signature
stringlengths
8
3.44k
body
stringlengths
0
1.41M
docstring
stringlengths
1
122k
id
stringlengths
5
17
def process_block(self, block):
ret = []<EOL>output = None<EOL>input_lines = None<EOL>lineno = self.IP.execution_count<EOL>input_prompt = self.promptin%lineno<EOL>output_prompt = self.promptout%lineno<EOL>image_file = None<EOL>image_directive = None<EOL>for token, data in block:<EOL><INDENT>if token==COMMENT:<EOL><INDENT>out_data = self.process_comment(data)<EOL><DEDENT>elif token==INPUT:<EOL><INDENT>(out_data, input_lines, output, is_doctest, image_file,<EOL>image_directive) =self.process_input(data, input_prompt, lineno)<EOL><DEDENT>elif token==OUTPUT:<EOL><INDENT>out_data =self.process_output(data, output_prompt,<EOL>input_lines, output, is_doctest,<EOL>image_file)<EOL><DEDENT>if out_data:<EOL><INDENT>ret.extend(out_data)<EOL><DEDENT><DEDENT>if image_file is not None:<EOL><INDENT>self.save_image(image_file)<EOL><DEDENT>return ret, image_directive<EOL>
process block from the block_parser and return a list of processed lines
f5151:c0:m8
def process_pure_python(self, content):
output = []<EOL>savefig = False <EOL>multiline = False <EOL>fmtin = self.promptin<EOL>for lineno, line in enumerate(content):<EOL><INDENT>line_stripped = line.strip()<EOL>if not len(line):<EOL><INDENT>output.append(line) <EOL>continue<EOL><DEDENT>if line_stripped.startswith('<STR_LIT:@>'):<EOL><INDENT>output.extend([line])<EOL>if '<STR_LIT>' in line:<EOL><INDENT>savefig = True <EOL><DEDENT>continue<EOL><DEDENT>if line_stripped.startswith('<STR_LIT:#>'):<EOL><INDENT>output.extend([line])<EOL>continue<EOL><DEDENT>if not multiline: <EOL><INDENT>if line_stripped.endswith('<STR_LIT:\\>'): <EOL><INDENT>multiline = True<EOL>cont_len = len(str(lineno)) + <NUM_LIT:2><EOL>line_to_process = line.strip('<STR_LIT:\\>')<EOL>output.extend(["<STR_LIT>" % (fmtin%lineno,line)])<EOL>continue<EOL><DEDENT>else: <EOL><INDENT>line_to_process = line.strip('<STR_LIT:\\>')<EOL><DEDENT><DEDENT>else: <EOL><INDENT>line_to_process += line.strip('<STR_LIT:\\>')<EOL>if line_stripped.endswith('<STR_LIT:\\>'): <EOL><INDENT>continuation = '<STR_LIT:.>' * cont_len<EOL>output.extend([('<STR_LIT>'+line_stripped) % continuation])<EOL>continue<EOL><DEDENT><DEDENT>self.process_input_line(str(line_to_process.strip()),<EOL>store_history=False)<EOL>out_line = self.cout.getvalue()<EOL>self.clear_cout()<EOL>if savefig:<EOL><INDENT>self.ensure_pyplot()<EOL>self.process_input_line('<STR_LIT>', store_history=False)<EOL>self.clear_cout()<EOL>savefig = False<EOL><DEDENT>if not multiline:<EOL><INDENT>in_line = "<STR_LIT>" % (fmtin%lineno,line)<EOL>output.extend([in_line])<EOL><DEDENT>else:<EOL><INDENT>output.extend([('<STR_LIT>'+line_stripped) % continuation])<EOL>multiline = False<EOL><DEDENT>if len(out_line):<EOL><INDENT>output.extend([out_line])<EOL><DEDENT>output.extend(['<STR_LIT>'])<EOL><DEDENT>return output<EOL>
content is a list of strings. it is unedited directive conent This runs it line by line in the InteractiveShell, prepends prompts as needed capturing stderr and stdout, then returns the content as a list as if it were ipython code
f5151:c0:m10
def process_pure_python2(self, content):
output = []<EOL>savefig = False <EOL>multiline = False <EOL>multiline_start = None<EOL>fmtin = self.promptin<EOL>ct = <NUM_LIT:0><EOL>content = [line for line in content if len(line.strip()) > <NUM_LIT:0>]<EOL>for lineno, line in enumerate(content):<EOL><INDENT>line_stripped = line.strip()<EOL>if not len(line):<EOL><INDENT>output.append(line)<EOL>continue<EOL><DEDENT>if line_stripped.startswith('<STR_LIT:@>'):<EOL><INDENT>output.extend([line])<EOL>if '<STR_LIT>' in line:<EOL><INDENT>savefig = True <EOL><DEDENT>continue<EOL><DEDENT>if line_stripped.startswith('<STR_LIT:#>'):<EOL><INDENT>output.extend([line])<EOL>continue<EOL><DEDENT>continuation = '<STR_LIT>'% '<STR_LIT>'.join(['<STR_LIT:.>']*(len(str(ct))+<NUM_LIT:2>))<EOL>if not multiline:<EOL><INDENT>modified = "<STR_LIT>" % (fmtin % ct, line_stripped)<EOL>output.append(modified)<EOL>ct += <NUM_LIT:1><EOL>try:<EOL><INDENT>ast.parse(line_stripped)<EOL>output.append('<STR_LIT>')<EOL><DEDENT>except Exception:<EOL><INDENT>multiline = True<EOL>multiline_start = lineno<EOL><DEDENT><DEDENT>else:<EOL><INDENT>modified = '<STR_LIT>' % (continuation, line)<EOL>output.append(modified)<EOL>try:<EOL><INDENT>ast.parse('<STR_LIT:\n>'.join(content[multiline_start:lineno+<NUM_LIT:1>]))<EOL>if (lineno < len(content) - <NUM_LIT:1> and<EOL>_count_indent(content[multiline_start]) <<EOL>_count_indent(content[lineno + <NUM_LIT:1>])):<EOL><INDENT>continue<EOL><DEDENT>output.extend([continuation, '<STR_LIT>'])<EOL>multiline = False<EOL><DEDENT>except Exception:<EOL><INDENT>pass<EOL><DEDENT><DEDENT>continue<EOL><DEDENT>return output<EOL>
content is a list of strings. it is unedited directive conent This runs it line by line in the InteractiveShell, prepends prompts as needed capturing stderr and stdout, then returns the content as a list as if it were ipython code
f5151:c0:m11
def setup(app):
<EOL>pass<EOL>
Setup as a sphinx extension.
f5152:m0
def __init__(self, weights_file: str=None, min_face_size: int=<NUM_LIT:20>, steps_threshold: list=None,<EOL>scale_factor: float=<NUM_LIT>):
if steps_threshold is None:<EOL><INDENT>steps_threshold = [<NUM_LIT>, <NUM_LIT>, <NUM_LIT>]<EOL><DEDENT>if weights_file is None:<EOL><INDENT>weights_file = pkg_resources.resource_stream('<STR_LIT>', '<STR_LIT>')<EOL><DEDENT>self.__min_face_size = min_face_size<EOL>self.__steps_threshold = steps_threshold<EOL>self.__scale_factor = scale_factor<EOL>config = tf.ConfigProto(log_device_placement=False)<EOL>config.gpu_options.allow_growth = True<EOL>self.__graph = tf.Graph()<EOL>with self.__graph.as_default():<EOL><INDENT>self.__session = tf.Session(config=config, graph=self.__graph)<EOL>weights = np.load(weights_file).item()<EOL>self.__pnet = PNet(self.__session, False)<EOL>self.__pnet.set_weights(weights['<STR_LIT>'])<EOL>self.__rnet = RNet(self.__session, False)<EOL>self.__rnet.set_weights(weights['<STR_LIT>'])<EOL>self.__onet = ONet(self.__session, False)<EOL>self.__onet.set_weights(weights['<STR_LIT>'])<EOL><DEDENT>weights_file.close()<EOL>
Initializes the MTCNN. :param weights_file: file uri with the weights of the P, R and O networks from MTCNN. By default it will load the ones bundled with the package. :param min_face_size: minimum size of the face to detect :param steps_threshold: step's thresholds values :param scale_factor: scale factor
f5158:c4:m0
@staticmethod<EOL><INDENT>def __scale_image(image, scale: float):<DEDENT>
height, width, _ = image.shape<EOL>width_scaled = int(np.ceil(width * scale))<EOL>height_scaled = int(np.ceil(height * scale))<EOL>im_data = cv2.resize(image, (width_scaled, height_scaled), interpolation=cv2.INTER_AREA)<EOL>im_data_normalized = (im_data - <NUM_LIT>) * <NUM_LIT><EOL>return im_data_normalized<EOL>
Scales the image to a given scale. :param image: :param scale: :return:
f5158:c4:m4
@staticmethod<EOL><INDENT>def __nms(boxes, threshold, method):<DEDENT>
if boxes.size == <NUM_LIT:0>:<EOL><INDENT>return np.empty((<NUM_LIT:0>, <NUM_LIT:3>))<EOL><DEDENT>x1 = boxes[:, <NUM_LIT:0>]<EOL>y1 = boxes[:, <NUM_LIT:1>]<EOL>x2 = boxes[:, <NUM_LIT:2>]<EOL>y2 = boxes[:, <NUM_LIT:3>]<EOL>s = boxes[:, <NUM_LIT:4>]<EOL>area = (x2 - x1 + <NUM_LIT:1>) * (y2 - y1 + <NUM_LIT:1>)<EOL>sorted_s = np.argsort(s)<EOL>pick = np.zeros_like(s, dtype=np.int16)<EOL>counter = <NUM_LIT:0><EOL>while sorted_s.size > <NUM_LIT:0>:<EOL><INDENT>i = sorted_s[-<NUM_LIT:1>]<EOL>pick[counter] = i<EOL>counter += <NUM_LIT:1><EOL>idx = sorted_s[<NUM_LIT:0>:-<NUM_LIT:1>]<EOL>xx1 = np.maximum(x1[i], x1[idx])<EOL>yy1 = np.maximum(y1[i], y1[idx])<EOL>xx2 = np.minimum(x2[i], x2[idx])<EOL>yy2 = np.minimum(y2[i], y2[idx])<EOL>w = np.maximum(<NUM_LIT:0.0>, xx2 - xx1 + <NUM_LIT:1>)<EOL>h = np.maximum(<NUM_LIT:0.0>, yy2 - yy1 + <NUM_LIT:1>)<EOL>inter = w * h<EOL>if method is '<STR_LIT>':<EOL><INDENT>o = inter / np.minimum(area[i], area[idx])<EOL><DEDENT>else:<EOL><INDENT>o = inter / (area[i] + area[idx] - inter)<EOL><DEDENT>sorted_s = sorted_s[np.where(o <= threshold)]<EOL><DEDENT>pick = pick[<NUM_LIT:0>:counter]<EOL>return pick<EOL>
Non Maximum Suppression. :param boxes: np array with bounding boxes. :param threshold: :param method: NMS method to apply. Available values ('Min', 'Union') :return:
f5158:c4:m6
def detect_faces(self, img) -> list:
if img is None or not hasattr(img, "<STR_LIT>"):<EOL><INDENT>raise InvalidImage("<STR_LIT>")<EOL><DEDENT>height, width, _ = img.shape<EOL>stage_status = StageStatus(width=width, height=height)<EOL>m = <NUM_LIT:12> / self.__min_face_size<EOL>min_layer = np.amin([height, width]) * m<EOL>scales = self.__compute_scale_pyramid(m, min_layer)<EOL>stages = [self.__stage1, self.__stage2, self.__stage3]<EOL>result = [scales, stage_status]<EOL>for stage in stages:<EOL><INDENT>result = stage(img, result[<NUM_LIT:0>], result[<NUM_LIT:1>])<EOL><DEDENT>[total_boxes, points] = result<EOL>bounding_boxes = []<EOL>for bounding_box, keypoints in zip(total_boxes, points.T):<EOL><INDENT>bounding_boxes.append({<EOL>'<STR_LIT>': [int(bounding_box[<NUM_LIT:0>]), int(bounding_box[<NUM_LIT:1>]),<EOL>int(bounding_box[<NUM_LIT:2>]-bounding_box[<NUM_LIT:0>]), int(bounding_box[<NUM_LIT:3>]-bounding_box[<NUM_LIT:1>])],<EOL>'<STR_LIT>': bounding_box[-<NUM_LIT:1>],<EOL>'<STR_LIT>': {<EOL>'<STR_LIT>': (int(keypoints[<NUM_LIT:0>]), int(keypoints[<NUM_LIT:5>])),<EOL>'<STR_LIT>': (int(keypoints[<NUM_LIT:1>]), int(keypoints[<NUM_LIT:6>])),<EOL>'<STR_LIT>': (int(keypoints[<NUM_LIT:2>]), int(keypoints[<NUM_LIT:7>])),<EOL>'<STR_LIT>': (int(keypoints[<NUM_LIT:3>]), int(keypoints[<NUM_LIT:8>])),<EOL>'<STR_LIT>': (int(keypoints[<NUM_LIT:4>]), int(keypoints[<NUM_LIT:9>])),<EOL>}<EOL>}<EOL>)<EOL><DEDENT>return bounding_boxes<EOL>
Detects bounding boxes from the specified image. :param img: image to process :return: list containing all the bounding boxes detected with their keypoints.
