Upload folder using huggingface_hub

README.md

@@ -0,0 +1,117 @@
+---
+license: apache-2.0
+language:
+ - en
+multilinguality:
+ - monolingual
+tags:
+ - text
+ - qa
+ - code
+ - lua
+ - roblox
+size_categories:
+ - 1K<n<10K
+task_categories:
+ - question-answering
+task_ids:
+ - closed-domain-qa
+pretty_name: "scriptinghelpers archive"
+configs:
+  - config_name: raw_data
+    data_files:
+      - split: train
+        path: "questions_raw.snappy.parquet"
+  - config_name: parsed_data
+    data_files:
+      - split: train
+        path: "data.snappy.parquet"
+dataset_info:
+  - config_name: raw_data
+    features:
+      - name: path
+        dtype: string
+      - name: content
+        dtype: string
+    splits:
+      - name: train
+        num_examples: 5733
+  - config_name: parsed_data
+    features:
+      - name: question_score
+        dtype: string
+      - name: title  
+        dtype: string
+      - name: question_userid
+        dtype: string
+      - name: question_user
+        dtype: string
+      - name: question_user_score
+        dtype: string
+      - name: question_datetime
+        dtype: string
+      - name: question
+        dtype: string
+      - name: question_id
+        dtype: string
+      - name: answer_count
+        dtype: string
+      - name: answer_id
+        dtype: string
+      - name: answer_score
+        dtype: string
+      - name: answer_userid
+        dtype: string
+      - name: answer_user
+        dtype: string
+      - name: answer_user_score
+        dtype: string
+      - name: answer_datetime
+        dtype: string
+      - name: answer
+        dtype: string
+    splits:
+      - name: train
+        num_examples: 5129
+---
+
+This is a small dataset containing question/answer articles on (now defunct) scriptinghelpers.org, which was a StackExchange-like website for Roblox developers. It was scraped from archive.org about a year ago and parsed with regex.
+
+## Trivia
+
+ - There are 2097 archived articles with accepted answers out of 5129.
+ - The earliest question archived was posted on `Sat Feb  1 23:57:31 2014` with question_id `1` by TheGuyWithAShortName entitled `HttpService - PostAsync `. It had an accepted answer posted on `2014-02-17 03:11:33` by MarkOtaris.
+ - The last question archived was posted on `Sun Apr  2 01:40:15 2023` with question_id `133258` by RedWirePlatinum entitled `Help with making a bouncy character script *without* creating instances?`.
+ - The last question archived with an accepted answer was posted on `Thu Mar 30 18:53:16 2023` with question_id `133254` by krLuCiEzkr entitled `LocalScript that checks if a player has this badge doesn't work? [CLOSED]`. Its answer was posted on `2023-03-30 19:29:01` by MaleficentMaestro.
+
+## Other Notes
+
+The `question` and `answer` columns have an endogenous HTML schema that has not been transformed into Markdown. For instance, in question id `129668`:
+
+```html
+<!-- question -->
+<p>I was trying to remove the HopperBin in user's inventory using the following script:</p>
+<pre class="brush: lua">game.Players.PlayerAdded:Connect(function(player)
+    for i,v in pairs(player.Backpack:GetChildren())
+        if v:IsA("HopperBin") then
+            v:Destroy()
+        end
+    end
+end)
+</pre>
+<p>Instead of removing, it does nothing.
+Any help would be great!</p>
+
+<!-- answer -->
+<p>1st of all, you should try it when the <strong>Character</strong> is added. Tools and such do not go straight to the backpack, it <strong>waits until the character is added and then adds tools to the backpack</strong>.</p>
+<p>2nd of all, I wouldn't use HopperBins to begin with, as they're <strong>deprecated</strong> (use tools).</p>
+
+```
+
+Parsing was done only with the single-pass regular expression and minimal post-processing to remove recurring anomalies.
+Comments and unaccepted answers were not included in the parsed data, but can be found in the raw data.
+
+## Requirements
+
+- `brute.py` requires an installation of [ripgrep](https://github.com/burntsushi/ripgrep#installation) on your machine.
+- `data_to_parquet.py` requires pyarrow.

