflow;
ScriptBlockAst scriptBlock = ScriptBlockRule(lCurly, isFilter);
var functionName = (functionNameToken.Kind == TokenKind.Generic)
? ((StringToken)functionNameToken).Value
: functionNameToken.Text;
FunctionDefinitionAst result = new FunctionDefinitionAst(ExtentOf(functionToken, scriptBlock),
isFilter, isWorkflow, functionNameToken, parameters, scriptBlock);
return result;
}
finally
{
_tokenizer.InWorkflowContext = oldTokenizerWorkflowContext;
}
}
private List<ParameterAst> FunctionParameterDeclarationRule(out IScriptExtent endErrorStatement, out Token rParen)
{
// G function-parameter-declaration:
// G new-lines:opt '(' parameter-list new-lines:opt ')'
List<ParameterAst> parameters = null;
endErrorStatement = null;
SkipNewlines();
rParen = null;
Token lParen = PeekToken();
if (lParen.Kind == TokenKind.LParen)
{
SkipToken();
parameters = ParameterListRule();
SkipNewlines();
rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: assume a body follows, so just keep parsing.
UngetToken(rParen);
endErrorStatement = parameters.Count > 0 ? parameters.Last().Extent : lParen.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingEndParenthesisInFunctionParameterList),
ParserStrings.MissingEndParenthesisInFunctionParameterList);
}
SkipNewlines();
}
return parameters;
}
private StatementAst TrapStatementRule(Token trapToken)
{
// G trap-statement:
// G 'trap' new-lines:opt type-literal:opt statement-block
var restorePoint = _tokenizer.GetRestorePoint();
SkipNewlines();
AttributeBaseAst type = AttributeRule();
var typeConstraintAst = type as TypeConstraintAst;
if (type != null && typeConstraintAst == null)
{
// Presumably we parsed an attribute instead of a type. Put it back and let the code
// below report a missing trap body. The attribute might belong to something that
// is after the trap, something like:
// trap
// [ValidateNotNullOrEmpty()]$var
Resync(restorePoint);
}
var body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: just return an error statement.
IScriptExtent endErrorStatement = ExtentFromFirstOf(typeConstraintAst, trapToken);
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingTrapStatement),
ParserStrings.MissingTrapStatement);
return new ErrorStatementAst(ExtentOf(trapToken, endErrorStatement), GetNestedErrorAsts(typeConstraintAst));
}
return new TrapStatementAst(ExtentOf(trapToken, body), typeConstraintAst, body);
}
/// <summary>
/// Parse a catch block.
/// </summary>
/// <param name="endErrorStatement">
/// Set to the last thing scanned that is definitely part of the catch, but only set after issuing an error.
/// </param>
/// <param name="errorAsts">
/// If there are any errors and CatchBlockRule is returning null, this list is used to return back any asts
/// consumed here (essentially the type constraints.)
/// </param>
/// <returns>A catch clause, or null there is no catch or there was some error.</returns>
private CatchClauseAst CatchBlockRule(ref IScriptExtent endErrorStatement, ref List<TypeConstraintAst> errorAsts)
{
// G catch-clause:
// G new-lines:opt 'catch' catch-type-list:opt statement-block
// G catch-type-list:
// G new-lines:opt type-literal
// G catch-type-list new-lines:opt ',' new-lines:opt type-literal
SkipNewlines();
Token catchToken = NextToken();
if (catchToken.Kind != TokenKind.Catch)
{
UngetToken(catchToken);
return null;
}
List<TypeConstraintAst> exceptionTypes = null;
Token commaToken = null;
while (true)
{
var restorePoint = _tokenizer.GetRestorePoint();
SkipNewlines();
AttributeBaseAst type = AttributeRule();
if (type == null)
{
if (commaToken != null)
{
// ErrorRecovery: Just consume the comma, pretend it wasn't there and look for the handler block.
endErrorStatement = commaToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingTypeLiteralToken),
ParserStrings.MissingTypeLiteralToken);
}
break;
}
var typeConstraintAst = type as TypeConstraintAst;
if (typeConstraintAst == null)
{
// Presumably we parsed an attribute instead of a type. Put it back and let the code
// below report a missing catch clause body. The attribute might belong to something that
// is after the try/catch, something like:
// catch
// [ValidateNotNullOrEmpty()]$var
Resync(restorePoint);
break;
}
exceptionTypes ??= new List<TypeConstraintAst>();
exceptionTypes.Add(typeConstraintAst);
SkipNewlines();
commaToken = PeekToken();
if (commaToken.Kind != TokenKind.Comma)
{
break;
}
SkipToken();
}
StatementBlockAst handler = StatementBlockRule();
if (handler == null)
{
// Avoid issuing an extra error if we've already issued one, plus the end of the ErrorStatementAst
// shouldn't be updated.
if (commaToken == null || endErrorStatement != commaToken.Extent)
{
// ErrorRecovery: just use the "missing" block in the result ast.
endErrorStatement = exceptionTypes != null ? exceptionTypes.Last().Extent : catchToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingCatchHandlerBlock),
ParserStrings.MissingCatchHandlerBlock);
}
if (exceptionTypes != null)
{
if (errorAsts == null)
{
errorAsts = exceptionTypes;
}
else
{
errorAsts.AddRange(exceptionTypes);
}
}
return null;
}
return new CatchClauseAst(ExtentOf(catchToken, handler), exceptionTypes, handler);
}
private StatementAst TryStatementRule(Token tryToken)
{
// G try-statement:
// G 'try' statement-block catch-clauses
// G 'try' statement-block finally-clause
// G 'try' statement-block catch-clauses finally-clause
// G catch-clauses:
// G catch-clause
// G catch-clauses catch-clause
// G catch-clause:
// G new-lines:opt 'catch' catch-type-list:opt statement-block
// G catch-type-list:
// G new-lines:opt type-literal
// G catch-type-list new-lines:opt ',' new-lines:opt type-literal
// G finally-clause:
// G new-lines:opt 'finally' statement-block
SkipNewlines();
StatementBlockAst body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: don't parse more, return an error statement.
ReportIncompleteInput(After(tryToken),
nameof(ParserStrings.MissingTryStatementBlock),
ParserStrings.MissingTryStatementBlock);
return new ErrorStatementAst(tryToken.Extent);
}
IScriptExtent endErrorStatement = null;
CatchClauseAst catchClause;
List<CatchClauseAst> catches = new List<CatchClauseAst>();
List<TypeConstraintAst> errorAsts = null;
while ((catchClause = CatchBlockRule(ref endErrorStatement, ref errorAsts)) != null)
{
catches.Add(catchClause);
}
SkipNewlines();
Token finallyToken = PeekToken();
StatementBlockAst finallyBlock = null;
if (finallyToken.Kind == TokenKind.Finally)
{
SkipToken();
finallyBlock = StatementBlockRule();
if (finallyBlock == null)
{
// ErrorRecovery: just return null, a null finally block is fine in the resulting ast (but maybe consider
// marking the resulting ast as having an error.)
endErrorStatement = finallyToken.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingFinallyStatementBlock),
ParserStrings.MissingFinallyStatementBlock,
finallyToken.Kind.Text());
}
}
if (catches.Count == 0 && finallyBlock == null && endErrorStatement == null)
{
// ErrorRecovery: don't parse more, return an error statement.
endErrorStatement = body.Extent;
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingCatchOrFinally),
ParserStrings.MissingCatchOrFinally);
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(tryToken, endErrorStatement), GetNestedErrorAsts(body, catches, errorAsts));
}
return new TryStatementAst(ExtentOf(tryToken, finallyBlock != null ? finallyBlock.Extent : catches.Last().Extent),
body, catches, finallyBlock);
}
private StatementAst DataStatementRule(Token dataToken)
{
// G data-statement:
// G 'data' new-lines:opt data-name data-commands-allowed:opt statement-block
// G data-name:
// G simple-name
// G data-commands-allowed:
// G new-lines:opt '-supportedcommand' data-commands-list
// G data-commands-list:
// G new-lines:opt data-command
// G data-commands-list ',' new-lines:opt data-command
// G data-command:
// G command-name-expr
IScriptExtent endErrorStatement = null;
SkipNewlines();
var dataVariableNameAst = SimpleNameRule();
string dataVariableName = dataVariableNameAst?.Value;
SkipNewlines();
Token supportedCommandToken = PeekToken();
List<ExpressionAst> commands = null;
if (supportedCommandToken.Kind == TokenKind.Parameter)
{
SkipToken();
if (!IsSpecificParameter(supportedCommandToken, "SupportedCommand"))
{
// ErrorRecovery: Assume the parameter is just misspelled, look for command names next
endErrorStatement = supportedCommandToken.Extent;
ReportError(endErrorStatement,
nameof(ParserStrings.InvalidParameterForDataSectionStatement),
ParserStrings.InvalidParameterForDataSectionStatement,
((ParameterToken)supportedCommandToken).ParameterName);
}
Token commaToken = null;
commands = new List<ExpressionAst>();
while (true)
{
SkipNewlines();
ExpressionAst ast = GetSingleCommandArgument(CommandArgumentContext.CommandName);
if (ast == null)
{
// ErrorRecovery: Look for the data statement body.
// Only report an error if an error hasn't already been issued.
if (endErrorStatement == null)
{
ReportIncompleteInput(After(commaToken ?? supportedCommandToken),
nameof(ParserStrings.MissingValueForSupportedCommandInDataSectionStatement),
ParserStrings.MissingValueForSupportedCommandInDataSectionStatement);
}
endErrorStatement = commaToken != null ? commaToken.Extent : supportedCommandToken.Extent;
break;
}
commands.Add(ast);
commaToken = PeekToken();
if (commaToken.Kind == TokenKind.Comma)
{
SkipToken();
continue;
}
break;
}
}
StatementBlockAst body = null;
if (endErrorStatement == null)
{
body = StatementBlockRule();
if (body == null)
{
// ErrorRecovery: return an error statement.
endErrorStatement = commands != null
? commands.Last().Extent
: ExtentFromFirstOf(dataVariableNameAst, dataToken);
ReportIncompleteInput(After(endErrorStatement),
nameof(ParserStrings.MissingStatementBlockForDataSection),
ParserStrings.MissingStatementBlockForDataSection);
}
}
if (endErrorStatement != null)
{
return new ErrorStatementAst(ExtentOf(dataToken, endErrorStatement), GetNestedErrorAsts(commands));
}
return new DataStatementAst(ExtentOf(dataToken, body), dataVariableName, commands, body);
}
#endregion Statements
#region Pipelines
private PipelineBaseAst PipelineChainRule()
{
// G pipeline-chain:
// G pipeline
// G pipeline-chain chain-operator pipeline
// G
// G chain-operator:
// G '||'
// G '&&'
// If this feature is not enabled,
// just look for pipelines as before.