f5158:c4:m10
def __stage1(self, image, scales: list, stage_status: StageStatus):
total_boxes = np.empty((<NUM_LIT:0>, <NUM_LIT:9>))<EOL>status = stage_status<EOL>for scale in scales:<EOL><INDENT>scaled_image = self.__scale_image(image, scale)<EOL>img_x = np.expand_dims(scaled_image, <NUM_LIT:0>)<EOL>img_y = np.transpose(img_x, (<NUM_LIT:0>, <NUM_LIT:2>, <NUM_LIT:1>, <NUM_LIT:3>))<EOL>out = self.__pnet.feed(img_y)<EOL>out0 = np.transpose(out[<NUM_LIT:0>], (<NUM_LIT:0>, <NUM_LIT:2>, <NUM_LIT:1>, <NUM_LIT:3>))<EOL>out1 = np.transpose(out[<NUM_LIT:1>], (<NUM_LIT:0>, <NUM_LIT:2>, <NUM_LIT:1>, <NUM_LIT:3>))<EOL>boxes, _ = self.__generate_bounding_box(out1[<NUM_LIT:0>, :, :, <NUM_LIT:1>].copy(),<EOL>out0[<NUM_LIT:0>, :, :, :].copy(), scale, self.__steps_threshold[<NUM_LIT:0>])<EOL>pick = self.__nms(boxes.copy(), <NUM_LIT:0.5>, '<STR_LIT>')<EOL>if boxes.size > <NUM_LIT:0> and pick.size > <NUM_LIT:0>:<EOL><INDENT>boxes = boxes[pick, :]<EOL>total_boxes = np.append(total_boxes, boxes, axis=<NUM_LIT:0>)<EOL><DEDENT><DEDENT>numboxes = total_boxes.shape[<NUM_LIT:0>]<EOL>if numboxes > <NUM_LIT:0>:<EOL><INDENT>pick = self.__nms(total_boxes.copy(), <NUM_LIT>, '<STR_LIT>')<EOL>total_boxes = total_boxes[pick, :]<EOL>regw = total_boxes[:, <NUM_LIT:2>] - total_boxes[:, <NUM_LIT:0>]<EOL>regh = total_boxes[:, <NUM_LIT:3>] - total_boxes[:, <NUM_LIT:1>]<EOL>qq1 = total_boxes[:, <NUM_LIT:0>] + total_boxes[:, <NUM_LIT:5>] * regw<EOL>qq2 = total_boxes[:, <NUM_LIT:1>] + total_boxes[:, <NUM_LIT:6>] * regh<EOL>qq3 = total_boxes[:, <NUM_LIT:2>] + total_boxes[:, <NUM_LIT:7>] * regw<EOL>qq4 = total_boxes[:, <NUM_LIT:3>] + total_boxes[:, <NUM_LIT:8>] * regh<EOL>total_boxes = np.transpose(np.vstack([qq1, qq2, qq3, qq4, total_boxes[:, <NUM_LIT:4>]]))<EOL>total_boxes = self.__rerec(total_boxes.copy())<EOL>total_boxes[:, <NUM_LIT:0>:<NUM_LIT:4>] = np.fix(total_boxes[:, <NUM_LIT:0>:<NUM_LIT:4>]).astype(np.int32)<EOL>status = StageStatus(self.__pad(total_boxes.copy(), stage_status.width, stage_status.height),<EOL>width=stage_status.width, height=stage_status.height)<EOL><DEDENT>return total_boxes, status<EOL>
First stage of the MTCNN. :param image: :param scales: :param stage_status: :return:
f5158:c4:m11
def __stage2(self, img, total_boxes, stage_status:StageStatus):
num_boxes = total_boxes.shape[<NUM_LIT:0>]<EOL>if num_boxes == <NUM_LIT:0>:<EOL><INDENT>return total_boxes, stage_status<EOL><DEDENT>tempimg = np.zeros(shape=(<NUM_LIT>, <NUM_LIT>, <NUM_LIT:3>, num_boxes))<EOL>for k in range(<NUM_LIT:0>, num_boxes):<EOL><INDENT>tmp = np.zeros((int(stage_status.tmph[k]), int(stage_status.tmpw[k]), <NUM_LIT:3>))<EOL>tmp[stage_status.dy[k] - <NUM_LIT:1>:stage_status.edy[k], stage_status.dx[k] - <NUM_LIT:1>:stage_status.edx[k], :] =img[stage_status.y[k] - <NUM_LIT:1>:stage_status.ey[k], stage_status.x[k] - <NUM_LIT:1>:stage_status.ex[k], :]<EOL>if tmp.shape[<NUM_LIT:0>] > <NUM_LIT:0> and tmp.shape[<NUM_LIT:1>] > <NUM_LIT:0> or tmp.shape[<NUM_LIT:0>] == <NUM_LIT:0> and tmp.shape[<NUM_LIT:1>] == <NUM_LIT:0>:<EOL><INDENT>tempimg[:, :, :, k] = cv2.resize(tmp, (<NUM_LIT>, <NUM_LIT>), interpolation=cv2.INTER_AREA)<EOL><DEDENT>else:<EOL><INDENT>return np.empty(shape=(<NUM_LIT:0>,)), stage_status<EOL><DEDENT><DEDENT>tempimg = (tempimg - <NUM_LIT>) * <NUM_LIT><EOL>tempimg1 = np.transpose(tempimg, (<NUM_LIT:3>, <NUM_LIT:1>, <NUM_LIT:0>, <NUM_LIT:2>))<EOL>out = self.__rnet.feed(tempimg1)<EOL>out0 = np.transpose(out[<NUM_LIT:0>])<EOL>out1 = np.transpose(out[<NUM_LIT:1>])<EOL>score = out1[<NUM_LIT:1>, :]<EOL>ipass = np.where(score > self.__steps_threshold[<NUM_LIT:1>])<EOL>total_boxes = np.hstack([total_boxes[ipass[<NUM_LIT:0>], <NUM_LIT:0>:<NUM_LIT:4>].copy(), np.expand_dims(score[ipass].copy(), <NUM_LIT:1>)])<EOL>mv = out0[:, ipass[<NUM_LIT:0>]]<EOL>if total_boxes.shape[<NUM_LIT:0>] > <NUM_LIT:0>:<EOL><INDENT>pick = self.__nms(total_boxes, <NUM_LIT>, '<STR_LIT>')<EOL>total_boxes = total_boxes[pick, :]<EOL>total_boxes = self.__bbreg(total_boxes.copy(), np.transpose(mv[:, pick]))<EOL>total_boxes = self.__rerec(total_boxes.copy())<EOL><DEDENT>return total_boxes, stage_status<EOL>
Second stage of the MTCNN. :param img: :param total_boxes: :param stage_status: :return:
f5158:c4:m12
def __stage3(self, img, total_boxes, stage_status: StageStatus):
num_boxes = total_boxes.shape[<NUM_LIT:0>]<EOL>if num_boxes == <NUM_LIT:0>:<EOL><INDENT>return total_boxes, np.empty(shape=(<NUM_LIT:0>,))<EOL><DEDENT>total_boxes = np.fix(total_boxes).astype(np.int32)<EOL>status = StageStatus(self.__pad(total_boxes.copy(), stage_status.width, stage_status.height),<EOL>width=stage_status.width, height=stage_status.height)<EOL>tempimg = np.zeros((<NUM_LIT>, <NUM_LIT>, <NUM_LIT:3>, num_boxes))<EOL>for k in range(<NUM_LIT:0>, num_boxes):<EOL><INDENT>tmp = np.zeros((int(status.tmph[k]), int(status.tmpw[k]), <NUM_LIT:3>))<EOL>tmp[status.dy[k] - <NUM_LIT:1>:status.edy[k], status.dx[k] - <NUM_LIT:1>:status.edx[k], :] =img[status.y[k] - <NUM_LIT:1>:status.ey[k], status.x[k] - <NUM_LIT:1>:status.ex[k], :]<EOL>if tmp.shape[<NUM_LIT:0>] > <NUM_LIT:0> and tmp.shape[<NUM_LIT:1>] > <NUM_LIT:0> or tmp.shape[<NUM_LIT:0>] == <NUM_LIT:0> and tmp.shape[<NUM_LIT:1>] == <NUM_LIT:0>:<EOL><INDENT>tempimg[:, :, :, k] = cv2.resize(tmp, (<NUM_LIT>, <NUM_LIT>), interpolation=cv2.INTER_AREA)<EOL><DEDENT>else:<EOL><INDENT>return np.empty(shape=(<NUM_LIT:0>,)), np.empty(shape=(<NUM_LIT:0>,))<EOL><DEDENT><DEDENT>tempimg = (tempimg - <NUM_LIT>) * <NUM_LIT><EOL>tempimg1 = np.transpose(tempimg, (<NUM_LIT:3>, <NUM_LIT:1>, <NUM_LIT:0>, <NUM_LIT:2>))<EOL>out = self.__onet.feed(tempimg1)<EOL>out0 = np.transpose(out[<NUM_LIT:0>])<EOL>out1 = np.transpose(out[<NUM_LIT:1>])<EOL>out2 = np.transpose(out[<NUM_LIT:2>])<EOL>score = out2[<NUM_LIT:1>, :]<EOL>points = out1<EOL>ipass = np.where(score > self.__steps_threshold[<NUM_LIT:2>])<EOL>points = points[:, ipass[<NUM_LIT:0>]]<EOL>total_boxes = np.hstack([total_boxes[ipass[<NUM_LIT:0>], <NUM_LIT:0>:<NUM_LIT:4>].copy(), np.expand_dims(score[ipass].copy(), <NUM_LIT:1>)])<EOL>mv = out0[:, ipass[<NUM_LIT:0>]]<EOL>w = total_boxes[:, <NUM_LIT:2>] - total_boxes[:, <NUM_LIT:0>] + <NUM_LIT:1><EOL>h = total_boxes[:, <NUM_LIT:3>] - total_boxes[:, <NUM_LIT:1>] + <NUM_LIT:1><EOL>points[<NUM_LIT:0>:<NUM_LIT:5>, :] = np.tile(w, (<NUM_LIT:5>, <NUM_LIT:1>)) * points[<NUM_LIT:0>:<NUM_LIT:5>, :] + np.tile(total_boxes[:, <NUM_LIT:0>], (<NUM_LIT:5>, <NUM_LIT:1>)) - <NUM_LIT:1><EOL>points[<NUM_LIT:5>:<NUM_LIT:10>, :] = np.tile(h, (<NUM_LIT:5>, <NUM_LIT:1>)) * points[<NUM_LIT:5>:<NUM_LIT:10>, :] + np.tile(total_boxes[:, <NUM_LIT:1>], (<NUM_LIT:5>, <NUM_LIT:1>)) - <NUM_LIT:1><EOL>if total_boxes.shape[<NUM_LIT:0>] > <NUM_LIT:0>:<EOL><INDENT>total_boxes = self.__bbreg(total_boxes.copy(), np.transpose(mv))<EOL>pick = self.__nms(total_boxes.copy(), <NUM_LIT>, '<STR_LIT>')<EOL>total_boxes = total_boxes[pick, :]<EOL>points = points[:, pick]<EOL><DEDENT>return total_boxes, points<EOL>
Third stage of the MTCNN. :param img: :param total_boxes: :param stage_status: :return:
f5158:c4:m13
def __init__(self, session, trainable: bool=True):
self._session = session<EOL>self.__trainable = trainable<EOL>self.__layers = {}<EOL>self.__last_layer_name = None<EOL>with tf.variable_scope(self.__class__.__name__.lower()):<EOL><INDENT>self._config()<EOL><DEDENT>
Initializes the network. :param trainable: flag to determine if this network should be trainable or not.
f5159:c0:m0
def _config(self):
raise NotImplementedError("<STR_LIT>")<EOL>
Configures the network layers. It is usually done using the LayerFactory() class.
f5159:c0:m1
def add_layer(self, name: str, layer_output):
self.__layers[name] = layer_output<EOL>self.__last_layer_name = name<EOL>
Adds a layer to the network. :param name: name of the layer to add :param layer_output: output layer.
f5159:c0:m2
def get_layer(self, name: str=None):
if name is None:<EOL><INDENT>name = self.__last_layer_name<EOL><DEDENT>return self.__layers[name]<EOL>
Retrieves the layer by its name. :param name: name of the layer to retrieve. If name is None, it will retrieve the last added layer to the network. :return: layer output
f5159:c0:m3
def is_trainable(self):
return self.__trainable<EOL>
Getter for the trainable flag.