brute.py

@@ -0,0 +1,71 @@
+import re
+from re import Match
+import subprocess
+from tqdm import tqdm
+from questions_tar_extract import extract_articles
+import shutil
+import os
+from pathlib import Path
+import stat
+
+blockregex  = re.compile(r'<<<(.*?)<<<', flags=re.MULTILINE | re.DOTALL)
+tr = str.maketrans({
+                   "\"": "\\\"",
+                   "\\": "\\\\",
+                   "\n": "\\n"
+                   })
+
+def render_block(mo: Match) -> str:
+    block = mo.group(1)
+    block = block.translate(tr)
+    return f'"{block}"'
+
+def prepare_with_ripgrep() -> list[str]:
+    articles = extract_articles()
+    
+    pbar = tqdm(total=5129, desc="Extracting and formatting scriptinghelpers posts")
+
+    rg_exec: str | None = None
+    if 'RIPGREP_PATH' in os.environ:
+        rg_exec_p = Path(os.path.normpath(os.environ['RIPGREP_PATH'].strip()))
+        if rg_exec_p.is_dir():
+            rg_exec = shutil.which("rg", path=rg_exec_p)
+        else:
+            rg_exec_mode = rg_exec_p.stat().st_mode
+            if bool(rg_exec_mode & (stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH)):
+                rg_exec = str(rg_exec_p)
+    else:
+        rg_exec = shutil.which("rg")
+
+    if rg_exec is None:
+        raise RuntimeError(f"ripgrep executable `{rg_exec}` is not installed or not visible in this environment")
+    
+    process = subprocess.Popen([rg_exec, "-U", "-uuu", "-o", "--multiline-dotall", "-N", "--color=never", "-I", "-f", "qa_regex4.txt", str(articles),
+                   "-r", """{"qscore": "$1", "title": "$2", "quserid": "$3", "quser": "$4", "quserscore": "$5", "qdatetime": "$6", "question": <<<$7<<<, "qid": "$8", "answercount": "$9", "aid": "$10", "ascore": "$11", "auserid": "$12", "auser": "$13", "auserscore": "$14", "adatetime": "$15", "answer": <<<$16<<<}"""], stdout=subprocess.PIPE, text=True, bufsize=1)
+    chunk_array = []
+    latest_chunk_part = []
+    while True:
+        line = process.stdout.readline()
+        if not line:
+            break
+        line = line.rstrip()
+        if line[-4:] == '<<<}':
+            latest_chunk = line if len(latest_chunk_part) == 0 else "\n".join(latest_chunk_part) + f"\n{line}"
+            latest_chunk = blockregex.sub(render_block, latest_chunk)
+            chunk_array.append(latest_chunk)
+            latest_chunk_part.clear()
+            pbar.update(1)
+        else:
+            latest_chunk_part.append(line)
+    pbar.total = len(chunk_array)
+    pbar.refresh()
+    pbar.close()
+
+    return chunk_array
+
+if __name__=='__main__':
+    chunk_array = "\n".join(prepare_with_ripgrep())
+    with open('data.jsonl', 'w') as f:
+        f.write(chunk_array)
+    # with open('extract5.txt', 'w') as f:
+    #     f.write(f'[{",\n".join(chunk_array)}]')

data.snappy.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11f7eda26f0d1001b03f6da3cbd9e0fbaf44f2b678931fbd63e31d619c97e5ba
+size 3491342