RuntimeHelpers.EnsureSufficientExecutionStack();
// First look for assignment, since PipelineRule once handled that and this supersedes that.
// We may end up with an expression here as a result,
// in which case we hang on to it to pass it into the first pipeline rule call.
Token assignToken = null;
ExpressionAst expr;
// Look for an expression,
// which could either be a variable to assign to,
// or the first expression in a pipeline: "Hi" | % { $_ }
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
expr = ExpressionRule();
if (expr != null)
{
// We peek here because we are in expression mode, otherwise =0 will get scanned
// as a single token.
var token = PeekToken();
if (token.Kind.HasTrait(TokenFlags.AssignmentOperator))
{
SkipToken();
assignToken = token;
}
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
// If we have an assign token, deal with assignment
if (expr != null && assignToken != null)
{
SkipNewlines();
StatementAst statement = StatementRule();
if (statement == null)
{
// ErrorRecovery: we are very likely at EOF because pretty much anything should result in some
// pipeline, so just keep parsing.
IScriptExtent errorExtent = After(assignToken);
ReportIncompleteInput(
errorExtent,
nameof(ParserStrings.ExpectedValueExpression),
ParserStrings.ExpectedValueExpression,
assignToken.Kind.Text());
statement = new ErrorStatementAst(errorExtent);
}
return new AssignmentStatementAst(
ExtentOf(expr, statement),
expr,
assignToken.Kind,
statement,
assignToken.Extent);
}
// Start scanning for a pipeline chain,
// possibly starting with the expression from earlier
ExpressionAst startExpression = expr;
ChainableAst currentPipelineChain = null;
Token currentChainOperatorToken = null;
Token nextToken = null;
bool background = false;
while (true)
{
// Look for the next pipeline in the chain,
// at this point we should already have parsed all assignments
// in enclosing calls to PipelineChainRule
PipelineAst nextPipeline;
Token firstPipelineToken = null;
if (startExpression != null)
{
nextPipeline = (PipelineAst)PipelineRule(startExpression);
startExpression = null;
}
else
{
// Remember the token for error reporting,
// since erroneous results in the rules still consume tokens
firstPipelineToken = PeekToken();
nextPipeline = (PipelineAst)PipelineRule();
}
if (nextPipeline == null)
{
if (currentChainOperatorToken == null)
{
// We haven't seen a chain token, so the caller is responsible
// for expecting a pipeline and must manage this
return null;
}
// See if we are responsible for reporting the issue
switch (firstPipelineToken.Kind)
{
case TokenKind.EndOfInput:
// If we're at EOF, we should allow input to complete
ReportIncompleteInput(
After(currentChainOperatorToken),
nameof(ParserStrings.EmptyPipelineChainElement),
ParserStrings.EmptyPipelineChainElement,
currentChainOperatorToken.Text);
break;
case TokenKind.Dot:
case TokenKind.Ampersand:
// If something like 'command && &' or 'command && .' was provided,
// CommandRule has already reported the error.
break;
default:
ReportError(
ExtentOf(currentChainOperatorToken, firstPipelineToken),
nameof(ParserStrings.EmptyPipelineChainElement),
ParserStrings.EmptyPipelineChainElement,
currentChainOperatorToken.Text);
break;
}
return new ErrorStatementAst(
ExtentOf(currentPipelineChain, currentChainOperatorToken),
currentChainOperatorToken,
new[] { currentPipelineChain });
}
// Look ahead for a chain operator
nextToken = PeekToken();
switch (nextToken.Kind)
{
case TokenKind.AndAnd:
case TokenKind.OrOr:
SkipToken();
SkipNewlines();
break;
// Background operators may also occur here
case TokenKind.Ampersand:
SkipToken();
nextToken = PeekToken();
switch (nextToken.Kind)
{
case TokenKind.AndAnd:
case TokenKind.OrOr:
SkipToken();
ReportError(nextToken.Extent, nameof(ParserStrings.BackgroundOperatorInPipelineChain), ParserStrings.BackgroundOperatorInPipelineChain);
return new ErrorStatementAst(ExtentOf(currentPipelineChain ?? nextPipeline, nextToken.Extent));
}
background = true;
goto default;
// No more chain operators -- return
default:
// If we haven't seen a chain yet, pass through the pipeline
// Simplifies the AST and prevents allocation
if (currentPipelineChain == null)
{
if (!background)
{
return nextPipeline;
}
// Set background on the pipeline AST
nextPipeline.Background = true;
return nextPipeline;
}
return new PipelineChainAst(
ExtentOf(currentPipelineChain.Extent, nextPipeline.Extent),
currentPipelineChain,
nextPipeline,
currentChainOperatorToken.Kind,
background);
}
// Assemble the new chain statement AST
currentPipelineChain = currentPipelineChain == null
? (ChainableAst)nextPipeline
: new PipelineChainAst(
ExtentOf(currentPipelineChain.Extent, nextPipeline.Extent),
currentPipelineChain,
nextPipeline,
currentChainOperatorToken.Kind);
// Remember the last operator to chain the coming pipeline
currentChainOperatorToken = nextToken;
// Look ahead to report incomplete input if needed
if (PeekToken().Kind == TokenKind.EndOfInput)
{
ReportIncompleteInput(
After(nextToken),
nameof(ParserStrings.EmptyPipelineChainElement),
ParserStrings.EmptyPipelineChainElement);
return currentPipelineChain;
}
}
}
private PipelineBaseAst PipelineRule(
ExpressionAst startExpression = null,
bool allowBackground = false)
{
// G pipeline:
// G assignment-expression
// G expression redirections:opt pipeline-tail:opt
// G command pipeline-tail:opt
// G
// G assignment-expression:
// G expression assignment-operator statement
// G
// G pipeline-tail:
// G new-lines:opt '|' new-lines:opt command pipeline-tail:opt
//
var pipelineElements = new List<CommandBaseAst>();
IScriptExtent startExtent = null;
Token nextToken = null;
bool scanning = true;
bool background = false;
ExpressionAst expr = startExpression;
while (scanning)
{
CommandBaseAst commandAst;
if (expr == null)
{
// Look for an expression at the beginning of a pipeline
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
expr = ExpressionRule();
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
if (expr != null)
{
if (pipelineElements.Count > 0)
{
// ErrorRecovery: this is a semantic error, so just keep parsing.
ReportError(
expr.Extent,
nameof(ParserStrings.ExpressionsMustBeFirstInPipeline),
ParserStrings.ExpressionsMustBeFirstInPipeline);
}
RedirectionAst[] redirections = null;
var redirectionToken = PeekToken() as RedirectionToken;
RedirectionAst lastRedirection = null;
while (redirectionToken != null)
{
SkipToken();
redirections ??= new RedirectionAst[CommandBaseAst.MaxRedirections];
IScriptExtent unused = null;
lastRedirection = RedirectionRule(redirectionToken, redirections, ref unused);
redirectionToken = PeekToken() as RedirectionToken;
}
var exprExtent = lastRedirection != null ? ExtentOf(expr, lastRedirection) : expr.Extent;
commandAst = new CommandExpressionAst(
exprExtent,
expr,
redirections?.Where(static r => r != null));
}
else
{
commandAst = (CommandAst)CommandRule(forDynamicKeyword: false);
}
if (commandAst != null)
{
startExtent ??= commandAst.Extent;
pipelineElements.Add(commandAst);
}
else if (pipelineElements.Count > 0 || PeekToken().Kind == TokenKind.Pipe)
{
// ErrorRecovery: just fall through
// If the first pipe element is null, the position points to the pipe (ideally it would
// point before, but the pipe could be the first character), otherwise the empty element
// is after the pipe character.
IScriptExtent errorPosition = nextToken != null ? After(nextToken) : PeekToken().Extent;
ReportIncompleteInput(
errorPosition,
nameof(ParserStrings.EmptyPipeElement),
ParserStrings.EmptyPipeElement);
}
// Reset the expression for the next loop
expr = null;
nextToken = PeekToken();
// Skip newlines before pipe tokens to support (pipe)line continuation when pipe
// tokens start the next line of script
if (nextToken.Kind == TokenKind.NewLine && _tokenizer.IsPipeContinuation(nextToken.Extent))
{
SkipNewlines();
nextToken = PeekToken();
}
switch (nextToken.Kind)
{
case TokenKind.Semi:
case TokenKind.NewLine:
case TokenKind.RParen:
case TokenKind.RCurly:
case TokenKind.EndOfInput:
// Handled by invoking rule
scanning = false;
continue;
case TokenKind.AndAnd:
case TokenKind.OrOr:
scanning = false;
continue;
case TokenKind.Ampersand:
if (!allowBackground)
{
// Handled by invoking rule
scanning = false;
continue;
}
SkipToken();
scanning = false;
background = true;
break;
case TokenKind.Pipe:
SkipToken();
SkipNewlines();
if (PeekToken().Kind == TokenKind.EndOfInput)
{
scanning = false;
ReportIncompleteInput(
After(nextToken),
nameof(ParserStrings.EmptyPipeElement),
ParserStrings.EmptyPipeElement);
}
break;
default:
// ErrorRecovery: don't eat the token, assume it belongs to something else.
ReportError(
nextToken.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
nextToken.Text);
scanning = false;
break;
}
}
if (pipelineElements.Count == 0)
{
return null;
}
return new PipelineAst(ExtentOf(startExtent, pipelineElements[pipelineElements.Count - 1]), pipelineElements, background);
}
private RedirectionAst RedirectionRule(RedirectionToken redirectionToken, RedirectionAst[] redirections, ref IScriptExtent extent)
{
// G redirections:
// G redirection
// G redirections redirection
// G redirection:
// G merging-redirection-operator
// G file-redirection-operator redirected-file-name
// G redirected-file-name:
// G command-argument
// G primary-expression
RedirectionAst result;
var fileRedirectionToken = redirectionToken as FileRedirectionToken;
if (fileRedirectionToken != null || (redirectionToken is InputRedirectionToken))
{
// get location
var filename = GetSingleCommandArgument(CommandArgumentContext.FileName);
if (filename == null)
{
// ErrorRecovery: Just pretend we have a filename and continue parsing.