f5159:c0:m4
def set_weights(self, weights_values: dict, ignore_missing=False):
network_name = self.__class__.__name__.lower()<EOL>with tf.variable_scope(network_name):<EOL><INDENT>for layer_name in weights_values:<EOL><INDENT>with tf.variable_scope(layer_name, reuse=True):<EOL><INDENT>for param_name, data in weights_values[layer_name].items():<EOL><INDENT>try:<EOL><INDENT>var = tf.get_variable(param_name)<EOL>self._session.run(var.assign(data))<EOL><DEDENT>except ValueError:<EOL><INDENT>if not ignore_missing:<EOL><INDENT>raise<EOL><DEDENT><DEDENT><DEDENT><DEDENT><DEDENT><DEDENT>
Sets the weights values of the network. :param weights_values: dictionary with weights for each layer
f5159:c0:m5
def feed(self, image):
network_name = self.__class__.__name__.lower()<EOL>with tf.variable_scope(network_name):<EOL><INDENT>return self._feed(image)<EOL><DEDENT>
Feeds the network with an image :param image: image (perhaps loaded with CV2) :return: network result
f5159:c0:m6
def __make_var(self, name: str, shape: list):
return tf.get_variable(name, shape, trainable=self.__network.is_trainable())<EOL>
Creates a tensorflow variable with the given name and shape. :param name: name to set for the variable. :param shape: list defining the shape of the variable. :return: created TF variable.
f5160:c0:m4
def new_feed(self, name: str, layer_shape: tuple):
feed_data = tf.placeholder(tf.float32, layer_shape, '<STR_LIT:input>')<EOL>self.__network.add_layer(name, layer_output=feed_data)<EOL>
Creates a feed layer. This is usually the first layer in the network. :param name: name of the layer :return:
f5160:c0:m5
def new_conv(self, name: str, kernel_size: tuple, channels_output: int,<EOL>stride_size: tuple, padding: str='<STR_LIT>',<EOL>group: int=<NUM_LIT:1>, biased: bool=True, relu: bool=True, input_layer_name: str=None):
<EOL>self.__validate_padding(padding)<EOL>input_layer = self.__network.get_layer(input_layer_name)<EOL>channels_input = int(input_layer.get_shape()[-<NUM_LIT:1>])<EOL>self.__validate_grouping(channels_input, channels_output, group)<EOL>convolve = lambda input_val, kernel: tf.nn.conv2d(input_val, kernel, [<NUM_LIT:1>, stride_size[<NUM_LIT:1>], stride_size[<NUM_LIT:0>], <NUM_LIT:1>],<EOL>padding=padding)<EOL>with tf.variable_scope(name) as scope:<EOL><INDENT>kernel = self.__make_var('<STR_LIT>', shape=[kernel_size[<NUM_LIT:1>], kernel_size[<NUM_LIT:0>], channels_input // group, channels_output])<EOL>output = convolve(input_layer, kernel)<EOL>if biased:<EOL><INDENT>biases = self.__make_var('<STR_LIT>', [channels_output])<EOL>output = tf.nn.bias_add(output, biases)<EOL><DEDENT>if relu:<EOL><INDENT>output = tf.nn.relu(output, name=scope.name)<EOL><DEDENT><DEDENT>self.__network.add_layer(name, layer_output=output)<EOL>
Creates a convolution layer for the network. :param name: name for the layer :param kernel_size: tuple containing the size of the kernel (Width, Height) :param channels_output: ¿? Perhaps number of channels in the output? it is used as the bias size. :param stride_size: tuple containing the size of the stride (Width, Height) :param padding: Type of padding. Available values are: ('SAME', 'VALID') :param group: groups for the kernel operation. More info required. :param biased: boolean flag to set if biased or not. :param relu: boolean flag to set if ReLu should be applied at the end of the layer or not. :param input_layer_name: name of the input layer for this layer. If None, it will take the last added layer of the network.
f5160:c0:m6
def new_prelu(self, name: str, input_layer_name: str=None):
input_layer = self.__network.get_layer(input_layer_name)<EOL>with tf.variable_scope(name):<EOL><INDENT>channels_input = int(input_layer.get_shape()[-<NUM_LIT:1>])<EOL>alpha = self.__make_var('<STR_LIT>', shape=[channels_input])<EOL>output = tf.nn.relu(input_layer) + tf.multiply(alpha, -tf.nn.relu(-input_layer))<EOL><DEDENT>self.__network.add_layer(name, layer_output=output)<EOL>
Creates a new prelu layer with the given name and input. :param name: name for this layer. :param input_layer_name: name of the layer that serves as input for this one.
f5160:c0:m7
def new_max_pool(self, name:str, kernel_size: tuple, stride_size: tuple, padding='<STR_LIT>',<EOL>input_layer_name: str=None):
self.__validate_padding(padding)<EOL>input_layer = self.__network.get_layer(input_layer_name)<EOL>output = tf.nn.max_pool(input_layer,<EOL>ksize=[<NUM_LIT:1>, kernel_size[<NUM_LIT:1>], kernel_size[<NUM_LIT:0>], <NUM_LIT:1>],<EOL>strides=[<NUM_LIT:1>, stride_size[<NUM_LIT:1>], stride_size[<NUM_LIT:0>], <NUM_LIT:1>],<EOL>padding=padding,<EOL>name=name)<EOL>self.__network.add_layer(name, layer_output=output)<EOL>
Creates a new max pooling layer. :param name: name for the layer. :param kernel_size: tuple containing the size of the kernel (Width, Height) :param stride_size: tuple containing the size of the stride (Width, Height) :param padding: Type of padding. Available values are: ('SAME', 'VALID') :param input_layer_name: name of the input layer for this layer. If None, it will take the last added layer of the network.
f5160:c0:m8
def new_fully_connected(self, name: str, output_count: int, relu=True, input_layer_name: str=None):
with tf.variable_scope(name):<EOL><INDENT>input_layer = self.__network.get_layer(input_layer_name)<EOL>vectorized_input, dimension = self.vectorize_input(input_layer)<EOL>weights = self.__make_var('<STR_LIT>', shape=[dimension, output_count])<EOL>biases = self.__make_var('<STR_LIT>', shape=[output_count])<EOL>operation = tf.nn.relu_layer if relu else tf.nn.xw_plus_b<EOL>fc = operation(vectorized_input, weights, biases, name=name)<EOL><DEDENT>self.__network.add_layer(name, layer_output=fc)<EOL>
Creates a new fully connected layer. :param name: name for the layer. :param output_count: number of outputs of the fully connected layer. :param relu: boolean flag to set if ReLu should be applied at the end of this layer. :param input_layer_name: name of the input layer for this layer. If None, it will take the last added layer of the network.
f5160:c0:m9
def new_softmax(self, name, axis, input_layer_name: str=None):
input_layer = self.__network.get_layer(input_layer_name)<EOL>if LooseVersion(tf.__version__) < LooseVersion("<STR_LIT>"):<EOL><INDENT>max_axis = tf.reduce_max(input_layer, axis, keep_dims=True)<EOL>target_exp = tf.exp(input_layer - max_axis)<EOL>normalize = tf.reduce_sum(target_exp, axis, keep_dims=True)<EOL><DEDENT>else:<EOL><INDENT>max_axis = tf.reduce_max(input_layer, axis, keepdims=True)<EOL>target_exp = tf.exp(input_layer - max_axis)<EOL>normalize = tf.reduce_sum(target_exp, axis, keepdims=True)<EOL><DEDENT>softmax = tf.div(target_exp, normalize, name)<EOL>self.__network.add_layer(name, layer_output=softmax)<EOL>
Creates a new softmax layer :param name: name to set for the layer :param axis: :param input_layer_name: name of the input layer for this layer. If None, it will take the last added layer of the network.
f5160:c0:m10
def dameraulevenshtein(seq1, seq2):
<EOL>oneago = None<EOL>thisrow = list(range_(<NUM_LIT:1>, len(seq2) + <NUM_LIT:1>)) + [<NUM_LIT:0>]<EOL>for x in range_(len(seq1)):<EOL><INDENT>twoago, oneago, thisrow = oneago, thisrow, [<NUM_LIT:0>] * len(seq2) + [x + <NUM_LIT:1>]<EOL>for y in range_(len(seq2)):<EOL><INDENT>delcost = oneago[y] + <NUM_LIT:1><EOL>addcost = thisrow[y - <NUM_LIT:1>] + <NUM_LIT:1><EOL>subcost = oneago[y - <NUM_LIT:1>] + (seq1[x] != seq2[y])<EOL>thisrow[y] = min(delcost, addcost, subcost)<EOL>if (x > <NUM_LIT:0> and y > <NUM_LIT:0> and seq1[x] == seq2[y - <NUM_LIT:1>] and<EOL>seq1[x - <NUM_LIT:1>] == seq2[y] and seq1[x] != seq2[y]):<EOL><INDENT>thisrow[y] = min(thisrow[y], twoago[y - <NUM_LIT:2>] + <NUM_LIT:1>)<EOL><DEDENT><DEDENT><DEDENT>return thisrow[len(seq2) - <NUM_LIT:1>]<EOL>
Calculate the Damerau-Levenshtein distance between sequences. This distance is the number of additions, deletions, substitutions, and transpositions needed to transform the first sequence into the second. Although generally used with strings, any sequences of comparable objects will work. Transpositions are exchanges of *consecutive* characters; all other operations are self-explanatory. This implementation is O(N*M) time and O(M) space, for N and M the lengths of the two sequences. >>> dameraulevenshtein('ba', 'abc') 2 >>> dameraulevenshtein('fee', 'deed') 2 It works with arbitrary sequences too: >>> dameraulevenshtein('abcd', ['b', 'a', 'c', 'd', 'e']) 2
f5162:m0
def evaluation_metrics(predicted, actual, bow=True):
if bow:<EOL><INDENT>p = set(predicted)<EOL>a = set(actual)<EOL>true_positive = <NUM_LIT:0><EOL>for token in p:<EOL><INDENT>if token in a:<EOL><INDENT>true_positive += <NUM_LIT:1><EOL><DEDENT><DEDENT><DEDENT>else:<EOL><INDENT>from collections import defaultdict<EOL>act = defaultdict(lambda: <NUM_LIT:0>)<EOL>for token in actual:<EOL><INDENT>act[token] += <NUM_LIT:1><EOL><DEDENT>true_positive = <NUM_LIT:0><EOL>for token in predicted:<EOL><INDENT>if act[token] > <NUM_LIT:0>:<EOL><INDENT>true_positive += <NUM_LIT:1><EOL>act[token] -= <NUM_LIT:1><EOL><DEDENT><DEDENT>p = predicted<EOL>a = actual<EOL><DEDENT>try:<EOL><INDENT>precision = true_positive / len(p)<EOL><DEDENT>except ZeroDivisionError:<EOL><INDENT>precision = <NUM_LIT:0.0><EOL><DEDENT>try:<EOL><INDENT>recall = true_positive / len(a)<EOL><DEDENT>except ZeroDivisionError:<EOL><INDENT>recall = <NUM_LIT:0.0><EOL><DEDENT>try:<EOL><INDENT>f1 = <NUM_LIT> * (precision * recall) / (precision + recall)<EOL><DEDENT>except ZeroDivisionError:<EOL><INDENT>f1 = <NUM_LIT:0.0><EOL><DEDENT>return (precision, recall, f1)<EOL>
Input: predicted, actual = lists of the predicted and actual tokens bow: if true use bag of words assumption Returns: precision, recall, F1, Levenshtein distance
f5162:m1
def get_and_union_features(features):
if not features:<EOL><INDENT>raise ValueError('<STR_LIT>')<EOL><DEDENT>if isinstance(features, (list, tuple)):<EOL><INDENT>if isinstance(features[<NUM_LIT:0>], tuple):<EOL><INDENT>return FeatureUnion(features)<EOL><DEDENT>elif isinstance(features[<NUM_LIT:0>], string_):<EOL><INDENT>return FeatureUnion([(feature, get_feature(feature)) for feature in features])<EOL><DEDENT>else:<EOL><INDENT>return make_union(*features)<EOL><DEDENT><DEDENT>elif isinstance(features, string_):<EOL><INDENT>return get_feature(features)<EOL><DEDENT>else:<EOL><INDENT>return features<EOL><DEDENT>
Get and combine features in a :class:`FeatureUnion`. Args: features (str or List[str], ``Features`` or List[``Features``], or List[Tuple[str, ``Features``]]): One or more features to be used to transform blocks into a matrix of numeric values. If more than one, a :class:`FeatureUnion` is automatically constructed. Example inputs:: features = 'weninger' features = ['weninger', 'kohlschuetter'] features = WeningerFeatures() features = [WeningerFeatures(), KohlschuetterFeatures()] features = [('weninger', WeningerFeatures()), ('kohlschuetter', KohlschuetterFeatures())] Returns: :class:`FeatureUnion` or ``Features``
f5162:m2
def load_pickled_model(filename, dirname=None):
if dirname is None:<EOL><INDENT>pkg_filename = pkgutil.get_loader('<STR_LIT>').get_filename('<STR_LIT>')<EOL>pkg_dirname = os.path.dirname(pkg_filename)<EOL>dirname = os.path.join(pkg_dirname, '<STR_LIT>', model_path)<EOL><DEDENT>filepath = os.path.join(dirname, filename)<EOL>return joblib.load(filepath)<EOL>