data_to_parquet.py

@@ -0,0 +1,106 @@
+from pathlib import Path
+import pyarrow as pa
+import pyarrow.parquet as pq
+import sys
+import json
+import time
+
+from questions_tar_extract import extract_articles
+from brute import prepare_with_ripgrep
+
+file_structure_schema = pa.schema([
+                                  ('path', pa.string()),
+                                  ('content', pa.large_string())
+                                  ])
+
+extracted_schema = pa.schema([
+                             ('question_score', pa.string()),
+                             ('title', pa.string()),
+                             ('question_userid', pa.string()),
+                             ('question_user', pa.string()),
+                             ('question_user_score', pa.string()),
+                             ('question_datetime', pa.string()),
+                             ('question', pa.string()),
+                             ('question_id', pa.string()),
+                             ('answer_count', pa.string()),
+                             ('answer_id', pa.string()),
+                             ('answer_score', pa.string()),
+                             ('answer_userid', pa.string()),
+                             ('answer_user', pa.string()),
+                             ('answer_user_score', pa.string()),
+                             ('answer_datetime', pa.string()),
+                             ('answer', pa.string())
+                             ])
+
+extracted_src_key_order = ["qscore","title","quserid","quser","quserscore","qdatetime","question","qid","answercount","aid","ascore","auserid","auser","auserscore","adatetime","answer"]
+
+script_dir = Path(__file__).resolve().parent
+
+if __name__ == '__main__':
+    articles = extract_articles()
+
+    scraped_files_root = articles.name + "_raw"
+    
+    article_pq_inputs = [[],[]]
+    for a in articles.iterdir():
+        a = next(next(a.iterdir()).iterdir())
+        with open(a, "r") as f:
+            content = f.read()
+        article_pq_inputs[0].append(str(a))
+        article_pq_inputs[1].append(content)
+
+    article_table = pa.Table.from_arrays(article_pq_inputs, schema=file_structure_schema)
+    
+    pq.write_table(article_table, f"{scraped_files_root}.snappy.parquet", compression='Snappy')
+
+    article_pq_file = Path(f"{scraped_files_root}.snappy.parquet")
+
+    print(f"wrote to cwd: {scraped_files_root}.snappy.parquet ({article_pq_file.stat().st_size / (1024 * 1024):.2f} MiB)")
+
+    extract_file_path = script_dir / "data.jsonl"
+    extract_pq_inputs: list[dict[str, str]] | list[list[str]] | None = None
+    if not extract_file_path.is_file():
+        try:
+            rem_attempts = 2
+            while rem_attempts > 0:
+                user_input = input(f"extracted data not found at {extract_file_path}. run extraction (need ripgrep)? [y/N] ").strip().lower()
+
+                if not user_input or user_input in ['n', 'no']:
+                    print("warning: parquet conversion only partially completed because extracted data could not be prepared.")
+                    sys.exit(0)
+                elif user_input in ['y', 'yes']:
+                    chunk_array = prepare_with_ripgrep()
+                    with open(script_dir / "data.jsonl", 'w', encoding='utf-8') as f:
+                        f.write("\n".join(chunk_array))
+                    extract_pq_inputs = list(map(json.loads, chunk_array))
+                    break
+                else:
+                    print(f"warning: unknown response '{user_input if len(user_input) < 4 else f"{user_input[:3]}..."}'. remaining attempts: {rem_attempts}")
+                    time.sleep(0.5)
+                    rem_attempts -= 1
+            if rem_attempts == 0:
+                print("fatal: failed to move on after a few attempts")
+                sys.exit(1)
+        except KeyboardInterrupt:
+            print("fatal: received KeyboardInterrupt")
+            sys.exit(1)
+    else:
+        with open(extract_file_path, 'r', encoding='utf-8') as f:
+            extract_pq_inputs = list(map(json.loads, f))
+
+    assert extract_pq_inputs is not None
+
+    _extract_pq_inputs = [[] for k in extracted_src_key_order]
+    for j, k_j in enumerate(extracted_src_key_order):
+        for i in range(len(extract_pq_inputs)):
+            _extract_pq_inputs[j].append(extract_pq_inputs[i][k_j])
+    extract_pq_inputs = _extract_pq_inputs
+            
+    extract_table = pa.Table.from_arrays(extract_pq_inputs, schema=extracted_schema)
+
+    pq.write_table(extract_table, "data.snappy.parquet", compression='Snappy')
+    
+    extract_pq_file = Path("data.snappy.parquet")
+
+    print(f"wrote to cwd: data.snappy.parquet ({extract_pq_file.stat().st_size / (1024 * 1024):.2f} MiB)")
+    