ReportError(After(redirectionToken),
nameof(ParserStrings.MissingFileSpecification),
ParserStrings.MissingFileSpecification);
filename = new ErrorExpressionAst(redirectionToken.Extent);
}
if (fileRedirectionToken == null)
{
// Must be an input redirection
ReportError(redirectionToken.Extent,
nameof(ParserStrings.RedirectionNotSupported),
ParserStrings.RedirectionNotSupported,
redirectionToken.Text);
extent = ExtentOf(redirectionToken, filename);
return null;
}
result = new FileRedirectionAst(ExtentOf(fileRedirectionToken, filename), fileRedirectionToken.FromStream,
filename, fileRedirectionToken.Append);
}
else
{
var mergingRedirectionToken = (MergingRedirectionToken)redirectionToken;
RedirectionStream fromStream = mergingRedirectionToken.FromStream;
RedirectionStream toStream = mergingRedirectionToken.ToStream;
if (toStream != RedirectionStream.Output)
{
// Have we seen something like 1>&2 or 2>&3
// ErrorRecovery: This is just a semantic error, so no special recovery.
ReportError(redirectionToken.Extent,
nameof(ParserStrings.RedirectionNotSupported),
ParserStrings.RedirectionNotSupported,
mergingRedirectionToken.Text);
toStream = RedirectionStream.Output;
}
else if (fromStream == toStream)
{
// Make sure 1>&1, 2>&2, etc. is an error.
// ErrorRecovery: This is just a semantic error, so no special recovery.
ReportError(redirectionToken.Extent,
nameof(ParserStrings.RedirectionNotSupported),
ParserStrings.RedirectionNotSupported,
mergingRedirectionToken.Text);
}
result = new MergingRedirectionAst(mergingRedirectionToken.Extent, mergingRedirectionToken.FromStream, toStream);
}
if (redirections[(int)result.FromStream] == null)
{
redirections[(int)result.FromStream] = result;
}
else
{
string errorStream;
switch (result.FromStream)
{
case RedirectionStream.All: errorStream = ParserStrings.AllStream; break;
case RedirectionStream.Output: errorStream = ParserStrings.OutputStream; break;
case RedirectionStream.Error: errorStream = ParserStrings.ErrorStream; break;
case RedirectionStream.Warning: errorStream = ParserStrings.WarningStream; break;
case RedirectionStream.Verbose: errorStream = ParserStrings.VerboseStream; break;
case RedirectionStream.Debug: errorStream = ParserStrings.DebugStream; break;
case RedirectionStream.Information: errorStream = ParserStrings.InformationStream; break;
default:
throw PSTraceSource.NewArgumentOutOfRangeException("result.FromStream", result.FromStream);
}
ReportError(result.Extent,
nameof(ParserStrings.StreamAlreadyRedirected),
ParserStrings.StreamAlreadyRedirected,
errorStream);
}
extent = result.Extent;
return result;
}
private ExpressionAst GetSingleCommandArgument(CommandArgumentContext context)
{
if (PeekToken().Kind == TokenKind.Comma || PeekToken().Kind == TokenKind.EndOfInput)
{
return null;
}
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Command);
return GetCommandArgument(context, NextToken());
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
[Flags]
private enum CommandArgumentContext
{
CommandName = 0x01,
CommandNameAfterInvocationOperator = 0x02 | CommandName,
FileName = 0x04,
CommandArgument = 0x08,
SwitchCondition = 0x10,
}
private ExpressionAst GetCommandArgument(CommandArgumentContext context, Token token)
{
Diagnostics.Assert(token.Kind != TokenKind.Comma, "A unary comma is an error in command mode, and should have already been reported.");
ExpressionAst exprAst;
List<ExpressionAst> commandArgs = null;
Token commaToken = null;
bool foundVerbatimArgument = false;
while (true)
{
switch (token.Kind)
{
// The following tokens are never allowed as command arguments.
case TokenKind.Pipe:
case TokenKind.RCurly:
case TokenKind.RParen:
case TokenKind.EndOfInput:
case TokenKind.NewLine:
case TokenKind.Semi:
case TokenKind.Redirection:
case TokenKind.RedirectInStd:
case TokenKind.AndAnd:
case TokenKind.OrOr:
case TokenKind.Ampersand:
case TokenKind.MinusMinus:
case TokenKind.Comma:
UngetToken(token);
// If we haven't seen an argument, the caller must issue an error. If we've seen at least one
// argument, then we will issue the error and return back the arguments seen so far.
if (commaToken == null)
{
return null;
}
// ErrorRecovery: stop looking for additional arguments, exclude the trailing comma
ReportIncompleteInput(After(commaToken),
nameof(ParserStrings.MissingExpression),
ParserStrings.MissingExpression,
",");
return new ErrorExpressionAst(ExtentOf(commandArgs[0], commaToken), commandArgs);
case TokenKind.SplattedVariable:
case TokenKind.Variable:
case TokenKind.Number:
case TokenKind.HereStringExpandable:
case TokenKind.StringExpandable:
case TokenKind.HereStringLiteral:
case TokenKind.StringLiteral:
case TokenKind.LParen:
case TokenKind.DollarParen:
case TokenKind.AtParen:
case TokenKind.AtCurly:
case TokenKind.LCurly:
UngetToken(token);
exprAst = PrimaryExpressionRule(withMemberAccess: true);
Diagnostics.Assert(exprAst != null, "PrimaryExpressionRule should never return null");
break;
case TokenKind.Generic:
if ((context & CommandArgumentContext.CommandName) != 0)
{
token.TokenFlags |= TokenFlags.CommandName;
}
var genericToken = (StringToken)token;
var expandableToken = genericToken as StringExpandableToken;
// A command name w/o invocation operator is not expandable even if the token has expandable parts
// If we have seen an invocation operator, the command name is expandable.
if (expandableToken != null && context != CommandArgumentContext.CommandName)
{
var nestedExpressions = ParseNestedExpressions(expandableToken);
exprAst = new ExpandableStringExpressionAst(expandableToken, expandableToken.Value,
expandableToken.FormatString, nestedExpressions);
}
else
{
exprAst = new StringConstantExpressionAst(genericToken.Extent, genericToken.Value, StringConstantType.BareWord);
// If this is a verbatim argument, then don't continue peeking
if (string.Equals(genericToken.Value, VERBATIM_ARGUMENT, StringComparison.OrdinalIgnoreCase))
{
foundVerbatimArgument = true;
}
}
break;
default:
exprAst = new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord);
// A command/argument that matches a keyword isn't really a keyword, so don't color it that way
token.TokenFlags &= ~TokenFlags.Keyword;
switch (context)
{
case CommandArgumentContext.CommandName:
case CommandArgumentContext.CommandNameAfterInvocationOperator:
token.TokenFlags |= TokenFlags.CommandName;
break;
case CommandArgumentContext.FileName:
case CommandArgumentContext.CommandArgument:
case CommandArgumentContext.SwitchCondition:
token.SetIsCommandArgument();
break;
}
break;
}
if (context != CommandArgumentContext.CommandArgument)
{
break;
}
if (foundVerbatimArgument)
{
break;
}
token = PeekToken();
if (token.Kind != TokenKind.Comma)
{
break;
}
commaToken = token;
commandArgs ??= new List<ExpressionAst>();
commandArgs.Add(exprAst);
SkipToken();
SkipNewlines();
token = NextToken();
}
Diagnostics.Assert(commandArgs != null || exprAst != null, "How did that happen?");
if (commandArgs != null)
{
commandArgs.Add(exprAst);
return new ArrayLiteralAst(ExtentOf(commandArgs[0], commandArgs[commandArgs.Count - 1]), commandArgs);
}
return exprAst;
}
internal Ast CommandRule(bool forDynamicKeyword)
{
// G command:
// G command-name command-elements:opt
// G command-invocation-operator command-module:opt command-name-expr command-elements:opt
// G command-invocation-operator: one of
// G '&' '.'
// G command-module:
// G primary-expression
// G command-name:
// G generic-token
// G generic-token-with-subexpr
// G generic-token-with-subexpr: No whitespace is allowed between ) and command-name.
// G generic-token-with-subexpr-start statement-list:opt ) command-name
// G command-name-expr:
// G command-name
// G primary-expression
// G command-elements:
// G command-element
// G command-elements command-element
// G command-element:
// G command-parameter
// G command-argument
// G redirection
// G command-argument:
// G command-name-expr
Token firstToken;
bool dotSource, ampersand;
bool sawDashDash = false;
bool foundVerbatimArgument = false;
RedirectionAst[] redirections = null;
IScriptExtent endExtent;
var elements = new List<CommandElementAst>();
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Command);
Token token = NextToken();
firstToken = token;
endExtent = token.Extent;
dotSource = false;
ampersand = false;
CommandArgumentContext context;
if (token.Kind == TokenKind.Dot)
{
dotSource = true;
token = NextToken();
context = CommandArgumentContext.CommandNameAfterInvocationOperator;
}
else if (token.Kind == TokenKind.Ampersand)
{
ampersand = true;
token = NextToken();
context = CommandArgumentContext.CommandNameAfterInvocationOperator;
}
else
{
context = CommandArgumentContext.CommandName;
}
bool scanning = true;
while (scanning)
{
switch (token.Kind)
{
case TokenKind.Pipe:
case TokenKind.RCurly:
case TokenKind.RParen:
case TokenKind.EndOfInput:
case TokenKind.NewLine:
case TokenKind.Semi:
case TokenKind.AndAnd:
case TokenKind.OrOr:
case TokenKind.Ampersand:
UngetToken(token);
scanning = false;
continue;
case TokenKind.MinusMinus:
endExtent = token.Extent;
// Add the first -- as a parameter, which is then ignored when constructing the command processor unless it's a native
// command. All subsequent -- are added as arguments.
elements.Add(sawDashDash
? (CommandElementAst)new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord)
: new CommandParameterAst(token.Extent, "-", null, token.Extent));
sawDashDash = true;
break;
case TokenKind.Comma:
endExtent = token.Extent;
ReportError(token.Extent,
nameof(ParserStrings.MissingArgument),
ParserStrings.MissingArgument);
SkipNewlines();
break;
case TokenKind.Parameter:
if ((context & CommandArgumentContext.CommandName) != 0 || sawDashDash)
{
endExtent = token.Extent;
token.TokenFlags |= TokenFlags.CommandName;
var commandName = new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord);
elements.Add(commandName);
break;
}
var parameterToken = (ParameterToken)token;
ExpressionAst parameterArgs;
IScriptExtent extent;
// If the next token is a comma, don't grab it as part of the argument. The next time through this
// loop will issue an error.
if (parameterToken.UsedColon && PeekToken().Kind != TokenKind.Comma)
{
parameterArgs = GetCommandArgument(CommandArgumentContext.CommandArgument, NextToken());
if (parameterArgs == null)
{
extent = parameterToken.Extent;
ReportError(After(extent),
nameof(ParserStrings.ParameterRequiresArgument),
ParserStrings.ParameterRequiresArgument,
parameterToken.Text);
}
else
{
extent = ExtentOf(token, parameterArgs);
}
}
else
{
parameterArgs = null;
extent = token.Extent;
}
endExtent = extent;
var paramAst = new CommandParameterAst(extent, parameterToken.ParameterName, parameterArgs, token.Extent);
elements.Add(paramAst);
break;
case TokenKind.Redirection:
case TokenKind.RedirectInStd:
if ((context & CommandArgumentContext.CommandName) == 0)
{
redirections ??= new RedirectionAst[CommandBaseAst.MaxRedirections];
RedirectionRule((RedirectionToken)token, redirections, ref endExtent);
}
else
{
// For backwards compatibility, we allow redirection operators as command names.