Load a pickled ``Extractor`` model from disk. Args: filename (str): Name of pickled model file under ``dirname``. dirname (str): Name of directory on disk containing the pickled model. If None, dragnet's default pickled model directory is used: /path/to/dragnet/pickled_models/[PY_VERSION]_[SKLEARN_VERSION] Returns: :class:`dragnet.extractor.Extractor`
f5162:m3
def str_cast(maybe_bytes, encoding='<STR_LIT:utf-8>'):
if isinstance(maybe_bytes, bytes_):<EOL><INDENT>return maybe_bytes.decode(encoding)<EOL><DEDENT>else:<EOL><INDENT>return maybe_bytes<EOL><DEDENT>
Converts any bytes-like input to a string-like output, with respect to python version Parameters ---------- maybe_bytes : if this is a bytes-like object, it will be converted to a string encoding : str, default='utf-8' encoding to be used when decoding bytes
f5163:m0
def bytes_cast(maybe_str, encoding='<STR_LIT:utf-8>'):
if isinstance(maybe_str, unicode_):<EOL><INDENT>return maybe_str.encode(encoding)<EOL><DEDENT>else:<EOL><INDENT>return maybe_str<EOL><DEDENT>
Converts any string-like input to a bytes-like output, with respect to python version Parameters ---------- maybe_str : if this is a string-like object, it will be converted to bytes encoding : str, default='utf-8' encoding to be used when encoding string
f5163:m1
def str_list_cast(list_, **kwargs):
return [str_cast(elem, **kwargs) for elem in list_]<EOL>
Converts any bytes-like items in input list to string-like values, with respect to python version Parameters ---------- list_ : list any bytes-like objects contained in the list will be converted to strings kwargs: encoding: str, default: 'utf-8' encoding to be used when decoding bytes
f5163:m2
def bytes_list_cast(list_, **kwargs):
return [bytes_cast(elem, **kwargs) for elem in list_]<EOL>
Converts any string-like items in input list to bytes-like values, with respect to python version Parameters ---------- list_ : list any string-like objects contained in the list will be converted to bytes kwargs: encoding: str, default: 'utf-8' encoding to be used when encoding string
f5163:m3
def str_dict_cast(dict_, include_keys=True, include_vals=True, **kwargs):
new_keys = str_list_cast(dict_.keys(), **kwargs) if include_keys else dict_.keys()<EOL>new_vals = str_list_cast(dict_.values(), **kwargs) if include_vals else dict_.values()<EOL>new_dict = dict(zip_(new_keys, new_vals))<EOL>return new_dict<EOL>
Converts any bytes-like items in input dict to string-like values, with respect to python version Parameters ---------- dict_ : dict any bytes-like objects contained in the dict will be converted to a string include_keys : bool, default=True if True, cast keys to a string, else ignore include_values : bool, default=True if True, cast values to a string, else ignore kwargs: encoding: str, default: 'utf-8' encoding to be used when decoding bytes
f5163:m4
def bytes_dict_cast(dict_, include_keys=True, include_vals=True, **kwargs):
new_keys = bytes_list_cast(dict_.keys(), **kwargs) if include_keys else dict_.keys()<EOL>new_vals = bytes_list_cast(dict_.values(), **kwargs) if include_vals else dict_.values()<EOL>new_dict = dict(zip_(new_keys, new_vals))<EOL>return new_dict<EOL>
Converts any string-like items in input dict to bytes-like values, with respect to python version Parameters ---------- dict_ : dict any string-like objects contained in the dict will be converted to bytes include_keys : bool, default=True if True, cast keys to bytes, else ignore include_values : bool, default=True if True, cast values to bytes, else ignore kwargs: encoding: str, default: 'utf-8' encoding to be used when encoding string
f5163:m5
def str_block_cast(block,<EOL>include_text=True,<EOL>include_link_tokens=True,<EOL>include_css=True,<EOL>include_features=True,<EOL>**kwargs):
if include_text:<EOL><INDENT>block.text = str_cast(block.text, **kwargs)<EOL><DEDENT>if include_link_tokens:<EOL><INDENT>block.link_tokens = str_list_cast(block.link_tokens, **kwargs)<EOL><DEDENT>if include_css:<EOL><INDENT>block.css = str_dict_cast(block.css, **kwargs)<EOL><DEDENT>if include_features:<EOL><INDENT>block.features = str_dict_cast(block.features, **kwargs)<EOL><DEDENT>return block<EOL>
Converts any bytes-like items in input Block object to string-like values, with respect to python version Parameters ---------- block : blocks.Block any bytes-like objects contained in the block object will be converted to a string include_text : bool, default=True if True, cast text to a string, else ignore include_link_tokens : bool, default=True if True, cast link_tokens to a string, else ignore include_css : bool, default=True if True, cast css to a string, else ignore include_features : bool, default=True if True, cast features to a string, else ignore kwargs: encoding: str, default: 'utf-8' encoding to be used when decoding bytes
f5163:m6
def bytes_block_cast(block,<EOL>include_text=True,<EOL>include_link_tokens=True,<EOL>include_css=True,<EOL>include_features=True,<EOL>**kwargs):
if include_text:<EOL><INDENT>block.text = bytes_cast(block.text, **kwargs)<EOL><DEDENT>if include_link_tokens:<EOL><INDENT>block.link_tokens = bytes_list_cast(block.link_tokens, **kwargs)<EOL><DEDENT>if include_css:<EOL><INDENT>block.css = bytes_dict_cast(block.css, **kwargs)<EOL><DEDENT>if include_features:<EOL><INDENT>block.features = bytes_dict_cast(block.features, **kwargs)<EOL><DEDENT>return block<EOL>
Converts any string-like items in input Block object to bytes-like values, with respect to python version Parameters ---------- block : blocks.Block any string-like objects contained in the block object will be converted to bytes include_text : bool, default=True if True, cast text to bytes, else ignore include_link_tokens : bool, default=True if True, cast link_tokens to bytes, else ignore include_css : bool, default=True if True, cast css to bytes, else ignore include_features : bool, default=True if True, cast features to bytes, else ignore kwargs: encoding: str, default: 'utf-8' encoding to be used when encoding string
f5163:m7
def str_block_list_cast(blocks, **kwargs):
return [str_block_cast(block, **kwargs) for block in blocks]<EOL>
Converts any bytes-like items in input lxml.Blocks to string-like values, with respect to python version Parameters ---------- blocks : list[lxml.Block] any bytes-like objects contained in the block object will be converted to a string kwargs: include_text : bool, default=True if True, cast text to a string, else ignore include_link_tokens : bool, default=True if True, cast link_tokens to a string, else ignore include_css : bool, default=True if True, cast css to a string, else ignore include_features : bool, default=True if True, cast features to a string, else ignore encoding: str, default: 'utf-8' encoding to be used when decoding bytes
f5163:m8
def bytes_block_list_cast(blocks, **kwargs):
return [bytes_block_cast(block, **kwargs) for block in blocks]<EOL>
Converts any string-like items in input lxml.Blocks to bytes-like values, with respect to python version Parameters ---------- blocks : list[lxml.Block] any string-like objects contained in the block object will be converted to bytes kwargs: include_text : bool, default=True if True, cast text to bytes, else ignore include_link_tokens : bool, default=True if True, cast link_tokens to bytes, else ignore include_css : bool, default=True if True, cast css to bytes, else ignore include_features : bool, default=True if True, cast features to bytes, else ignore encoding: str, default: 'utf-8' encoding to be used when decoding bytes
f5163:m9
def evaluate_model_predictions(y_true, y_pred, weights=None):
if isinstance(y_pred[<NUM_LIT:0>], np.ndarray):<EOL><INDENT>y_pred = np.concatenate(y_pred)<EOL><DEDENT>if isinstance(y_true[<NUM_LIT:0>], np.ndarray):<EOL><INDENT>y_true = np.concatenate(y_true)<EOL><DEDENT>if (weights is not None) and (isinstance(weights[<NUM_LIT:0>], np.ndarray)):<EOL><INDENT>weights = np.concatenate(weights)<EOL><DEDENT>accuracy = accuracy_score(<EOL>y_true, y_pred, normalize=True, sample_weight=weights)<EOL>precision = precision_score(<EOL>y_true, y_pred, average='<STR_LIT>', pos_label=<NUM_LIT:1>, sample_weight=weights)<EOL>recall = recall_score(<EOL>y_true, y_pred, average='<STR_LIT>', pos_label=<NUM_LIT:1>, sample_weight=weights)<EOL>f1 = f1_score(<EOL>y_true, y_pred, average='<STR_LIT>', pos_label=<NUM_LIT:1>, sample_weight=weights)<EOL>return {'<STR_LIT>': accuracy, '<STR_LIT>': precision, '<STR_LIT>': recall, '<STR_LIT>': f1}<EOL>
Evaluate the performance of an extractor model's binary classification predictions, typically at the block level, of whether a block is content or not. Args: y_true (``np.ndarray``) y_pred (``np.ndarray``) weights (``np.ndarray``) Returns: Dict[str, float]
f5164:m0
def evaluate_extracted_tokens(gold_content, extr_content):
if isinstance(gold_content, string_):<EOL><INDENT>gold_content = simple_tokenizer(gold_content)<EOL><DEDENT>if isinstance(extr_content, string_):<EOL><INDENT>extr_content = simple_tokenizer(extr_content)<EOL><DEDENT>gold_set = set(gold_content)<EOL>extr_set = set(extr_content)<EOL>jaccard = len(gold_set & extr_set) / len(gold_set | extr_set)<EOL>levenshtein = dameraulevenshtein(gold_content, extr_content)<EOL>return {'<STR_LIT>': jaccard, '<STR_LIT>': levenshtein}<EOL>
Evaluate the similarity between gold-standard and extracted content, typically for a single HTML document, as another way of evaluating the performance of an extractor model. Args: gold_content (str or Sequence[str]): Gold-standard content, either as a string or as an already-tokenized list of tokens. extr_content (str or Sequence[str]): Extracted content, either as a string or as an already-tokenized list of tokens. Returns: Dict[str, float]
f5164:m1
def train_model(extractor, data_dir, output_dir=None):
<EOL>output_dir, fname_prefix = _set_up_output_dir_and_fname_prefix(output_dir, extractor)<EOL>logging.info('<STR_LIT>')<EOL>data = prepare_all_data(data_dir)<EOL>training_data, test_data = train_test_split(<EOL>data, test_size=<NUM_LIT>, random_state=<NUM_LIT>)<EOL>train_html, train_labels, train_weights = extractor.get_html_labels_weights(training_data)<EOL>test_html, test_labels, test_weights = extractor.get_html_labels_weights(test_data)<EOL>logging.info('<STR_LIT>')<EOL>try:<EOL><INDENT>extractor.fit(train_html, train_labels, weights=train_weights)<EOL><DEDENT>except (TypeError, ValueError):<EOL><INDENT>extractor.fit(train_html, train_labels)<EOL><DEDENT>train_eval = evaluate_model_predictions(<EOL>np.concatenate(train_labels), extractor.predict(train_html),<EOL>np.concatenate(train_weights))<EOL>test_eval = evaluate_model_predictions(<EOL>np.concatenate(test_labels), extractor.predict(test_html),<EOL>np.concatenate(test_weights))<EOL>_report_model_performance(output_dir, fname_prefix, train_eval, test_eval)<EOL>_write_model_to_disk(output_dir, fname_prefix, extractor)<EOL>return extractor<EOL>
Train an extractor model, then write train/test block-level classification performance as well as the model itself to disk in ``output_dir``. Args: extractor (:class:`Extractor`): Instance of the ``Extractor`` class to be trained. data_dir (str): Directory on disk containing subdirectories for all training data, including raw html and gold standard blocks files output_dir (str): Directory on disk to which the trained model files, errors, etc. are to be written. If None, outputs are not saved. Returns: :class:`Extractor`: A trained extractor model.