format.md

@@ -0,0 +1,26 @@
+# Format of Scripting Helpers Q&A posts
+
+## QA Regex
+
+### Prototype
+
+```regex
+points-number">(?P<qscore>\d+)[^h]+<h1 class="question-title">\r?\n\s*(?P<title>[^<]+)(?:\s*<span class="muted">\[closed\]</span>)?</h1>\s*<div class="question-poster sans-serif">\r?\n\s*<span class="muted">Asked by</span>\r?\n\s*<div class="question-poster-username">\r?\n\s*<img src="resources/avatar/(?P<quserid>\d+)" alt(?:="[^"]*")? class="loggedin-avatar" /><a href="/user/(?P<quser>\w+)" class="loggedin-name">[^<]+</a> <span class="muted small">(?P<quserscore>\d+)</span>[^f]+float-right" title="(?P<qdatetime>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})[^q]+question-markdown">(?P<question>(?:<p>(?:[^<]|<code>[^<]+</code>)*</p>|<pre class=[\x27"]brush: lua[\x27"]>[^<]*</pre>)*)[^I]+Id = \x27q(?P<qid>\d+)(?:.*?h3 class="wide-line-title">(?P<answercount>\d+) answers?</h3>\r?\n\s*<div class="question" id="(?P<aid>\d+)">[^b]+ber">(?P<ascore>\d+)[^A]+Accepted answer[^A]Answered by</span>\r?\n\s*<div class="question-poster-username">\r?\n\s*<img src="/resources/avatar/(?P<auserid>\d+)" alt(?:="[^"]*")? class="loggedin-avatar" /><a href="/user/(?P<auser>\w+)" class="loggedin-name">\w+</a> <span class="muted small">(?P<auserscore>\d+)[^f]float right" title="(?P<adatetime>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})[^q]question-markdown">(?P<answer>(?:<p>(?:[^<]|<code>[^<]+</code>)*</p>|<pre class=[\x27"]brush: lua[\x27"]>[^<]*</pre>)*))?
+```
+
+Unusable. Use [./qa_regex4.txt/](qa_regex4.txt)
+
+### Command
+
+Requires [ripgrep](https://github.com/burntsushi/ripgrep).
+
+```
+rg -U -uuu -o --multiline-dotall -N --color never -I -f qa_regex4.txt questions \
+  -r '{"qscore": "$1", "title": "$2", "quserid": "$3", "quser": "$4", "quserscore": "$5", "qdatetime": "$6", "question": <<<$7<<<, "qid": "$8", "answercount": "$9", "aid": "$10", "ascore": "$11", "auserid": "$12", "auser": "$13", "auserscore": "$14", "adatetime": "$15", "answer": >>>$16>>>}' \
+```
+
+Must replace `<<<` blocks with long, programmatically escaped content strings. Some post content includes single backslashes. All post content has many newlines.
+
+### Entry Specification
+
+There is one entry for every asked question, so there may be no accepted answer associated with it. In that case, all of the answer's properties (from and including `"answercount"`) will be empty strings.