// V2 did not allow:
// & <<
// but V3 and on will because it falls out rather naturally here.
endExtent = token.Extent;
elements.Add(new StringConstantExpressionAst(token.Extent, token.Text, StringConstantType.BareWord));
}
break;
default:
if (token.Kind == TokenKind.InlineScript && context == CommandArgumentContext.CommandName)
{
scanning = InlineScriptRule(token, elements);
Diagnostics.Assert(elements.Count >= 1, "We should at least have the command name: inlinescript");
endExtent = elements.Last().Extent;
if (!scanning) { continue; }
}
else
{
var ast = GetCommandArgument(context, token);
// If this is the special verbatim argument syntax, look for the next element
StringToken argumentToken = token as StringToken;
if ((argumentToken != null) && string.Equals(argumentToken.Value, VERBATIM_ARGUMENT, StringComparison.OrdinalIgnoreCase))
{
elements.Add(ast);
endExtent = ast.Extent;
var verbatimToken = GetVerbatimCommandArgumentToken();
if (verbatimToken != null)
{
foundVerbatimArgument = true;
scanning = false;
ast = new StringConstantExpressionAst(verbatimToken.Extent, verbatimToken.Value, StringConstantType.BareWord);
elements.Add(ast);
endExtent = ast.Extent;
}
break;
}
endExtent = ast.Extent;
elements.Add(ast);
}
break;
}
if (!foundVerbatimArgument)
{
context = CommandArgumentContext.CommandArgument;
token = NextToken();
}
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
if (elements.Count == 0)
{
Diagnostics.Assert(!forDynamicKeyword, "DynamicKeyword should have command name at least");
if (dotSource || ampersand)
{
IScriptExtent extent = firstToken.Extent;
ReportError(extent,
nameof(ParserStrings.MissingExpression),
ParserStrings.MissingExpression,
firstToken.Text);
}
return null;
}
if (forDynamicKeyword)
{
// TODO: report error if any redirections
return new DynamicKeywordStatementAst(ExtentOf(firstToken, endExtent), elements);
}
return new CommandAst(ExtentOf(firstToken, endExtent), elements,
dotSource || ampersand ? firstToken.Kind : TokenKind.Unknown,
redirections?.Where(static r => r != null));
}
#endregion Pipelines
#region Expressions
/// <summary>Parse an expression.</summary>
/// <param name="endNumberOnTernaryOpChars">
/// When it's known for sure that we are expecting an expression, allowing a generic token like '12?' or '12:' is
/// not useful. In those cases, we force to start a new token upon seeing '?' and ':' when scanning for a number
/// by setting this parameter to true, hoping to find a ternary expression.
/// </param>
private ExpressionAst ExpressionRule(bool endNumberOnTernaryOpChars = false)
{
// G expression:
// G logical-expression
// G
// G logical-expression:
// G binary-expression
// G ternary-expression
// G
// G ternary-expression:
// G binary-expression '?' new-lines:opt ternary-expression new-lines:opt ':' new-lines:opt ternary-expression
RuntimeHelpers.EnsureSufficientExecutionStack();
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
ExpressionAst condition = BinaryExpressionRule(endNumberOnTernaryOpChars);
if (condition == null)
{
return null;
}
Token token = PeekToken();
if (token.Kind != TokenKind.QuestionMark)
{
return condition;
}
SkipToken();
SkipNewlines();
// We have seen the ternary operator '?' and now expecting the 'IfTrue' expression.
ExpressionAst ifTrue = ExpressionRule(endNumberOnTernaryOpChars: true);
if (ifTrue == null)
{
// ErrorRecovery: create an error expression to fill out the ast and keep parsing.
IScriptExtent extent = After(token);
ReportIncompleteInput(
extent,
nameof(ParserStrings.ExpectedValueExpression),
ParserStrings.ExpectedValueExpression,
token.Text);
ifTrue = new ErrorExpressionAst(extent);
}
SkipNewlines();
token = NextToken();
if (token.Kind != TokenKind.Colon)
{
var componentAsts = new List<Ast>() { condition };
// ErrorRecovery: we have done the expression parsing and should try parsing something else.
UngetToken(token);
// Don't bother reporting this error if we already reported an empty 'IfTrue' operand error.
if (ifTrue is not ErrorExpressionAst)
{
componentAsts.Add(ifTrue);
ReportIncompleteInput(
token.Extent,
nameof(ParserStrings.MissingColonInTernaryExpression),
ParserStrings.MissingColonInTernaryExpression);
}
return new ErrorExpressionAst(ExtentOf(condition, Before(token)), componentAsts);
}
SkipNewlines();
ExpressionAst ifFalse = ExpressionRule(endNumberOnTernaryOpChars: true);
if (ifFalse == null)
{
// ErrorRecovery: create an error expression to fill out the ast and keep parsing.
IScriptExtent extent = After(token);
ReportIncompleteInput(
extent,
nameof(ParserStrings.ExpectedValueExpression),
ParserStrings.ExpectedValueExpression,
token.Text);
ifFalse = new ErrorExpressionAst(extent);
}
return new TernaryExpressionAst(ExtentOf(condition, ifFalse), condition, ifTrue, ifFalse);
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
/// <summary>Parse a binary expression.</summary>
/// <param name="endNumberOnTernaryOpChars">
/// When it's known for sure that we are expecting an expression, allowing a generic token like '12?' or '12:' is
/// not useful. In those cases, we force to start a new token upon seeing '?' and ':' when scanning for a number
/// by setting this parameter to true, hoping to find a ternary expression.
/// </param>
private ExpressionAst BinaryExpressionRule(bool endNumberOnTernaryOpChars = false)
{
// G binary-expression:
// G bitwise-expression
// G binary-expression '-and' new-lines:opt bitwise-expression
// G binary-expression '-or' new-lines:opt bitwise-expression
// G binary-expression '-xor' new-lines:opt bitwise-expression
// G
// G bitwise-expression:
// G comparison-expression
// G bitwise-expression '-band' new-lines:opt comparison-expression
// G bitwise-expression '-bor' new-lines:opt comparison-expression
// G bitwise-expression '-bxor' new-lines:opt comparison-expression
// G
// G comparison-expression:
// G nullcoalesce-expression
// G comparison-expression comparison-operator new-lines:opt nullcoalesce-expression
// G
// G nullcoalesce-expression:
// G additive-expression
// G nullcoalesce-expression '??' new-lines:opt additive-expression
// G
// G additive-expression:
// G multiplicative-expression
// G additive-expression '+' new-lines:opt multiplicative-expression
// G additive-expression dash new-lines:opt multiplicative-expression
// G
// G multiplicative-expression:
// G format-expression
// G multiplicative-expression '*' new-lines:opt format-expression
// G multiplicative-expression '/' new-lines:opt format-expression
// G multiplicative-expression '%' new-lines:opt format-expression
// G
// G format-expression:
// G range-expression
// G format-expression format-operator new-lines:opt range-expression
// G
// G range-expression:
// G array-literal-expression
// G range-expression '..' new-lines:opt array-literal-expression
RuntimeHelpers.EnsureSufficientExecutionStack();
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
ExpressionAst lhs, rhs;
ExpressionAst expr = ArrayLiteralRule(endNumberOnTernaryOpChars);
if (expr == null)
{
return null;
}
ParameterToken paramToken;
Token token = PeekToken();
if (!token.Kind.HasTrait(TokenFlags.BinaryOperator))
{
paramToken = token as ParameterToken;
if (paramToken != null)
{
return ErrorRecoveryParameterInExpression(paramToken, expr);
}
return expr;
}
else if (token.Kind == TokenKind.AndAnd || token.Kind == TokenKind.OrOr)
{
return expr;
}
SkipToken();
Stack<ExpressionAst> operandStack = new Stack<ExpressionAst>();
Stack<Token> operatorStack = new Stack<Token>();
operandStack.Push(expr);
operatorStack.Push(token);
int precedence = token.Kind.GetBinaryPrecedence();
while (true)
{
SkipNewlines();
// We have seen a binary operator token and now expecting the right-hand-side expression.
expr = ArrayLiteralRule(endNumberOnTernaryOpChars: true);
if (expr == null)
{
// ErrorRecovery: create an error expression to fill out the ast and keep parsing.
IScriptExtent extent = After(token);
// Use token.Text, not token.Kind.Text() b/c the kind might not match the actual operator used
// when a case insensitive operator is used.