f5164:m2
def train_many_models(extractor, param_grid, data_dir, output_dir=None,<EOL>**kwargs):
<EOL>output_dir, fname_prefix = _set_up_output_dir_and_fname_prefix(output_dir, extractor)<EOL>logging.info('<STR_LIT>')<EOL>data = prepare_all_data(data_dir)<EOL>training_data, test_data = train_test_split(<EOL>data, test_size=<NUM_LIT>, random_state=<NUM_LIT>)<EOL>train_html, train_labels, train_weights = extractor.get_html_labels_weights(training_data)<EOL>test_html, test_labels, test_weights = extractor.get_html_labels_weights(test_data)<EOL>train_blocks = np.array([extractor.blockifier.blockify(doc)<EOL>for doc in train_html])<EOL>train_mask = [extractor._has_enough_blocks(blocks) for blocks in train_blocks]<EOL>train_blocks = train_blocks[train_mask]<EOL>train_labels = np.concatenate(train_labels[train_mask])<EOL>train_weights = np.concatenate(train_weights[train_mask])<EOL>test_labels = np.concatenate(test_labels)<EOL>test_weights = np.concatenate(test_weights)<EOL>train_features = np.concatenate([extractor.features.fit_transform(blocks)<EOL>for blocks in train_blocks])<EOL>gscv = GridSearchCV(<EOL>extractor.model, param_grid, fit_params={'<STR_LIT>': train_weights},<EOL>scoring=kwargs.get('<STR_LIT>', '<STR_LIT>'), cv=kwargs.get('<STR_LIT>', <NUM_LIT:5>),<EOL>n_jobs=kwargs.get('<STR_LIT>', <NUM_LIT:1>), verbose=kwargs.get('<STR_LIT>', <NUM_LIT:1>))<EOL>gscv = gscv.fit(train_features, train_labels)<EOL>logging.info('<STR_LIT>', gscv.best_score_)<EOL>logging.info('<STR_LIT>', gscv.best_params_)<EOL>extractor.model = gscv.best_estimator_<EOL>train_eval = evaluate_model_predictions(<EOL>train_labels, extractor.predict(train_html[train_mask]), weights=train_weights)<EOL>test_eval = evaluate_model_predictions(<EOL>test_labels, extractor.predict(test_html), weights=test_weights)<EOL>_write_model_to_disk(output_dir, fname_prefix, extractor)<EOL>return extractor<EOL>
Train many extractor models, then for the best-scoring model, write train/test block-level classification performance as well as the model itself to disk in ``output_dir``. Args: extractor (:class:`Extractor`): Instance of the ``Extractor`` class to be trained. param_grid (dict or List[dict]): Dictionary with parameters names (str) as keys and lists of parameter settings to try as values, or a list of such dictionaries, in which case the grids spanned by each are explored. See documentation for :class:`GridSearchCV` for details. data_dir (str): Directory on disk containing subdirectories for all training data, including raw html and gold standard blocks files output_dir (str): Directory on disk to which the trained model files, errors, etc. are to be written. If None, outputs are not saved. **kwargs: scoring (str or Callable): default 'f1' cv (int): default 5 n_jobs (int): default 1 verbose (int): default 1 Returns: :class:`Extractor`: The trained extractor model with the best-scoring set of params. See Also: Documentation for grid search :class:`GridSearchCV` in ``scikit-learn``: http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html
f5164:m3
def extract_all_gold_standard_data(data_dir, nprocesses=<NUM_LIT:1>,<EOL>overwrite=False, **kwargs):
use_pool = nprocesses > <NUM_LIT:1><EOL>if use_pool:<EOL><INDENT>pool = multiprocessing.Pool(processes=nprocesses)<EOL><DEDENT>if overwrite is False:<EOL><INDENT>gs_blocks_dir = os.path.join(data_dir, GOLD_STANDARD_BLOCKS_DIRNAME)<EOL>if not os.path.isdir(gs_blocks_dir):<EOL><INDENT>os.mkdir(gs_blocks_dir)<EOL><DEDENT>gs_blocks_filenames = get_filenames(<EOL>gs_blocks_dir, full_path=False, match_regex=re.escape(GOLD_STANDARD_BLOCKS_EXT))<EOL>gs_blocks_fileroots = {<EOL>re.search(r'<STR_LIT>' + re.escape(GOLD_STANDARD_BLOCKS_EXT), gs_blocks_filename).group(<NUM_LIT:1>)<EOL>for gs_blocks_filename in gs_blocks_filenames}<EOL><DEDENT>else:<EOL><INDENT>gs_blocks_fileroots = set()<EOL><DEDENT>gs_dir = os.path.join(data_dir, GOLD_STANDARD_DIRNAME)<EOL>gs_filenames = get_filenames(<EOL>gs_dir, full_path=False, match_regex=re.escape(GOLD_STANDARD_EXT))<EOL>for i, gs_filename in enumerate(gs_filenames):<EOL><INDENT>gs_fileroot = re.search(r'<STR_LIT>' + re.escape(GOLD_STANDARD_EXT), gs_filename).group(<NUM_LIT:1>)<EOL>if gs_fileroot in gs_blocks_fileroots:<EOL><INDENT>continue<EOL><DEDENT>if i % <NUM_LIT:100> == <NUM_LIT:0>:<EOL><INDENT>print('<STR_LIT>'.format(gs_filename))<EOL><DEDENT>if use_pool:<EOL><INDENT>pool.apply_async(extract_gold_standard_blocks, (data_dir, gs_fileroot), kwargs)<EOL><DEDENT>else:<EOL><INDENT>extract_gold_standard_blocks(data_dir, gs_fileroot, **kwargs)<EOL><DEDENT><DEDENT>if use_pool:<EOL><INDENT>pool.close()<EOL>pool.join()<EOL><DEDENT>
Extract the gold standard block-level content and comment percentages from a directory of labeled data (only those for which the gold standard blocks are not found), and save results to corresponding files in a block-level gold standard directory under ``data_dir``. Args: data_dir (str): Directory on disk containing subdirectories for all training data, including raw html files and gold standard content + comments text files nprocesses (int): If > 1, use a :class:`multiprocessing.Pool` to parallelize the extractions overwrite (bool): If True, overwrite existing gold-standard blocks files. **kwargs: passed into :func:`extract_gold_standard_blocks` See Also: :func:`extract_gold_standard_blocks`
f5165:m0
def extract_gold_standard_blocks(data_dir, fileroot, encoding=None,<EOL>tokenizer=simple_tokenizer, cetr=False):
<EOL>raw_html = read_html_file(data_dir, fileroot, encoding=encoding) <EOL>from dragnet.blocks import BlockifyError<EOL>try:<EOL><INDENT>blocks = [b.text for b in Blockifier.blockify(raw_html)] <EOL><DEDENT>except BlockifyError as e:<EOL><INDENT>print('<STR_LIT>'.format(fileroot))<EOL>return<EOL><DEDENT>blocks_tokens = [tokenizer(block) for block in blocks]<EOL>num_blocks_tokens = [len(block_tokens) for block_tokens in blocks_tokens]<EOL>all_blocks_tokens = []<EOL>all_blocks_tokens_block_id = []<EOL>for i, block_tokens in enumerate(blocks_tokens):<EOL><INDENT>all_blocks_tokens.extend(block_tokens)<EOL>all_blocks_tokens_block_id.extend([i] * len(block_tokens))<EOL><DEDENT>def get_frac_and_str_tokens_in_gs(gs_txt):<EOL><INDENT>"""<STR_LIT>"""<EOL>gs_tokens = tokenizer(gs_txt)<EOL>tokens_in_gs = check_inclusion(all_blocks_tokens, gs_tokens)<EOL>num_blocks_tokens_in_gs = [<NUM_LIT:0> for _ in range(len(blocks))]<EOL>blocks_tokens_in_gs_tokens = [[] for _ in range(len(blocks))]<EOL>for token, token_in_gs, block_id in zip(all_blocks_tokens, tokens_in_gs, all_blocks_tokens_block_id):<EOL><INDENT>if token_in_gs is True:<EOL><INDENT>num_blocks_tokens_in_gs[block_id] += <NUM_LIT:1><EOL>blocks_tokens_in_gs_tokens[block_id].append(token)<EOL><DEDENT><DEDENT>blocks_tokens_strs_in_gs = [<EOL>'<STR_LIT:U+0020>'.join(block_tokens_in_gs_tokens)<EOL>for block_tokens_in_gs_tokens in blocks_tokens_in_gs_tokens]<EOL>frac_blocks_tokens_in_gs = [<EOL>num_block_tokens_in_gs / num_block_tokens<EOL>for num_block_tokens_in_gs, num_block_tokens<EOL>in zip(num_blocks_tokens_in_gs, num_blocks_tokens)]<EOL>return (frac_blocks_tokens_in_gs, blocks_tokens_strs_in_gs)<EOL><DEDENT>gs_content, gs_comments = read_gold_standard_file(data_dir, fileroot, cetr)<EOL>frac_blocks_tokens_in_gs_content, blocks_tokens_strs_in_gs_content =get_frac_and_str_tokens_in_gs(gs_content)<EOL>frac_blocks_tokens_in_gs_comments, blocks_tokens_strs_in_gs_comments =get_frac_and_str_tokens_in_gs(gs_comments)<EOL>output_fname = os.path.join(<EOL>data_dir, GOLD_STANDARD_BLOCKS_DIRNAME, fileroot + GOLD_STANDARD_BLOCKS_EXT)<EOL>line_fmt = u'<STR_LIT>'<EOL>with io.open(output_fname, mode='<STR_LIT:w>') as f:<EOL><INDENT>for block_id, block_tokens in enumerate(blocks_tokens):<EOL><INDENT>line = line_fmt.format(<EOL>frac_content=frac_blocks_tokens_in_gs_content[block_id],<EOL>frac_comments=frac_blocks_tokens_in_gs_comments[block_id],<EOL>block_tokens='<STR_LIT:U+0020>'.join(block_tokens),<EOL>content_tokens=blocks_tokens_strs_in_gs_content[block_id],<EOL>comment_tokens=blocks_tokens_strs_in_gs_comments[block_id])<EOL>f.write(line)<EOL><DEDENT><DEDENT>
Extract the gold standard block-level content and comments for a single observation identified by ``fileroot``, and write the results to file. Args: data_dir (str): The root directory containing sub-directories for raw HTML, gold standard extracted content, and gold standard blocks. fileroot (str): Unique identifier for a single observation of training data, corresponding to the start of its raw html and gold standard filenames under ``data_dir``. encoding (str) tokenizer (Callable): Object that takes a string and returns the tokens as a list of strings. cetr (bool): If True, parse the gold standard in clean eval format. Notes: Results are written to a text file in the block-level gold standard dir :obj:`GOLD_STANDARD_BLOCKS_DIRNAME` below ``data_dir``. Each line corresponds to a single block in its order of appearance, and has the following format:: content_frac comments_frac all_tokens content_tokens comments_tokens where each item is separated by a tab. ``content_frac`` is equal to the fraction of ``all_tokens`` found in the corresponding gold parse content text; ``comments_frac`` is the same but for comments text.