qa_regex4.txt

@@ -0,0 +1 @@
+points-number">(?P<qscore>\d+)[^h]+h1 class="question-title">\r?\n\s*(?P<title>[^<]+)(?:\s*<span class="muted">\[closed\]</span>)?</h1>\s*<div class="question-poster sans-serif">\r?\n\s*<span class="muted">Asked by</span>\r?\n\s*<div class="question-poster-username">\r?\n\s*<img src="/resources/avatar/(?P<quserid>\d+)" alt(?:="[^"]*")? class="loggedin-avatar" /><a href="/user/(?P<quser>\w+)" class="loggedin-name">[^<]+</a> <span class="muted small">(?P<quserscore>\d+)</span>[^f]+float-right" title="(?P<qdatetime>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})[^q]+question-markdown">(?:<div id="markdown-[a-z0-9]+">)?(?P<question>(?:(?:<(?:p|h[1-6]|blockquote)>(?:[^<]|<(?:code|strong|em|blockquote|p|a(?: target="[^"]*")? href="[^"]+"(?: rev="footnote")?(?: class="[^"]+")?)>(?:[^<]|<(?:code|strong|em|a(?: target="[^.]+") href="[^"]+")>[^<]+</(?:code|strong|em|a)>)*</(?:code|strong|em|blockquote|p|a)>|<sup id="fnref:\d+"><a href="#fn:\d+" class="footnote-ref">\d+</a></sup>|<br />)*</(?:p|h[1-6]|blockquote)>|<(?:ul|ol)>(?:\s|<li>.*?</li>)+</(?:ul|ol)>|<div class="footnotes">\n<hr />\n<ol>\n\n(?:<li id="fn:\d+">.*?</li>\n\n)*</ol>\n</div>|<pre class=[\x27"]brush: lua[\x27"]>(?:<br />)?[^<]*</pre>|<hr ?/>)(?:\r?\n)+)*).*?Id = [\x27]q(?P<qid>\d+)(?:.*?h3 class="wide-line-title">(?P<answercount>\d+) answers?</h3>\r?\n\s*<div class="question" id="(?P<aid>\d+)">[^b]+ber">(?P<ascore>\d+)[^A]+Accepted answer[^A]+Answered by</span>\r?\n\s*<div class="question-poster-username">\r?\n\s*<img src="/resources/avatar/(?P<auserid>\d+)" alt(?:="[^"]*")? class="loggedin-avatar" /><a href="/user/(?P<auser>\w+)" class="loggedin-name">\w+</a> <span class="muted small">(?P<auserscore>\d+)[^f]+float-right" title="(?P<adatetime>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})[^q]+question-markdown">(?:<div id="markdown-[a-z0-9]+">)?(?P<answer>(?:(?:<(?:p|h[1-6]|blockquote)>(?:[^<]|<(?:code|strong|em|blockquote|p|a(?: target="[^"]*")? href="[^"]+"(?: rev="footnote")?(?: class="[^"]+")?)>(?:[^<]|<(?:code|strong|em|a(?: target="[^.]+") href="[^"]+")>[^<]+</(?:code|strong|em|a)>)*</(?:code|strong|em|blockquote|p|a)>|<sup id="fnref:\d+"><a href="#fn:\d+" class="footnote-ref">\d+</a></sup>|<br />)*</(?:p|h[1-6]|blockquote)>|<(?:ul|ol)>(?:\s|<li>.*?</li>)+</(?:ul|ol)>|<div class="footnotes">\n<hr />\n<ol>\n\n(?:<li id="fn:\d+">.*?</li>\n\n)*</ol>\n</div>|<pre class=[\x27"]brush: lua[\x27"]>(?:<br />)?[^<]*</pre>|<hr ?/>)(?:\r?\n)+)*))?

questions.tar.xz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a5a79ac3cada431dc6aed57550c8905ef12de36b80f50df6aeb5eb0e889fa43
+size 4715488

questions_raw.snappy.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52d0ef1044e938d9a37fee956cec4d303fb195c09616d8e29fcc57b89f3c1438
+size 22622502

questions_tar_extract.py

@@ -0,0 +1,18 @@
+import tarfile
+from pathlib import Path
+
+script_dir = Path(__file__).resolve().parent
+scraped_files_root = "questions"
+
+def extract_articles() -> Path:
+    articles = script_dir / scraped_files_root
+    if not articles.is_dir():
+        articles_archive = script_dir / f"{scraped_files_root}.tar.xz"
+        if not articles_archive.is_file():
+            raise RuntimeError(f"{scraped_files_root} directory / {scraped_files_root}.tar.xz archive not present in cwd")
+        with tarfile.open(articles_archive, "r:*") as tar:
+            tar.extractall()
+        articles = script_dir / scraped_files_root
+        if not articles.is_dir():
+            raise RuntimeError(f"tarfile extraction failed to produce {scraped_files_root} directory")
+    return articles