ReportIncompleteInput(
extent,
nameof(ParserStrings.ExpectedValueExpression),
ParserStrings.ExpectedValueExpression,
token.Text);
expr = new ErrorExpressionAst(extent);
}
operandStack.Push(expr);
token = NextToken();
if (!token.Kind.HasTrait(TokenFlags.BinaryOperator))
{
// Remember the token that we stopped on, but only parameters, used for error recovery
paramToken = token as ParameterToken;
UngetToken(token);
break;
}
int newPrecedence = token.Kind.GetBinaryPrecedence();
while (newPrecedence <= precedence)
{
rhs = operandStack.Pop();
lhs = operandStack.Pop();
Token op = operatorStack.Pop();
operandStack.Push(new BinaryExpressionAst(ExtentOf(lhs, rhs), lhs, op.Kind, rhs, op.Extent));
if (operatorStack.Count == 0)
{
break;
}
precedence = operatorStack.Peek().Kind.GetBinaryPrecedence();
}
operatorStack.Push(token);
precedence = newPrecedence;
}
rhs = operandStack.Pop();
Diagnostics.Assert(operandStack.Count == operatorStack.Count, "Stacks out of sync");
while (operandStack.Count > 0)
{
lhs = operandStack.Pop();
token = operatorStack.Pop();
rhs = new BinaryExpressionAst(ExtentOf(lhs, rhs), lhs, token.Kind, rhs, token.Extent);
}
if (paramToken != null)
{
return ErrorRecoveryParameterInExpression(paramToken, rhs);
}
return rhs;
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
}
private ExpressionAst ErrorRecoveryParameterInExpression(ParameterToken paramToken, ExpressionAst expr)
{
// ErrorRecovery - when we have a parameter in an expression, eat it under the assumption
// that it's an incomplete operator. This simplifies analysis later, e.g. trying to autocomplete
// operators.
ReportError(paramToken.Extent,
nameof(ParserStrings.UnexpectedToken),
ParserStrings.UnexpectedToken,
paramToken.Text);
SkipToken();
return new ErrorExpressionAst(ExtentOf(expr, paramToken),
new Ast[] {
expr,
new CommandParameterAst(paramToken.Extent, paramToken.ParameterName, null, paramToken.Extent)});
}
/// <summary>Parse an array literal expression.</summary>
/// <param name="endNumberOnTernaryOpChars">
/// When it's known for sure that we are expecting an expression, allowing a generic token like '12?' or '12:' is
/// not useful. In those cases, we force to start a new token upon seeing '?' and ':' when scanning for a number
/// by setting this parameter to true, hoping to find a ternary expression.
/// </param>
private ExpressionAst ArrayLiteralRule(bool endNumberOnTernaryOpChars = false)
{
// G array-literal-expression:
// G unary-expression
// G unary-expression ',' new-lines:opt array-literal-expression
ExpressionAst lastExpr = UnaryExpressionRule(endNumberOnTernaryOpChars);
if (lastExpr == null)
{
return null;
}
ExpressionAst firstExpr = lastExpr;
Token commaToken = PeekToken();
if (commaToken.Kind != TokenKind.Comma || _disableCommaOperator)
{
return lastExpr;
}
var arrayValues = new List<ExpressionAst> { lastExpr };
while (commaToken.Kind == TokenKind.Comma)
{
SkipToken();
SkipNewlines();
// We have seen a comma token and now expecting an expression as an array element.
lastExpr = UnaryExpressionRule(endNumberOnTernaryOpChars: true);
if (lastExpr == null)
{
// ErrorRecovery: create an error expression for the ast and break.
ReportIncompleteInput(After(commaToken),
nameof(ParserStrings.MissingExpressionAfterToken),
ParserStrings.MissingExpressionAfterToken,
commaToken.Text);
lastExpr = new ErrorExpressionAst(commaToken.Extent);
arrayValues.Add(lastExpr);
break;
}
arrayValues.Add(lastExpr);
commaToken = PeekToken();
}
return new ArrayLiteralAst(ExtentOf(firstExpr, lastExpr), arrayValues);
}
/// <summary>Parse an unary expression.</summary>
/// <param name="endNumberOnTernaryOpChars">
/// When it's known for sure that we are expecting an expression, allowing a generic token like '12?' or '12:' is
/// not useful. In those cases, we force to start a new token upon seeing '?' and ':' when scanning for a number
/// by setting this parameter to true, hoping to find a ternary expression.
/// </param>
private ExpressionAst UnaryExpressionRule(bool endNumberOnTernaryOpChars = false)
{
// G unary-expression:
// G primary-expression
// G expression-with-unary-operator
// G
// G expression-with-unary-operator:
// G ',' new-lines:opt unary-expression
// G '-not' new-lines:opt unary-expression
// G '!' new-lines:opt unary-expression
// G '-bnot' new-lines:opt unary-expression
// G '+' new-lines:opt unary-expression
// G dash new-lines:opt unary-expression
// G pre-increment-expression
// G pre-decrement-expression
// G cast-expression
// G '-split' new-lines:opt unary-expression
// G '-join' new-lines:opt unary-expression
// G
// G pre-increment-expression:
// G '++' new-lines:opt unary-expression
// G
// G pre-decrement-expression:
// G dashdash new-lines:opt unary-expression
// G
// G cast-expression:
// G type-literal unary-expression
RuntimeHelpers.EnsureSufficientExecutionStack();
ExpressionAst expr = null;
Token token;
bool oldAllowSignedNumbers = _tokenizer.AllowSignedNumbers;
bool oldForceEndNumberOnTernaryOperators = _tokenizer.ForceEndNumberOnTernaryOpChars;
try
{
_tokenizer.AllowSignedNumbers = true;
_tokenizer.ForceEndNumberOnTernaryOpChars = endNumberOnTernaryOpChars;
if (_ungotToken != null)
{
// Possibly a signed number. Need to resync.
bool needResync = _ungotToken.Kind == TokenKind.Minus;
if (!needResync)
{
// A generic token possibly composed of numbers and ternary operator chars. Need to resync.
needResync = endNumberOnTernaryOpChars && _ungotToken.Kind == TokenKind.Generic;
}
if (needResync)
{
Resync(_ungotToken);
}
}
token = PeekToken();
}
finally
{
_tokenizer.AllowSignedNumbers = oldAllowSignedNumbers;
_tokenizer.ForceEndNumberOnTernaryOpChars = oldForceEndNumberOnTernaryOperators;
}
ExpressionAst child;
if (token.Kind.HasTrait(TokenFlags.UnaryOperator))
{
if (_disableCommaOperator && token.Kind == TokenKind.Comma)
{
return null;
}
SkipToken();
SkipNewlines();
// We have seen a unary operator token and now expecting an expression.
child = UnaryExpressionRule(endNumberOnTernaryOpChars: true);
if (child != null)
{
if (token.Kind == TokenKind.Comma)
{
expr = new ArrayLiteralAst(ExtentOf(token, child), new ExpressionAst[] { child });
}
else
{
expr = new UnaryExpressionAst(ExtentOf(token, child), token.Kind, child);
}
}
else
{
// ErrorRecovery: don't bother constructing a unary expression, but we know we must have
// some sort of expression, so return an error expression.
//
// Use token.Text, not token.Kind.Text() b/c the kind might not match the actual operator used
// when a case insensitive operator is used.
ReportIncompleteInput(
After(token),
nameof(ParserStrings.MissingExpressionAfterOperator),
ParserStrings.MissingExpressionAfterOperator,
token.Text);
return new ErrorExpressionAst(token.Extent);
}
}
else if (token.Kind == TokenKind.LBracket)
{
// Possibly a type literal or attribute.
var attributes = AttributeListRule(true);
if (attributes == null)
{
return null;
}
AttributeBaseAst lastAttribute = attributes.Last();
if (lastAttribute is AttributeAst)
{
SkipNewlines();
// We are now expecting a child expression.
child = UnaryExpressionRule(endNumberOnTernaryOpChars: true);
if (child == null)
{
// ErrorRecovery: We have a list of attributes, and we know it's not before a param statement,
// so we know we must have some sort of expression. Return an error expression then.
ReportIncompleteInput(
lastAttribute.Extent,
nameof(ParserStrings.UnexpectedAttribute),
ParserStrings.UnexpectedAttribute,
lastAttribute.TypeName.FullName);
return new ErrorExpressionAst(ExtentOf(token, lastAttribute), attributes);
}
expr = new AttributedExpressionAst(ExtentOf(lastAttribute, child), lastAttribute, child);
}
else
{
Diagnostics.Assert(
_ungotToken == null || ErrorList.Count > 0,
"Unexpected lookahead from AttributeListRule.");
// If we've looked ahead, don't go looking for a member access token, we've already issued an error,
// just assume we're not trying to access a member.
var memberAccessToken = _ungotToken != null ? null : NextMemberAccessToken(false);
if (memberAccessToken != null)
{
expr = CheckPostPrimaryExpressionOperators(
memberAccessToken,
new TypeExpressionAst(lastAttribute.Extent, lastAttribute.TypeName));
}
else
{
token = PeekToken();
if (token.Kind != TokenKind.NewLine && token.Kind != TokenKind.Comma)
{
// We are now expecting a child expression.