f5165:m1
def get_filenames(dirname, full_path=False, match_regex=None, extension=None):
if not os.path.exists(dirname):<EOL><INDENT>raise OSError('<STR_LIT>'.format(dirname))<EOL><DEDENT>match_regex = re.compile(match_regex) if match_regex else None<EOL>for filename in sorted(os.listdir(dirname)):<EOL><INDENT>if extension and not os.path.splitext(filename)[-<NUM_LIT:1>] == extension:<EOL><INDENT>continue<EOL><DEDENT>if match_regex and not match_regex.search(filename):<EOL><INDENT>continue<EOL><DEDENT>if full_path is True:<EOL><INDENT>yield os.path.join(dirname, filename)<EOL><DEDENT>else:<EOL><INDENT>yield filename<EOL><DEDENT><DEDENT>
Get all filenames under ``dirname`` that match ``match_regex`` or have file extension equal to ``extension``, optionally prepending the full path. Args: dirname (str): /path/to/dir on disk where files to read are saved full_path (bool): if False, return filenames without path; if True, return filenames with path, as ``os.path.join(dirname, fname)`` match_regex (str): include files whose names match this regex pattern extension (str): if files only of a certain type are wanted, specify the file extension (e.g. ".txt") Yields: str: next matching filename
f5165:m2
def read_html_file(data_dir, fileroot, encoding=None):
fname = os.path.join(<EOL>data_dir, RAW_HTML_DIRNAME, fileroot + RAW_HTML_EXT)<EOL>encodings = (encoding,) if encoding else ('<STR_LIT:utf-8>', '<STR_LIT>') <EOL>for encoding in encodings:<EOL><INDENT>try:<EOL><INDENT>with io.open(fname, mode='<STR_LIT>', encoding=encoding) as f:<EOL><INDENT>raw_html = f.read()<EOL><DEDENT>break<EOL><DEDENT>except (UnicodeDecodeError, UnicodeError):<EOL><INDENT>raw_html = None<EOL><DEDENT><DEDENT>return ftfy.fix_encoding(raw_html).strip()<EOL>
Read the HTML file corresponding to identifier ``fileroot`` in the raw HTML directory below the root ``data_dir``. Args: data_dir (str) fileroot (str) encoding (str) Returns: str
f5165:m3
def read_gold_standard_file(data_dir, fileroot, encoding=None, cetr=False):
fname = os.path.join(<EOL>data_dir, GOLD_STANDARD_DIRNAME, fileroot + GOLD_STANDARD_EXT)<EOL>encodings = (encoding,) if encoding else ('<STR_LIT:utf-8>', '<STR_LIT>', '<STR_LIT>')<EOL>for encoding in encodings:<EOL><INDENT>try:<EOL><INDENT>with io.open(fname, mode='<STR_LIT>', encoding=encoding) as f:<EOL><INDENT>gold_standard = f.read()<EOL><DEDENT>break<EOL><DEDENT>except (UnicodeDecodeError, UnicodeError):<EOL><INDENT>gold_standard = None<EOL><DEDENT><DEDENT>if not gold_standard:<EOL><INDENT>return [u'<STR_LIT>', u'<STR_LIT>']<EOL><DEDENT>if not cetr:<EOL><INDENT>content_comments = RE_COMMENTS_DELIM.split(gold_standard, maxsplit=<NUM_LIT:1>)<EOL>if len(content_comments) == <NUM_LIT:1>:<EOL><INDENT>content_comments = [content_comments[<NUM_LIT:0>], u'<STR_LIT>']<EOL><DEDENT><DEDENT>else:<EOL><INDENT>tree = etree.fromstring(gold_standard, parser=etree.HTMLParser())<EOL>content_comments = [u'<STR_LIT:U+0020>'.join(text_from_subtree(tree)), u'<STR_LIT>']<EOL><DEDENT>content_comments = [ftfy.fix_encoding(content_comments[<NUM_LIT:0>]).strip(),<EOL>ftfy.fix_encoding(content_comments[<NUM_LIT:1>]).strip()]<EOL>return content_comments<EOL>
Read the gold standard content file corresponding to identifier ``fileroot`` in the gold standard directory below the root ``data_dir``. Args: data_dir (str) fileroot (str) encoding (str) cetr (bool): if True, assume no comments and parse the gold standard to remove tags Returns: List[str, str]: contents string and comments string, respectively
f5165:m4
def read_gold_standard_blocks_file(data_dir, fileroot, split_blocks=True):
fname = os.path.join(<EOL>data_dir, GOLD_STANDARD_BLOCKS_DIRNAME, fileroot + GOLD_STANDARD_BLOCKS_EXT)<EOL>with io.open(fname, mode='<STR_LIT:r>') as f:<EOL><INDENT>data = f.read()<EOL><DEDENT>if split_blocks:<EOL><INDENT>return filter(None, data[:-<NUM_LIT:1>].split('<STR_LIT:\n>'))<EOL><DEDENT>return filter(None, data)<EOL>
Read the gold standard blocks file corresponding to identifier ``fileroot`` in the gold standard blocks directory below the root ``data_dir``. Args: data_dir (str) fileroot (str) split_blocks (bool): If True, split the file's content into blocks. Returns: str or List[str]
f5165:m5
def prepare_data(data_dir, fileroot, block_pct_tokens_thresh=<NUM_LIT:0.1>):
if not <NUM_LIT:0.0> <= block_pct_tokens_thresh <= <NUM_LIT:1.0>:<EOL><INDENT>raise ValueError('<STR_LIT>')<EOL><DEDENT>html = read_html_file(data_dir, fileroot)<EOL>blocks = read_gold_standard_blocks_file(data_dir, fileroot, split_blocks=True)<EOL>content_blocks = []<EOL>comments_blocks = []<EOL>for block in blocks:<EOL><INDENT>block_split = block.split('<STR_LIT:\t>')<EOL>num_block_tokens = len(block_split[<NUM_LIT:2>].split())<EOL>content_blocks.append(<EOL>(float(block_split[<NUM_LIT:0>]), num_block_tokens, block_split[<NUM_LIT:3>].split()))<EOL>comments_blocks.append(<EOL>(float(block_split[<NUM_LIT:1>]), num_block_tokens, block_split[<NUM_LIT:4>].split()))<EOL><DEDENT>parsed_content_blocks = _parse_content_or_comments_blocks(<EOL>content_blocks, block_pct_tokens_thresh)<EOL>parsed_comments_blocks = _parse_content_or_comments_blocks(<EOL>comments_blocks, block_pct_tokens_thresh)<EOL>return (html, parsed_content_blocks, parsed_comments_blocks)<EOL>
Prepare data for a single HTML + gold standard blocks example, uniquely identified by ``fileroot``. Args: data_dir (str) fileroot (str) block_pct_tokens_thresh (float): must be in [0.0, 1.0] Returns: Tuple[str, Tuple[np.array[int], np.array[int], List[str]], Tuple[np.array[int], np.array[int], List[str]]]: The first element is simply the raw html as a string. The second and third elements are 3-tuples for content and comments, respectively, where the first element is a numpy array of 1s and 0s whose values correspond to whether or not a given block is considered non-content or not; the second element is a numpy integer array whose values are the total number of tokens in each block; and the third element is a flat list of content or comment tokens as strings, concatenated from all blocks. See Also: :func:`prepare_all_data`
f5165:m7
def prepare_all_data(data_dir, block_pct_tokens_thresh=<NUM_LIT:0.1>):
gs_blocks_dir = os.path.join(data_dir, GOLD_STANDARD_BLOCKS_DIRNAME)<EOL>gs_blocks_filenames = get_filenames(<EOL>gs_blocks_dir, full_path=False, match_regex=re.escape(GOLD_STANDARD_BLOCKS_EXT))<EOL>gs_blocks_fileroots = (<EOL>re.search(r'<STR_LIT>' + re.escape(GOLD_STANDARD_BLOCKS_EXT), gs_blocks_filename).group(<NUM_LIT:1>)<EOL>for gs_blocks_filename in gs_blocks_filenames)<EOL>return [prepare_data(data_dir, fileroot, block_pct_tokens_thresh)<EOL>for fileroot in gs_blocks_fileroots]<EOL>
Prepare data for all HTML + gold standard blocks examples in ``data_dir``. Args: data_dir (str) block_pct_tokens_thresh (float): must be in [0.0, 1.0] Returns: List[Tuple[str, List[float, int, List[str]], List[float, int, List[str]]]] See Also: :func:`prepare_data`
f5165:m8
def fit(self, documents, labels, weights=None):
block_groups = np.array([self.blockifier.blockify(doc) for doc in documents])<EOL>mask = [self._has_enough_blocks(blocks) for blocks in block_groups]<EOL>block_groups = block_groups[mask]<EOL>labels = np.concatenate(np.array(labels)[mask])<EOL>features_mat = np.concatenate([self.features.fit_transform(blocks)<EOL>for blocks in block_groups])<EOL>if weights is None:<EOL><INDENT>self.model.fit(features_mat, labels)<EOL><DEDENT>else:<EOL><INDENT>weights = np.concatenate(np.array(weights)[mask])<EOL>self.model.fit(features_mat, labels, sample_weight=weights)<EOL><DEDENT>return self<EOL>
Fit :class`Extractor` features and model to a training dataset. Args: blocks (List[Block]) labels (``np.ndarray``) weights (``np.ndarray``) Returns: :class`Extractor`
f5166:c0:m3
def get_html_labels_weights(self, data):
all_html = []<EOL>all_labels = []<EOL>all_weights = []<EOL>for html, content, comments in data:<EOL><INDENT>all_html.append(html)<EOL>labels, weights = self._get_labels_and_weights(<EOL>content, comments)<EOL>all_labels.append(labels)<EOL>all_weights.append(weights)<EOL><DEDENT>return np.array(all_html), np.array(all_labels), np.array(all_weights)<EOL>
Gather the html, labels, and weights of many files' data. Primarily useful for training/testing an :class`Extractor`. Args: data: Output of :func:`dragnet.data_processing.prepare_all_data`. Returns: Tuple[List[Block], np.array(int), np.array(int)]: All blocks, all labels, and all weights, respectively.
f5166:c0:m4
def _get_labels_and_weights(self, content, comments):
<EOL>if '<STR_LIT:content>' in self.to_extract and '<STR_LIT>' in self.to_extract:<EOL><INDENT>labels = np.logical_or(content[<NUM_LIT:0>], comments[<NUM_LIT:0>]).astype(int)<EOL>weights = content[<NUM_LIT:1>],<EOL><DEDENT>elif '<STR_LIT:content>' in self.to_extract:<EOL><INDENT>labels = content[<NUM_LIT:0>]<EOL>weights = content[<NUM_LIT:1>]<EOL><DEDENT>else:<EOL><INDENT>labels = comments[<NUM_LIT:0>]<EOL>weights = comments[<NUM_LIT:1>]<EOL><DEDENT>if self.max_block_weight is None:<EOL><INDENT>weights = np.minimum(weights, self.max_block_weight)<EOL><DEDENT>return labels, weights<EOL>
Args: content (Tuple[np.array[int], np.array[int], List[str]]) comments (Tuple[np.array[int], np.array[int], List[str]]) Returns: Tuple[np.array[int], np.array[int], List[str]]
f5166:c0:m6
def extract(self, html, encoding=None, as_blocks=False):
preds, blocks = self.predict(html, encoding=encoding, return_blocks=True)<EOL>if as_blocks is False:<EOL><INDENT>return str_cast(b'<STR_LIT:\n>'.join(blocks[ind].text for ind in np.flatnonzero(preds)))<EOL><DEDENT>else:<EOL><INDENT>return [blocks[ind] for ind in np.flatnonzero(preds)]<EOL><DEDENT>
Extract the main content and/or comments from an HTML document and return it as a string or as a sequence of block objects. Args: html (str): HTML document as a string. encoding (str): Encoding of ``html``. If None (encoding unknown), the original encoding will be guessed from the HTML itself. as_blocks (bool): If False, return the main content as a combined string; if True, return the content-holding blocks as a list of block objects. Returns: str or List[Block]
f5166:c0:m7
def predict(self, documents, **kwargs):
if isinstance(documents, (str, bytes, unicode_, np.unicode_)):<EOL><INDENT>return self._predict_one(documents, **kwargs)<EOL><DEDENT>else:<EOL><INDENT>return np.concatenate([self._predict_one(doc, **kwargs) for doc in documents])<EOL><DEDENT>
Predict class (content=1 or not-content=0) of the blocks in one or many HTML document(s). Args: documents (str or List[str]): HTML document(s) Returns: ``np.ndarray`` or List[``np.ndarray``]: array of binary predictions for content (1) or not-content (0).
f5166:c0:m8
def _predict_one(self, document, encoding=None, return_blocks=False):
<EOL>blocks = self.blockifier.blockify(document, encoding=encoding)<EOL>try:<EOL><INDENT>features = self.features.transform(blocks)<EOL><DEDENT>except ValueError: <EOL><INDENT>preds = np.zeros((len(blocks)))<EOL><DEDENT>else:<EOL><INDENT>if self.prob_threshold is None:<EOL><INDENT>preds = self.model.predict(features)<EOL><DEDENT>else:<EOL><INDENT>self._positive_idx = (<EOL>self._positive_idx or list(self.model.classes_).index(<NUM_LIT:1>))<EOL>preds = self.model.predict_proba(features) > self.prob_threshold<EOL>preds = preds[:, self._positive_idx].astype(int)<EOL><DEDENT><DEDENT>if return_blocks:<EOL><INDENT>return preds, blocks<EOL><DEDENT>else:<EOL><INDENT>return preds<EOL><DEDENT>
Predict class (content=1 or not-content=0) of each block in an HTML document. Args: documents (str): HTML document Returns: ``np.ndarray``: array of binary predictions for content (1) or not-content (0).
f5166:c0:m9
def fit(self, blocks, y=None):
feature_array = self.feature.fit_transform(blocks)<EOL>self.scaler = self.scaler.fit(feature_array)<EOL>return self<EOL>
Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: :class:`StandardizedFeature`: an instance of this class with the ``self.scaler`` attribute fit to the ``blocks`` data Note: When fitting the :class:`StandardScaler` object, you'll probably want to determine the mean and/or std of *multiple* HTML files' blocks, rather than just a single observation. To do that, just concatenate all of the blocks together in a single iterable. In contrast, you'll typically apply :meth:`transform` to a *single* HTML file's blocks at a time.
f5168:c0:m1
def transform(self, blocks, y=None):
return self.scaler.transform(self.feature.transform(blocks))<EOL>
Transform an ordered sequence of blocks into a 2D features matrix with shape (num blocks, num features) and standardized feature values. Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: `np.ndarray`: 2D array of shape (num blocks, num sub-features), where ``blocks`` data has been transformed by ``self.feature`` and optionally standardized by ``self.scaler``.
f5168:c0:m2
def fit(self, blocks, y=None):
return self<EOL>
This method returns the current instance unchanged, since no fitting is required for this ``Feature``. It's here only for API consistency.
f5169:c0:m0
def transform(self, blocks, y=None):
feature_vecs = (<EOL>tuple(re.search(token, block.css[attrib]) is not None<EOL>for block in blocks)<EOL>for attrib, tokens in self.attribute_tokens<EOL>for token in tokens<EOL>)<EOL>return np.column_stack(tuple(feature_vecs)).astype(int)<EOL>
Transform an ordered sequence of blocks into a 2D features matrix with shape (num blocks, num features). Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: `np.ndarray`: 2D array of shape (num blocks, num CSS attributes), where values are either 0 or 1, indicating the absence or presence of a given token in a CSS attribute on a given block.
f5169:c0:m1
def fit(self, blocks, y=None):
return self<EOL>
This method returns the current instance unchanged, since no fitting is required for this ``Feature``. It's here only for API consistency.
f5170:c0:m0
def transform(self, blocks, y=None):
return make_kohlschuetter_features(blocks)<EOL>
Transform an ordered sequence of blocks into a 2D features matrix with shape (num blocks, num features). Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: `np.ndarray`: 2D array of shape (num blocks, 6), where values are floats corresponding to the link and text densities of a block and its immediate neighbors in the sequence.
f5170:c0:m1
def get_feature(name):
if name == '<STR_LIT>':<EOL><INDENT>return CSSFeatures()<EOL><DEDENT>elif name == '<STR_LIT>':<EOL><INDENT>return KohlschuetterFeatures()<EOL><DEDENT>elif name == '<STR_LIT>':<EOL><INDENT>return ReadabilityFeatures()<EOL><DEDENT>elif name == '<STR_LIT>':<EOL><INDENT>return WeningerFeatures()<EOL><DEDENT>elif name == '<STR_LIT>':<EOL><INDENT>return ClusteredWeningerFeatures()<EOL><DEDENT>else:<EOL><INDENT>raise ValueError('<STR_LIT>'.format(name))<EOL><DEDENT>
Get an instance of a ``Features`` class by ``name`` (str).
f5171:m0
def fit(self, blocks, y=None):
return self<EOL>
This method returns the current instance unchanged, since no fitting is required for this ``Feature``. It's here only for API consistency.
f5172:c0:m0
def transform(self, blocks, y=None):
return make_readability_features(blocks)<EOL>
Transform an ordered sequence of blocks into a 2D features matrix with shape (num blocks, num features). Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: `np.ndarray`: 2D array of shape (num blocks, 1)
f5172:c0:m1
def fit(self, blocks, y=None):
return self<EOL>
This method returns the current instance unchanged, since no fitting is required for this ``Feature``. It's here only for API consistency.
f5173:c0:m1
def transform(self, blocks, y=None):
return make_weninger_features(blocks, sigma=self.sigma)<EOL>
Computes the content to tag ratio per block and returns the smoothed values and the smoothed absolute differences for each block. Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: :class:`np.ndarray`: 2D array of shape (len(x), 2), where values are floats corresponding to the smoothed and smoothed absolute difference values for each block.
f5173:c0:m2
def fit(self, blocks, y=None):
self.kmeans.fit(make_weninger_features(blocks))<EOL>self.kmeans.cluster_centers_.sort(axis=<NUM_LIT:0>)<EOL>self.kmeans.cluster_centers_[<NUM_LIT:0>, :] = np.zeros(<NUM_LIT:2>)<EOL>return self<EOL>
Fit a k-means clustering model using an ordered sequence of blocks.
f5173:c1:m1
def transform(self, blocks, y=None):
preds = (self.kmeans.predict(make_weninger_features(blocks)) > <NUM_LIT:0>).astype(int)<EOL>return np.reshape(preds, (-<NUM_LIT:1>, <NUM_LIT:1>))<EOL>
Computes the content to tag ratio per block, smooths the values, then predicts content (1) or not-content (0) using a fit k-means cluster model. Args: blocks (List[Block]): as output by :class:`Blockifier.blockify` y (None): This isn't used, it's only here for API consistency. Returns: :class:`np.ndarray`: 2D array of shape (len(feature_mat), 1), where values are either 0 or 1, corresponding to the kmeans prediction of content (1) or not-content (0).
f5173:c1:m2
def block_output_tokens(blocks, true_tokens):
assert len(blocks) == len(true_tokens)<EOL>for k in range_(len(blocks)):<EOL><INDENT>block_tokens = re.split(r"<STR_LIT>", blocks[k].text.strip())<EOL>assert block_tokens == true_tokens[k]<EOL><DEDENT>
blocks = the output from blockify true_tokens = a list of true tokens
f5178:m1
@property<EOL><INDENT>def sources(<EOL>self):<DEDENT>
sourceResultsList = []<EOL>sourceResultsList[:] = [dict(l) for l in self.sourceResultsList]<EOL>return sourceResultsList<EOL>
*The results of the search returned as a python list of dictionaries* **Usage:** .. code-block:: python sources = tns.sources
f5187:c0:m1
@property<EOL><INDENT>def spectra(<EOL>self):<DEDENT>
specResultsList = []<EOL>specResultsList[:] = [dict(l) for l in self.specResultsList]<EOL>return specResultsList<EOL>
*The associated source spectral data* **Usage:** .. code-block:: python sourceSpectra = tns.spectra
f5187:c0:m2
@property<EOL><INDENT>def files(<EOL>self):<DEDENT>
relatedFilesResultsList = []<EOL>relatedFilesResultsList[:] = [dict(l)<EOL>for l in self.relatedFilesResultsList]<EOL>return relatedFilesResultsList<EOL>
*The associated source files* **Usage:** .. code-block:: python sourceFiles = tns.files
f5187:c0:m3
@property<EOL><INDENT>def photometry(<EOL>self):<DEDENT>
photResultsList = []<EOL>photResultsList[:] = [dict(l) for l in self.photResultsList]<EOL>return photResultsList<EOL>
*The associated source photometry* **Usage:** .. code-block:: python sourcePhotometry = tns.photometry
f5187:c0:m4
@property<EOL><INDENT>def url(<EOL>self):<DEDENT>
return self._searchURL<EOL>
*The generated URL used for searching of the TNS* **Usage:** .. code-block:: python searchURL = tns.url
f5187:c0:m5
def csv(<EOL>self,<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>csvSources = self.sourceResults.csv(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>csvPhot = self.photResults.csv(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>csvSpec = self.specResults.csv(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>csvFiles = self.relatedFilesResults.csv(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>csvSources = self.sourceResults.csv()<EOL>csvPhot = self.photResults.csv()<EOL>csvSpec = self.specResults.csv()<EOL>csvFiles = self.relatedFilesResults.csv()<EOL><DEDENT>return csvSources, csvPhot, csvSpec, csvFiles<EOL>
*Render the results in csv format* **Key Arguments:** - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `csvSources` -- the top-level transient data - `csvPhot` -- all photometry associated with the transients - `csvSpec` -- all spectral data associated with the transients - `csvFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in csv format: .. code-block:: python csvSources, csvPhot, csvSpec, csvFiles = tns.csv() print csvSources .. code-block:: text TNSId,TNSName,discoveryName,discSurvey,raSex,decSex,raDeg,decDeg,transRedshift,specType,discMag,discMagFilter,discDate,objectUrl,hostName,hostRedshift,separationArcsec,separationNorthArcsec,separationEastArcsec 2016asf,SN2016asf,ASASSN-16cs,ASAS-SN,06:50:36.73,+31:06:45.36,102.6530,31.1126,0.021,SN Ia,17.1,V-Johnson,2016-03-06 08:09:36,http://wis-tns.weizmann.ac.il/object/2016asf,KUG 0647+311,,0.66,0.65,-0.13 You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.csv("~/tns") .. image:: https://i.imgur.com/BwwqMBg.png :width: 800px :alt: csv output
f5187:c0:m6
def json(<EOL>self,<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>jsonSources = self.sourceResults.json(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>jsonPhot = self.photResults.json(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>jsonSpec = self.specResults.json(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>jsonFiles = self.relatedFilesResults.json(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>jsonSources = self.sourceResults.json()<EOL>jsonPhot = self.photResults.json()<EOL>jsonSpec = self.specResults.json()<EOL>jsonFiles = self.relatedFilesResults.json()<EOL><DEDENT>return jsonSources, jsonPhot, jsonSpec, jsonFiles<EOL>
*Render the results in json format* **Key Arguments:** - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `jsonSources` -- the top-level transient data - `jsonPhot` -- all photometry associated with the transients - `jsonSpec` -- all spectral data associated with the transients - `jsonFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in json format: .. code-block:: python jsonSources, jsonPhot, jsonSpec, jsonFiles = tns.json() print jsonSources .. code-block:: text [ { "TNSId": "2016asf", "TNSName": "SN2016asf", "decDeg": 31.1126, "decSex": "+31:06:45.36", "discDate": "2016-03-06 08:09:36", "discMag": "17.1", "discMagFilter": "V-Johnson", "discSurvey": "ASAS-SN", "discoveryName": "ASASSN-16cs", "hostName": "KUG 0647+311", "hostRedshift": null, "objectUrl": "http://wis-tns.weizmann.ac.il/object/2016asf", "raDeg": 102.65304166666667, "raSex": "06:50:36.73", "separationArcsec": "0.66", "separationEastArcsec": "-0.13", "separationNorthArcsec": "0.65", "specType": "SN Ia", "transRedshift": "0.021" } ] You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.json("~/tns") .. image:: https://i.imgur.com/wAHqARI.png :width: 800px :alt: json output
f5187:c0:m7
def yaml(<EOL>self,<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>yamlSources = self.sourceResults.yaml(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>yamlPhot = self.photResults.yaml(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>yamlSpec = self.specResults.yaml(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>yamlFiles = self.relatedFilesResults.yaml(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>yamlSources = self.sourceResults.yaml()<EOL>yamlPhot = self.photResults.yaml()<EOL>yamlSpec = self.specResults.yaml()<EOL>yamlFiles = self.relatedFilesResults.yaml()<EOL><DEDENT>return yamlSources, yamlPhot, yamlSpec, yamlFiles<EOL>
*Render the results in yaml format* **Key Arguments:** - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `yamlSources` -- the top-level transient data - `yamlPhot` -- all photometry associated with the transients - `yamlSpec` -- all spectral data associated with the transients - `yamlFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in yaml format: .. code-block:: python yamlSources, yamlPhot, yamlSpec, yamlFiles = tns.yaml() print yamlSources .. code-block:: text - TNSId: 2016asf TNSName: SN2016asf decDeg: 31.1126 decSex: '+31:06:45.36' discDate: '2016-03-06 08:09:36' discMag: '17.1' discMagFilter: V-Johnson discSurvey: ASAS-SN discoveryName: ASASSN-16cs hostName: KUG 0647+311 hostRedshift: null objectUrl: http://wis-tns.weizmann.ac.il/object/2016asf raDeg: 102.65304166666667 raSex: '06:50:36.73' separationArcsec: '0.66' separationEastArcsec: '-0.13' separationNorthArcsec: '0.65' specType: SN Ia transRedshift: '0.021' You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.yaml("~/tns") .. image:: https://i.imgur.com/ZpJIC6p.png :width: 800px :alt: yaml output
f5187:c0:m8
def markdown(<EOL>self,<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>markdownSources = self.sourceResults.markdown(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>markdownPhot = self.photResults.markdown(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>markdownSpec = self.specResults.markdown(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>markdownFiles = self.relatedFilesResults.markdown(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>markdownSources = self.sourceResults.markdown()<EOL>markdownPhot = self.photResults.markdown()<EOL>markdownSpec = self.specResults.markdown()<EOL>markdownFiles = self.relatedFilesResults.markdown()<EOL><DEDENT>return markdownSources, markdownPhot, markdownSpec, markdownFiles<EOL>
*Render the results in markdown format* **Key Arguments:** - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `markdownSources` -- the top-level transient data - `markdownPhot` -- all photometry associated with the transients - `markdownSpec` -- all spectral data associated with the transients - `markdownFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in markdown table format: .. code-block:: python markdownSources, markdownPhot, markdownSpec, markdownFiles = tns.markdown() print markdownSources .. code-block:: text | TNSId | TNSName | discoveryName | discSurvey | raSex | decSex | raDeg | decDeg | transRedshift | specType | discMag | discMagFilter | discDate | objectUrl | hostName | hostRedshift | separationArcsec | separationNorthArcsec | separationEastArcsec | |:---------|:-----------|:---------------|:------------|:-------------|:--------------|:----------|:---------|:---------------|:----------|:---------|:---------------|:---------------------|:----------------------------------------------|:--------------|:--------------|:------------------|:-----------------------|:----------------------| | 2016asf | SN2016asf | ASASSN-16cs | ASAS-SN | 06:50:36.73 | +31:06:45.36 | 102.6530 | 31.1126 | 0.021 | SN Ia | 17.1 | V-Johnson | 2016-03-06 08:09:36 | http://wis-tns.weizmann.ac.il/object/2016asf | KUG 0647+311 | | 0.66 | 0.65 | -0.13 | You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.markdown("~/tns") .. image:: https://i.imgur.com/AYLBQoJ.png :width: 800px :alt: markdown output