child = UnaryExpressionRule(endNumberOnTernaryOpChars: true);
if (child != null)
{
expr = new ConvertExpressionAst(
ExtentOf(lastAttribute, child),
(TypeConstraintAst)lastAttribute, child);
}
}
}
expr ??= new TypeExpressionAst(lastAttribute.Extent, lastAttribute.TypeName);
}
for (int i = attributes.Count - 2; i >= 0; --i)
{
var typeConstraint = attributes[i] as TypeConstraintAst;
expr = typeConstraint != null
? new ConvertExpressionAst(ExtentOf(typeConstraint, expr), typeConstraint, expr)
: new AttributedExpressionAst(ExtentOf(attributes[i], expr), attributes[i], expr);
}
}
else
{
expr = PrimaryExpressionRule(withMemberAccess: true);
}
if (expr != null)
{
token = PeekToken();
TokenKind operation = (token.Kind == TokenKind.PlusPlus)
? TokenKind.PostfixPlusPlus
: (token.Kind == TokenKind.MinusMinus)
? TokenKind.PostfixMinusMinus
: TokenKind.Unknown;
if (operation != TokenKind.Unknown)
{
SkipToken();
expr = new UnaryExpressionAst(ExtentOf(expr, token), operation, expr);
}
}
return expr;
}
private ExpressionAst PrimaryExpressionRule(bool withMemberAccess)
{
// G primary-expression:
// G value
// G member-access
// G element-access
// G invocation-expression
// G post-increment-expression
// G post-decrement-expression
// G value:
// G parenthesized-expression
// G sub-expression
// G array-expression
// G script-block-expression
// G hash-literal-expression
// G literal
// G type-literal
// G variable
ExpressionAst expr;
Token token = NextToken();
switch (token.Kind)
{
case TokenKind.SplattedVariable:
case TokenKind.Variable:
expr = CheckUsingVariable((VariableToken)token, withMemberAccess: withMemberAccess);
break;
case TokenKind.Number:
expr = new ConstantExpressionAst((NumberToken)token);
break;
case TokenKind.HereStringExpandable:
case TokenKind.StringExpandable:
expr = ExpandableStringRule((StringExpandableToken)token);
break;
case TokenKind.HereStringLiteral:
case TokenKind.StringLiteral:
expr = new StringConstantExpressionAst((StringToken)token);
break;
case TokenKind.LParen:
expr = ParenthesizedExpressionRule(token);
break;
case TokenKind.AtParen:
case TokenKind.DollarParen:
expr = SubExpressionRule(token);
break;
case TokenKind.AtCurly:
expr = HashExpressionRule(token, false /* parsingSchemaElement */ );
break;
case TokenKind.LCurly:
expr = ScriptBlockExpressionRule(token);
break;
default:
UngetToken(token);
return null;
}
if (!withMemberAccess)
{
return expr;
}
return CheckPostPrimaryExpressionOperators(NextMemberAccessToken(true), expr);
}
private ExpressionAst CheckUsingVariable(VariableToken variableToken, bool withMemberAccess)
{
var variablePath = variableToken.VariablePath;
if (variablePath.IsDriveQualified && variablePath.DriveName.Equals("using", StringComparison.OrdinalIgnoreCase) && variablePath.UnqualifiedPath.Length > 0)
{
var realVariablePath = new VariablePath(variablePath.UnqualifiedPath);
ExpressionAst childExpr = new VariableExpressionAst(variableToken.Extent, realVariablePath, (variableToken.Kind == TokenKind.SplattedVariable));
if (withMemberAccess)
{
childExpr = CheckPostPrimaryExpressionOperators(NextMemberAccessToken(true), childExpr);
}
return new UsingExpressionAst(childExpr.Extent, childExpr);
}
return new VariableExpressionAst(variableToken);
}
private ExpressionAst CheckPostPrimaryExpressionOperators(Token token, ExpressionAst expr)
{
while (token != null)
{
// To support fluent style programming, allow newlines after the member access operator.
SkipNewlines();
if (token.Kind == TokenKind.Dot || token.Kind == TokenKind.ColonColon || token.Kind == TokenKind.QuestionDot)
{
expr = MemberAccessRule(expr, token);
}
else if (token.Kind == TokenKind.LBracket || token.Kind == TokenKind.QuestionLBracket)
{
expr = ElementAccessRule(expr, token);
}
token = NextMemberAccessToken(true);
}
return expr;
}
private ExpressionAst HashExpressionRule(Token atCurlyToken, bool parsingSchemaElement)
{
// G hash-literal-expression:
// G '@{' new-lines:opt hash-literal-body:opt new-lines:opt '}'
// G hash-literal-body:
// G hash-entry
// G hash-literal-body statement-terminators hash-entry
// G statement-terminators:
// G statement-terminator
// G statement-terminators statement-terminator
SkipNewlines();
List<KeyValuePair> keyValuePairs = new List<KeyValuePair>();
while (true)
{
KeyValuePair pair = GetKeyValuePair(parsingSchemaElement);
if (pair == null)
{
break;
}
keyValuePairs.Add(pair);
Token token = PeekToken();
if (token.Kind != TokenKind.NewLine && token.Kind != TokenKind.Semi)
{
break;
}
SkipNewlinesAndSemicolons();
}
IScriptExtent endExtent;
Token rCurly = NextToken();
if (rCurly.Kind != TokenKind.RCurly)
{
UngetToken(rCurly);
string errorId;
string errorMsg;
if (parsingSchemaElement)
{
errorId = nameof(ParserStrings.IncompletePropertyAssignmentBlock);
errorMsg = ParserStrings.IncompletePropertyAssignmentBlock;
}
else
{
errorId = nameof(ParserStrings.MissingEndCurlyBrace);
errorMsg = ParserStrings.MissingEndCurlyBrace;
}
ReportIncompleteInput(After(atCurlyToken), rCurly.Extent, errorId, errorMsg);
endExtent = Before(rCurly);
}
else
{
endExtent = rCurly.Extent;
}
var hashAst = new HashtableAst(ExtentOf(atCurlyToken, endExtent), keyValuePairs);
hashAst.IsSchemaElement = parsingSchemaElement;
return hashAst;
}
private KeyValuePair GetKeyValuePair(bool parsingSchemaElement)
{
// G hash-entry:
// G key-expression '=' new-lines:opt statement
Token equals;
ExpressionAst key;
var oldTokenizerMode = _tokenizer.Mode;
try
{
SetTokenizerMode(TokenizerMode.Expression);
key = LabelOrKeyRule();
if (key == null)
{
return null;
}
equals = NextToken();
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
if (equals.Kind != TokenKind.Equals)
{
// ErrorRecovery: Pretend we saw the '=' and a statement.
UngetToken(equals);
IScriptExtent errorExtent = After(key);
string errorId;
string errorMsg;
if (parsingSchemaElement)
{
errorId = nameof(ParserStrings.MissingEqualsInPropertyAssignmentBlock);
errorMsg = ParserStrings.MissingEqualsInPropertyAssignmentBlock;
}
else
{
errorId = nameof(ParserStrings.MissingEqualsInHashLiteral);
errorMsg = ParserStrings.MissingEqualsInHashLiteral;
}
ReportError(errorExtent, errorId, errorMsg);
SyncOnError(false, TokenKind.RCurly, TokenKind.Semi, TokenKind.NewLine);
return new KeyValuePair(key, new ErrorStatementAst(errorExtent));
}
StatementAst statement;
try
{
SetTokenizerMode(TokenizerMode.Command);
SkipNewlines();
statement = StatementRule();
if (statement == null)
{
// ErrorRecovery: pretend we saw a statement and keep parsing.
IScriptExtent errorExtent = After(equals);
string errorId;
string errorMsg;
if (parsingSchemaElement)
{
errorId = nameof(ParserStrings.MissingEqualsInPropertyAssignmentBlock);
errorMsg = ParserStrings.MissingEqualsInPropertyAssignmentBlock;
}
else
{
errorId = nameof(ParserStrings.MissingStatementInHashLiteral);
errorMsg = ParserStrings.MissingStatementInHashLiteral;
}
ReportIncompleteInput(errorExtent, errorId, errorMsg);
statement = new ErrorStatementAst(errorExtent);
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
}
return new KeyValuePair(key, statement);
}
private ExpressionAst ScriptBlockExpressionRule(Token lCurly)
{
// G script-block-expression:
// G '{' new-lines:opt script-block new-lines:opt '}'
ScriptBlockAst scriptBlockAst;
bool oldDisableCommaOperator = _disableCommaOperator;
var oldTokenizerMode = _tokenizer.Mode;
try
{
_disableCommaOperator = false;
SetTokenizerMode(TokenizerMode.Command);
SkipNewlines();
scriptBlockAst = ScriptBlockRule(lCurly, isFilter: false);
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
SetTokenizerMode(oldTokenizerMode);
}
return new ScriptBlockExpressionAst(scriptBlockAst.Extent, scriptBlockAst);
}
private ExpressionAst SubExpressionRule(Token firstToken)
{
// G array-expression:
// G '@(' new-lines:opt statement-list:opt new-lines:opt ')'
// G sub-expression:
// G '$(' new-lines:opt statement-list:opt new-lines:opt ')'
IScriptExtent statementListExtent;
List<TrapStatementAst> traps = new List<TrapStatementAst>();
List<StatementAst> statements = new List<StatementAst>();
Token rParen;
bool oldDisableCommaOperator = _disableCommaOperator;
var oldTokenizerMode = _tokenizer.Mode;
try
{
_disableCommaOperator = false;
SetTokenizerMode(TokenizerMode.Command);
SkipNewlines();
statementListExtent = StatementListRule(statements, traps);
SkipNewlines();
rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: Assume only the closing paren is missing, continue as though it was present.
UngetToken(rParen);
ReportIncompleteInput(rParen.Extent,
nameof(ParserStrings.MissingEndParenthesisInSubexpression),
ParserStrings.MissingEndParenthesisInSubexpression);
}
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
SetTokenizerMode(oldTokenizerMode);
}
// End extent is rparen, end of the statement list (if no rparen), or the first token (if no statements).
IScriptExtent extent = ExtentOf(firstToken,
rParen.Kind == TokenKind.RParen ? rParen.Extent : statementListExtent ?? firstToken.Extent);
if (firstToken.Kind == TokenKind.DollarParen)
{
return new SubExpressionAst(extent,
new StatementBlockAst(statementListExtent ?? PositionUtilities.EmptyExtent,
statements, traps));
}
Diagnostics.Assert(firstToken.Kind == TokenKind.AtParen, "only support $() and @() here.");
return new ArrayExpressionAst(extent,
new StatementBlockAst(statementListExtent ?? PositionUtilities.EmptyExtent, statements, traps));
}
private ExpressionAst ParenthesizedExpressionRule(Token lParen)
{
// G parenthesized-expression:
// G '(' new-lines:opt pipeline-chain new-lines:opt ')'
Token rParen;
PipelineBaseAst pipelineAst;
var oldTokenizerMode = _tokenizer.Mode;
var oldDisableCommaOperator = _disableCommaOperator;
try
{
SetTokenizerMode(TokenizerMode.Command);
_disableCommaOperator = false;
SkipNewlines();
pipelineAst = PipelineChainRule();
if (pipelineAst == null)
{
IScriptExtent errorPosition = After(lParen);
ReportIncompleteInput(
errorPosition,
nameof(ParserStrings.ExpectedExpression),
ParserStrings.ExpectedExpression);
pipelineAst = new ErrorStatementAst(errorPosition);
}
SkipNewlines();
rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: Assume only the closing paren is missing, continue as though it was present.
UngetToken(rParen);
ReportIncompleteInput(After(pipelineAst),
nameof(ParserStrings.MissingEndParenthesisInExpression),
ParserStrings.MissingEndParenthesisInExpression);
rParen = null;
}
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
SetTokenizerMode(oldTokenizerMode);
}
return new ParenExpressionAst(ExtentOf(lParen, ExtentFromFirstOf(rParen, pipelineAst)), pipelineAst);
}
private List<ExpressionAst> ParseNestedExpressions(StringExpandableToken expandableStringToken)
{
List<ExpressionAst> nestedExpressions = new List<ExpressionAst>();
List<Token> newNestedTokens = _savingTokens ? new List<Token>() : null;
foreach (var token in expandableStringToken.NestedTokens)
{
Diagnostics.Assert(!token.HasError || ErrorList.Count > 0, "No nested tokens should have unreported errors.");
ExpressionAst exprAst;
var varToken = token as VariableToken;
if (varToken != null)
{
exprAst = CheckUsingVariable(varToken, false);
if (_savingTokens) { newNestedTokens.Add(varToken); }
}
// Enable if we decide we still need to support
// "${}" or "$var:"
// else if (token.Kind == TokenKind.Unknown)
// {
// // Diagnostics.Assert(token.Text.Equals("${}", StringComparison.OrdinalIgnoreCase),
// // "The unknown token is only used in an expandable string when it's an empty variable name.");
// // TODO: Need strict-mode check at runtime.
// // TODO: in V2, "${}" expanded to '$', but "$var:" expanded to the empty string
// exprAst = new StringConstantExpressionAst(token.Extent, "$", StringConstantType.BareWord);
// if (_savingTokens) { newNestedTokens.Add(token); }
// }
else
{
TokenizerState ts = null;
try
{
ts = _tokenizer.StartNestedScan((UnscannedSubExprToken)token);
exprAst = PrimaryExpressionRule(withMemberAccess: true);
if (_savingTokens) { newNestedTokens.AddRange(_tokenizer.TokenList); }
}
finally
{
// _ungotToken is probably <EOF>, but if there were errors, it could be something
// else. Either way, we don't want it.
_ungotToken = null;
_tokenizer.FinishNestedScan(ts);
}
}
nestedExpressions.Add(exprAst);
}
if (_savingTokens) { expandableStringToken.NestedTokens = new ReadOnlyCollection<Token>(newNestedTokens); }
return nestedExpressions;
}
private ExpressionAst ExpandableStringRule(StringExpandableToken strToken)
{
// G value:
// G literal
ExpressionAst expr;
// We need to scan the nested tokens even if there was some error. This is used by the tab completion: "pshome is $psh<tab>
if (strToken.NestedTokens != null)
{
List<ExpressionAst> nestedExpressions = ParseNestedExpressions(strToken);
expr = new ExpandableStringExpressionAst(strToken, strToken.Value, strToken.FormatString, nestedExpressions);
}
else
{
expr = new StringConstantExpressionAst(strToken);
}
return expr;
}
private ExpressionAst MemberNameRule()
{
// G member-name:
// G simple-name
// G string-literal
// G string-literal-with-subexpression
// G expression-with-unary-operator
// G value
ExpressionAst simpleName = SimpleNameRule();
if (simpleName != null)
{
return simpleName;
}
Token token = PeekToken();
if (token.Kind.HasTrait(TokenFlags.UnaryOperator) || token.Kind == TokenKind.LBracket)
{
return UnaryExpressionRule();
}
return PrimaryExpressionRule(withMemberAccess: false);
}
private ExpressionAst MemberAccessRule(ExpressionAst targetExpr, Token operatorToken)
{
// G member-access: No whitespace is allowed between terms in these productions.
// G primary-expression '.' member-name
// G primary-expression '::' member-name
// On entry, we've verified that operatorToken is not preceded by whitespace.
CommandElementAst member = MemberNameRule();
if (member == null)
{
// ErrorRecovery: pretend we saw a property name, don't bother looking for an invocation,
// and keep parsing.
ReportIncompleteInput(After(operatorToken),
nameof(ParserStrings.MissingPropertyName),
ParserStrings.MissingPropertyName);
member = GetSingleCommandArgument(CommandArgumentContext.CommandArgument) ??
new ErrorExpressionAst(ExtentOf(targetExpr, operatorToken));
}
else if (_ungotToken == null)
{
// Member name may be an incomplete token like `$a.$(Command-Name`, in which case, '_ungotToken != null'.
// We do not look for generic args or invocation token if the member name token is recognisably incomplete.
int resyncIndex = _tokenizer.GetRestorePoint();
List<ITypeName> genericTypeArguments = GenericMethodArgumentsRule(resyncIndex, out Token rBracket);
Token lParen = NextInvokeMemberToken();
if (lParen != null)
{
// When we reach here, we either had a legit section of generic arguments (in which case, `rBracket`
// won't be null), or we saw `lParen` directly following the member token (in which case, `rBracket`
// will be null).
int endColumnNumber = rBracket is null ? member.Extent.EndColumnNumber : rBracket.Extent.EndColumnNumber;
Diagnostics.Assert(lParen.Kind == TokenKind.LParen || lParen.Kind == TokenKind.LCurly, "token kind incorrect");
Diagnostics.Assert(
endColumnNumber == lParen.Extent.StartColumnNumber,
"member and paren must be adjacent when the method is not generic");
return MemberInvokeRule(targetExpr, lParen, operatorToken, member, genericTypeArguments);
}
else if (rBracket != null)
{
// We had a legit section of generic arguments but no 'lParen' following that, so this is not a method
// invocation, but an invalid indexing operation. Resync the tokenizer back to before the generic arg
// parsing and then continue.
Resync(resyncIndex);
}
}
return new MemberExpressionAst(
ExtentOf(targetExpr, member),
targetExpr,
member,
@static: operatorToken.Kind == TokenKind.ColonColon,
nullConditional: operatorToken.Kind == TokenKind.QuestionDot);
}
private List<ITypeName> GenericMethodArgumentsRule(int resyncIndex, out Token rBracketToken)
{
List<ITypeName> genericTypes = null;
Token lBracket = NextToken();
rBracketToken = null;
if (lBracket.Kind != TokenKind.LBracket)
{
// We cannot avoid this Resync(); if we use PeekToken() to try to avoid a Resync(), the method called
// after this [`NextInvokeMemberToken()` or `NextMemberAccessToken()`] will note that an _ungotToken
// is present and assume an error state. That will cause any property accesses or non-generic method
// calls to throw a parse error.
Resync(resyncIndex);
return null;
}
// This is either a InvokeMember expression with generic type arguments, or some sort of collection index
// on a property.
TokenizerMode oldTokenizerMode = _tokenizer.Mode;
try
{
// Switch to typename mode to avoid aggressive argument tokenization.
SetTokenizerMode(TokenizerMode.TypeName);
SkipNewlines();
Token firstToken = NextToken();
// For method generic arguments, we only support the syntax `$var.Method[TypeName1 <, TypeName2 ...>]`,
// not the syntax `$var.Method[[TypeName1] <, [TypeName2] ...>]`.
// The latter syntax has been supported for type expression since the beginning, but it's ambiguous in
// this scenario because we could be looking at an indexing operation on a property like:
// `$var.Property[<expression>]`
// and the `<expression>` could start with a type expression like `[TypeName]::Method()`, or even just
// a single type expression acting as a key to a hashtable property. Such cases will cause ambiguities.
//
// It could be possible to write code that sorts out the ambiguity and continue to support the latter
// syntax for method generic arguments, and thus to allow assembly-qualified type names. But we choose
// not to do so because:
// 1. that will definitely increase the complexity of the parsing code and also make it fragile;
// 2. the latter syntax hurts readability a lot due to the number of opening/closing brackets.
// The downside is that the assembly-qualified type names won't be supported for method generic args,
// but that's likely not a problem in practice, and we can revisit if it turns out otherwise.
if (firstToken.Kind == TokenKind.Identifier)
{
resyncIndex = -1;
genericTypes = GenericTypeArgumentsRule(firstToken, out rBracketToken);
if (rBracketToken.Kind != TokenKind.RBracket)
{
UngetToken(rBracketToken);
ReportIncompleteInput(
Before(rBracketToken),
nameof(ParserStrings.EndSquareBracketExpectedAtEndOfType),
ParserStrings.EndSquareBracketExpectedAtEndOfType);
rBracketToken = null;
}
}
}
finally
{
SetTokenizerMode(oldTokenizerMode);
if (resyncIndex > 0)
{
Resync(resyncIndex);
}
}
return genericTypes;
}
private ExpressionAst MemberInvokeRule(
ExpressionAst targetExpr,
Token lBracket,
Token operatorToken,
CommandElementAst member,
IList<ITypeName> genericTypes)
{
// G invocation-expression: target-expression passed as a parameter. lBracket can be '(' or '{'.
// G target-expression member-name invoke-param-list
// G invoke-param-list:
// G '(' invoke-param-paren-list
// G script-block
IScriptExtent lastExtent = null;
List<ExpressionAst> arguments;
if (lBracket.Kind == TokenKind.LParen)
{
arguments = this.InvokeParamParenListRule(lBracket, out lastExtent);
}
else
{
arguments = new List<ExpressionAst>();
// handle the construct $x.methodName{2+2} as through it had been written $x.methodName({2+2})
SkipNewlines();
ExpressionAst argument = ScriptBlockExpressionRule(lBracket);
arguments.Add(argument);
lastExtent = argument.Extent;
}
return new InvokeMemberExpressionAst(
ExtentOf(targetExpr, lastExtent),
targetExpr,
member,
arguments,
operatorToken.Kind == TokenKind.ColonColon,
operatorToken.Kind == TokenKind.QuestionDot,
genericTypes);
}
private List<ExpressionAst> InvokeParamParenListRule(Token lParen, out IScriptExtent lastExtent)
{
// G argument-list: '(' is passed in lParen
// G argument-expression-list:opt new-lines:opt ')'
// G argument-expression-list:
// G argument-expression
// G argument-expression new-lines:opt ',' argument-expression-list
// G argument-expression:
// G See grammar for expression - the only difference is that an
// G array-literal-expression is not allowed - the comma is used
// G to separate argument-expressions.
List<ExpressionAst> arguments = new List<ExpressionAst>();
Token comma = null;
Token rParen = null;
bool oldDisableCommaOperator = _disableCommaOperator;
bool reportedError = false;
try
{
_disableCommaOperator = true;
while (true)
{
SkipNewlines();
ExpressionAst argument = ExpressionRule();
if (argument == null)
{
if (comma != null)
{
// ErrorRecovery: sync at closing paren or newline.
ReportIncompleteInput(After(comma),
nameof(ParserStrings.MissingExpressionAfterToken),
ParserStrings.MissingExpressionAfterToken,
TokenKind.Comma.Text());
reportedError = true;
}
break;
}
arguments.Add(argument);
SkipNewlines();
comma = NextToken();
if (comma.Kind != TokenKind.Comma)
{
UngetToken(comma);
comma = null;
break;
}
}
SkipNewlines();
rParen = NextToken();
if (rParen.Kind != TokenKind.RParen)
{
// ErrorRecovery: pretend we saw a closing paren or curly and keep parsing.
UngetToken(rParen);
if (!reportedError)
{
ReportIncompleteInput(arguments.Count > 0 ? After(arguments.Last()) : After(lParen),
nameof(ParserStrings.MissingEndParenthesisInMethodCall),
ParserStrings.MissingEndParenthesisInMethodCall);
}
rParen = null;
}
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
}
lastExtent = ExtentFromFirstOf(rParen, comma, arguments.LastOrDefault(), lParen);
return arguments;
}
private ExpressionAst ElementAccessRule(ExpressionAst primaryExpression, Token lBracket)
{
// G element-access:
// G primary-expression '[' new-lines:opt expression new-lines:opt ']'
SkipNewlines();
bool oldDisableCommaOperator = _disableCommaOperator;
_disableCommaOperator = false;
ExpressionAst indexExpr = null;
try
{
indexExpr = ExpressionRule();
}
finally
{
_disableCommaOperator = oldDisableCommaOperator;
}
if (indexExpr == null)
{
// ErrorRecovery: hope we see a closing bracket. If we don't, we'll pretend we saw
// the closing bracket, but build an expression that can't compile.
var errorExtent = After(lBracket);
ReportIncompleteInput(
errorExtent,
nameof(ParserStrings.MissingArrayIndexExpression),
ParserStrings.MissingArrayIndexExpression);
indexExpr = new ErrorExpressionAst(lBracket.Extent);
}
SkipNewlines();
Token rBracket = NextToken();
if (rBracket.Kind != TokenKind.RBracket)
{
// ErrorRecovery: just pretend we had a closing bracket and continue parsing.
UngetToken(rBracket);
// Skip reporting the error if we've already reported a missing index.
if (indexExpr is not ErrorExpressionAst)
{
ReportIncompleteInput(After(indexExpr),
nameof(ParserStrings.MissingEndSquareBracket),
ParserStrings.MissingEndSquareBracket);
}
rBracket = null;
}
return new IndexExpressionAst(ExtentOf(primaryExpression, ExtentFromFirstOf(rBracket, indexExpr)), primaryExpression, indexExpr, lBracket.Kind == TokenKind.QuestionLBracket);
}
#endregion Expressions
#region Error Reporting
private void SaveError(ParseError error)
{
if (ErrorList.Count > 0)
{
foreach (ParseError err in ErrorList)
{
if (err.ErrorId.Equals(error.ErrorId, StringComparison.Ordinal)
&& err.Extent.EndColumnNumber == error.Extent.EndColumnNumber
&& err.Extent.EndLineNumber == error.Extent.EndLineNumber
&& err.Extent.StartColumnNumber == error.Extent.StartColumnNumber
&& err.Extent.StartLineNumber == error.Extent.StartLineNumber)
{
return;
}
}
}
ErrorList.Add(error);
}
private void SaveError(IScriptExtent extent, string errorId, string errorMsg, bool incompleteInput, params object[] args)
{
AssertErrorIdCorrespondsToMsgString(errorId, errorMsg);
if (args != null && args.Length > 0)
{
errorMsg = string.Format(CultureInfo.CurrentCulture, errorMsg, args);
}
ParseError errorToSave = new ParseError(extent, errorId, errorMsg, incompleteInput);
SaveError(errorToSave);
}
/// <summary>
/// Debug assertion to ensure that all errors saved by the parser come
/// from resource (.resx) files.
/// </summary>
/// <param name="errorId">The error ID string (.resx key).</param>
/// <param name="errorMsg">The error message, which may be a template string (.resx value).</param>
[System.Diagnostics.Conditional("DEBUG")]
[System.Diagnostics.Conditional("ASSERTIONS_TRACE")]
private static void AssertErrorIdCorrespondsToMsgString(string errorId, string errorMsg)
{
// These types are the ones known to contain
// strings used by the parser as errors
Type[] resxTypes = new[]
{
typeof(ParserStrings),
typeof(DiscoveryExceptions),
typeof(ExtendedTypeSystem),
typeof(MshSnapInCmdletResources),
typeof(ParameterBinderStrings)
};
// Go through each resource type and see if the errorId key is in it, and whether the value corresponds to the errorMsg
bool msgCorrespondsToString = false;
foreach (Type resxType in resxTypes)
{
string resxErrorBody = resxType.GetProperty(errorId, BindingFlags.Static | BindingFlags.NonPublic)?.GetValue(null) as string;
if (string.Equals(errorMsg, resxErrorBody, StringComparison.Ordinal))
{
msgCorrespondsToString = true;
break;
}
}
Diagnostics.Assert(msgCorrespondsToString, $"Parser error ID \"{errorId}\" must correspond to the error message \"{errorMsg}\"");
}
private static object[] arrayOfOneArg
{
get { return t_arrayOfOneArg ??= new object[1]; }
}
[ThreadStatic]
private static object[] t_arrayOfOneArg;
private static object[] arrayOfTwoArgs
{
get { return t_arrayOfTwoArgs ??= new object[2]; }
}
[ThreadStatic]
private static object[] t_arrayOfTwoArgs;
internal bool ReportIncompleteInput(IScriptExtent extent, string errorId, string errorMsg)
{
// If the error position isn't at the end of the input, then we don't want to mark the error
// as incomplete input.
bool incompleteInput = _tokenizer.IsAtEndOfScript(extent, checkCommentsAndWhitespace: true);
SaveError(extent, errorId, errorMsg, incompleteInput, null);
return incompleteInput;
}
internal bool ReportIncompleteInput(IScriptExtent extent, string errorId, string errorMsg, object arg)
{
// If the error position isn't at the end of the input, then we don't want to mark the error
// as incomplete input.
bool incompleteInput = _tokenizer.IsAtEndOfScript(extent, checkCommentsAndWhitespace: true);
arrayOfOneArg[0] = arg;
SaveError(extent, errorId, errorMsg, incompleteInput, arrayOfOneArg);
return incompleteInput;
}
internal bool ReportIncompleteInput(IScriptExtent errorPosition,
IScriptExtent errorDetectedPosition,
string errorId,
string errorMsg,
params object[] args)
{
// If the error position isn't at the end of the input, then we don't want to mark the error
// as incomplete input.
bool incompleteInput = _tokenizer.IsAtEndOfScript(errorDetectedPosition, checkCommentsAndWhitespace: true);
SaveError(errorPosition, errorId, errorMsg, incompleteInput, args);
return incompleteInput;
}
internal void ReportError(IScriptExtent extent, string errorId, string errorMsg)
{
SaveError(extent, errorId, errorMsg, false, null);
}
internal void ReportError(IScriptExtent extent, string errorId, string errorMsg, object arg)
{
arrayOfOneArg[0] = arg;
SaveError(extent, errorId, errorMsg, false, arrayOfOneArg);
}
internal void ReportError(IScriptExtent extent, string errorId, string errorMsg, object arg1, object arg2)
{
arrayOfTwoArgs[0] = arg1;
arrayOfTwoArgs[1] = arg2;
SaveError(extent, errorId, errorMsg, false, arrayOfTwoArgs);
}
internal void ReportError(IScriptExtent extent, string errorId, string errorMsg, params object[] args)
{
SaveError(extent, errorId, errorMsg, false, args);
}
internal void ReportError(ParseError error)
{
SaveError(error);
}
private static void ReportErrorsAsWarnings(Collection<Exception> errors)
{
var executionContext = Runspaces.Runspace.DefaultRunspace.ExecutionContext;
if (executionContext != null && executionContext.InternalHost != null && executionContext.InternalHost.UI != null)
{
foreach (var error in errors)
{
if (error != null)
{
executionContext.InternalHost.UI.WriteWarningLine(error.ToString());
}
}
}
}
#endregion Error Reporting
}
#region Error related classes
/// <summary>
/// </summary>
public class ParseError
{
/// <summary>
/// Creates a new parse error.
/// </summary>
/// <param name="extent">The IScriptExtent that represents the location of the error.</param>
/// <param name="errorId">The error ID to associate with the error.</param>
/// <param name="message">The message of the error.</param>
public ParseError(IScriptExtent extent, string errorId, string message)
: this(extent, errorId, message, false)
{
}
internal ParseError(IScriptExtent extent, string errorId, string message, bool incompleteInput)
{
Diagnostics.Assert(extent != null, "can't have a null position for a parse error");
Diagnostics.Assert(!string.IsNullOrEmpty(message), "can't have a null error message");
Diagnostics.Assert(!string.IsNullOrEmpty(errorId), "can't have a null error id");
Extent = extent;
ErrorId = errorId;
Message = message;
IncompleteInput = incompleteInput;
}
/// <summary>
/// </summary>
/// <returns></returns>
public override string ToString()
{
return PositionUtilities.VerboseMessage(Extent) + Environment.NewLine + Message;
}
/// <summary>
/// </summary>
public IScriptExtent Extent { get; }
/// <summary>
/// </summary>
public string ErrorId { get; }
/// <summary>
/// </summary>
public string Message { get; }
/// <summary>
/// </summary>
public bool IncompleteInput { get; }
}
#endregion Error related classes
// Guid is {eba789d9-533b-58d4-cd1f-2e6520e3a9c2}
[EventSource(Name = "Microsoft-PowerShell-Parser")]
internal class ParserEventSource : EventSource
{
internal static readonly ParserEventSource Log = new ParserEventSource();
internal const int MaxScriptLengthToLog = 50;
public void ParseStart(string FileName, int Length) { WriteEvent(1, FileName, Length); }
public void ParseStop() { WriteEvent(2); }
public void ResolveSymbolsStart() { WriteEvent(3); }
public void ResolveSymbolsStop() { WriteEvent(4); }
public void SemanticChecksStart() { WriteEvent(5); }
public void SemanticChecksStop() { WriteEvent(6); }
public void CheckSecurityStart(string FileName) { WriteEvent(7, FileName); }
public void CheckSecurityStop(string FileName) { WriteEvent(8, FileName); }
public void CompileStart(string FileName, int Length, bool Optimized) { WriteEvent(9, FileName, Length, Optimized); }
public void CompileStop() { WriteEvent(10); }
internal static string GetFileOrScript(string fileName, string input)
{
return fileName ?? input.AsSpan(0, Math.Min(256, input.Length)).Trim().ToString();
}
}
}