f5187:c0:m9
def table(<EOL>self,<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>tableSources = self.sourceResults.table(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>tablePhot = self.photResults.table(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>tableSpec = self.specResults.table(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL>tableFiles = self.relatedFilesResults.table(<EOL>filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>tableSources = self.sourceResults.table()<EOL>tablePhot = self.photResults.table()<EOL>tableSpec = self.specResults.table()<EOL>tableFiles = self.relatedFilesResults.table()<EOL><DEDENT>return tableSources, tablePhot, tableSpec, tableFiles<EOL>
*Render the results as an ascii table* **Key Arguments:** - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `tableSources` -- the top-level transient data - `tablePhot` -- all photometry associated with the transients - `tableSpec` -- all spectral data associated with the transients - `tableFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in ascii table format: .. code-block:: python tableSources, tablePhot, tableSpec, tableFiles = tns.table() print tableSources .. code-block:: text +----------+------------+----------------+-------------+--------------+---------------+-----------+----------+----------------+-----------+----------+----------------+----------------------+-----------------------------------------------+---------------+---------------+-------------------+------------------------+-----------------------+ | TNSId | TNSName | discoveryName | discSurvey | raSex | decSex | raDeg | decDeg | transRedshift | specType | discMag | discMagFilter | discDate | objectUrl | hostName | hostRedshift | separationArcsec | separationNorthArcsec | separationEastArcsec | +----------+------------+----------------+-------------+--------------+---------------+-----------+----------+----------------+-----------+----------+----------------+----------------------+-----------------------------------------------+---------------+---------------+-------------------+------------------------+-----------------------+ | 2016asf | SN2016asf | ASASSN-16cs | ASAS-SN | 06:50:36.73 | +31:06:45.36 | 102.6530 | 31.1126 | 0.021 | SN Ia | 17.1 | V-Johnson | 2016-03-06 08:09:36 | http://wis-tns.weizmann.ac.il/object/2016asf | KUG 0647+311 | | 0.66 | 0.65 | -0.13 | +----------+------------+----------------+-------------+--------------+---------------+-----------+----------+----------------+-----------+----------+----------------+----------------------+-----------------------------------------------+---------------+---------------+-------------------+------------------------+-----------------------+ You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.table("~/tns") .. image:: https://i.imgur.com/m09M0ho.png :width: 800px :alt: ascii files
f5187:c0:m10
def mysql(<EOL>self,<EOL>tableNamePrefix="<STR_LIT>",<EOL>dirPath=None):
if dirPath:<EOL><INDENT>p = self._file_prefix()<EOL>createStatement = """<STR_LIT>""" % locals()<EOL>mysqlSources = self.sourceResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>", filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>", createStatement=createStatement)<EOL>createStatement = """<STR_LIT>""" % locals()<EOL>mysqlPhot = self.photResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>", filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>", createStatement=createStatement)<EOL>createStatement = """<STR_LIT>""" % locals()<EOL>mysqlSpec = self.specResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>", filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>", createStatement=createStatement)<EOL>createStatement = """<STR_LIT>""" % locals()<EOL>mysqlFiles = self.relatedFilesResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>", filepath=dirPath + "<STR_LIT:/>" + p + "<STR_LIT>", createStatement=createStatement)<EOL><DEDENT>else:<EOL><INDENT>mysqlSources = self.sourceResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>")<EOL>mysqlPhot = self.photResults.mysql(tableNamePrefix + "<STR_LIT>")<EOL>mysqlSpec = self.specResults.mysql(tableNamePrefix + "<STR_LIT>")<EOL>mysqlFiles = self.relatedFilesResults.mysql(<EOL>tableNamePrefix + "<STR_LIT>")<EOL><DEDENT>return mysqlSources, mysqlPhot, mysqlSpec, mysqlFiles<EOL>
*Render the results as MySQL Insert statements* **Key Arguments:** - ``tableNamePrefix`` -- the prefix for the database table names to assign the insert statements to. Default *TNS*. - ``dirPath`` -- the path to the directory to save the rendered results to. Default *None* **Return:** - `mysqlSources` -- the top-level transient data - `mysqlPhot` -- all photometry associated with the transients - `mysqlSpec` -- all spectral data associated with the transients - `mysqlFiles` -- all files associated with the matched transients found on the tns **Usage:** To render the results in mysql insert format: .. code-block:: python mysqlSources, mysqlPhot, mysqlSpec, mysqlFiles = tns.mysql("TNS") print mysqlSources .. code-block:: text INSERT INTO `TNS_sources` (TNSId,TNSName,dateCreated,decDeg,decSex,discDate,discMag,discMagFilter,discSurvey,discoveryName,hostName,hostRedshift,objectUrl,raDeg,raSex,separationArcsec,separationEastArcsec,separationNorthArcsec,specType,transRedshift) VALUES ("2016asf" ,"SN2016asf" ,"2016-09-20T11:22:13" ,"31.1126" ,"+31:06:45.36" ,"2016-03-06 08:09:36" ,"17.1" ,"V-Johnson" ,"ASAS-SN" ,"ASASSN-16cs" ,"KUG 0647+311" ,null ,"http://wis-tns.weizmann.ac.il/object/2016asf" ,"102.653041667" ,"06:50:36.73" ,"0.66" ,"-0.13" ,"0.65" ,"SN Ia" ,"0.021") ON DUPLICATE KEY UPDATE TNSId="2016asf", TNSName="SN2016asf", dateCreated="2016-09-20T11:22:13", decDeg="31.1126", decSex="+31:06:45.36", discDate="2016-03-06 08:09:36", discMag="17.1", discMagFilter="V-Johnson", discSurvey="ASAS-SN", discoveryName="ASASSN-16cs", hostName="KUG 0647+311", hostRedshift=null, objectUrl="http://wis-tns.weizmann.ac.il/object/2016asf", raDeg="102.653041667", raSex="06:50:36.73", separationArcsec="0.66", separationEastArcsec="-0.13", separationNorthArcsec="0.65", specType="SN Ia", transRedshift="0.021", updated=1, dateLastModified=NOW() ; You can save the results to file by passing in a directory path within which to save the files to. The four flavours of data (sources, photometry, spectra and files) are saved to separate files but all data can be assoicated with its transient source using the transient's unique `TNSId`. .. code-block:: python tns.mysql("TNS", "~/tns") .. image:: https://i.imgur.com/CozySPW.png :width: 800px :alt: mysql output
f5187:c0:m11
def _query_tns(self):
self.log.info('<STR_LIT>')<EOL>sourceTable = []<EOL>photoTable = []<EOL>specTable = []<EOL>relatedFilesTable = []<EOL>stop = False<EOL>sourceCount = <NUM_LIT:0><EOL>while not stop:<EOL><INDENT>status_code, content, self._searchURL = self._get_tns_search_results()<EOL>if status_code != <NUM_LIT:200>:<EOL><INDENT>self.log.error(<EOL>'<STR_LIT>' % locals())<EOL>return None<EOL><DEDENT>if "<STR_LIT>" in content:<EOL><INDENT>print("<STR_LIT>")<EOL>return sourceTable, photoTable, specTable, relatedFilesTable<EOL><DEDENT>if self._parse_transient_rows(content, True) < self.batchSize:<EOL><INDENT>stop = True<EOL><DEDENT>else:<EOL><INDENT>self.page += <NUM_LIT:1><EOL>thisPage = self.page<EOL>print("<STR_LIT>" % locals())<EOL>sourceCount += self.batchSize<EOL>print("<STR_LIT:\t>" + self._searchURL)<EOL>timesleep.sleep(<NUM_LIT:1>)<EOL><DEDENT>for transientRow in self._parse_transient_rows(content):<EOL><INDENT>sourceContent = transientRow.group()<EOL>discInfo, TNSId = self._parse_discovery_information(<EOL>sourceContent)<EOL>sourceTable.append(discInfo)<EOL>phot, relatedFiles = self._parse_photometry_data(<EOL>sourceContent, TNSId)<EOL>photoTable += phot<EOL>relatedFilesTable += relatedFiles<EOL>spec, relatedFiles = self._parse_spectral_data(<EOL>sourceContent, TNSId)<EOL>specTable += spec<EOL>relatedFilesTable += relatedFiles<EOL><DEDENT><DEDENT>try:<EOL><INDENT>sourceTable = sorted(sourceTable, key=itemgetter(<EOL>'<STR_LIT>'), reverse=False)<EOL><DEDENT>except:<EOL><INDENT>pass<EOL><DEDENT>self.log.info('<STR_LIT>')<EOL>return sourceTable, photoTable, specTable, relatedFilesTable<EOL>
*determine how to query the TNS, send query and parse the results* **Return:** - ``results`` -- a list of dictionaries (one dictionary for each result set returned from the TNS)
f5187:c0:m12
def _get_tns_search_results(<EOL>self):
self.log.info('<STR_LIT>')<EOL>try:<EOL><INDENT>response = requests.get(<EOL>url="<STR_LIT>",<EOL>params={<EOL>"<STR_LIT>": self.page,<EOL>"<STR_LIT>": self.ra,<EOL>"<STR_LIT>": self.dec,<EOL>"<STR_LIT>": self.radiusArcsec,<EOL>"<STR_LIT:name>": self.name,<EOL>"<STR_LIT>": self.internal_name,<EOL>"<STR_LIT>": self.start,<EOL>"<STR_LIT>": self.end,<EOL>"<STR_LIT>": self.batchSize,<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>"<STR_LIT>": "<STR_LIT:1>",<EOL>},<EOL>)<EOL><DEDENT>except requests.exceptions.RequestException:<EOL><INDENT>print('<STR_LIT>')<EOL><DEDENT>self.log.info('<STR_LIT>')<EOL>return response.status_code, response.content, response.url<EOL>
*query the tns and result the response*
f5187:c0:m13
def _file_prefix(<EOL>self):
self.log.info('<STR_LIT>')<EOL>if self.ra:<EOL><INDENT>now = datetime.now()<EOL>prefix = now.strftime("<STR_LIT>")<EOL><DEDENT>elif self.name:<EOL><INDENT>prefix = self.name + "<STR_LIT>"<EOL><DEDENT>elif self.internal_name:<EOL><INDENT>prefix = self.internal_name + "<STR_LIT>"<EOL><DEDENT>elif self.discInLastDays:<EOL><INDENT>discInLastDays = str(self.discInLastDays)<EOL>now = datetime.now()<EOL>prefix = now.strftime(<EOL>discInLastDays + "<STR_LIT>")<EOL><DEDENT>self.log.info('<STR_LIT>')<EOL>return prefix<EOL>
*Generate a file prefix based on the type of search for saving files to disk* **Return:** - ``prefix`` -- the file prefix
f5187:c0:m14
def _parse_transient_rows(<EOL>self,<EOL>content,<EOL>count=False):
self.log.info('<STR_LIT>')<EOL>regexForRow = r"""<STR_LIT>"""<EOL>if count:<EOL><INDENT>matchedSources = re.findall(<EOL>regexForRow,<EOL>content,<EOL>flags=re.S <EOL>)<EOL>return len(matchedSources)<EOL><DEDENT>matchedSources = re.finditer(<EOL>regexForRow,<EOL>content,<EOL>flags=re.S <EOL>)<EOL>self.log.info('<STR_LIT>')<EOL>return matchedSources<EOL>
* parse transient rows from the TNS result page content* **Key Arguments:** - ``content`` -- the content from the TNS results page. - ``count`` -- return only the number of rows **Return:** - ``transientRows``
f5187:c0:m15
def _parse_discovery_information(<EOL>self,<EOL>content):
self.log.info('<STR_LIT>')<EOL>converter = unit_conversion(<EOL>log=self.log<EOL>)<EOL>matches = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>content,<EOL>flags=<NUM_LIT:0> <EOL>)<EOL>discoveryData = []<EOL>for match in matches:<EOL><INDENT>row = match.groupdict()<EOL>for k, v in row.items():<EOL><INDENT>row[k] = v.strip()<EOL>if len(v) == <NUM_LIT:0>:<EOL><INDENT>row[k] = None<EOL><DEDENT><DEDENT>if row["<STR_LIT>"] == <NUM_LIT:0>:<EOL><INDENT>row["<STR_LIT>"] = None<EOL><DEDENT>if row["<STR_LIT>"][<NUM_LIT:0>] in ["<STR_LIT:1>", "<STR_LIT:2>"]:<EOL><INDENT>row["<STR_LIT>"] = "<STR_LIT>" + row["<STR_LIT>"]<EOL><DEDENT>row["<STR_LIT>"] = "<STR_LIT>" +row["<STR_LIT>"]<EOL>row["<STR_LIT>"] = converter.ra_sexegesimal_to_decimal(<EOL>ra=row["<STR_LIT>"]<EOL>)<EOL>row["<STR_LIT>"] = converter.dec_sexegesimal_to_decimal(<EOL>dec=row["<STR_LIT>"]<EOL>)<EOL>if self.ra:<EOL><INDENT>from astrocalc.coords import separations<EOL>calculator = separations(<EOL>log=self.log,<EOL>ra1=self.ra,<EOL>dec1=self.dec,<EOL>ra2=row["<STR_LIT>"],<EOL>dec2=row["<STR_LIT>"],<EOL>)<EOL>angularSeparation, north, east = calculator.get()<EOL>row["<STR_LIT>"] = angularSeparation<EOL>row["<STR_LIT>"] = north<EOL>row["<STR_LIT>"] = east<EOL><DEDENT>if not row["<STR_LIT>"]:<EOL><INDENT>row["<STR_LIT>"] = row["<STR_LIT>"]<EOL><DEDENT>del row["<STR_LIT>"]<EOL>del row["<STR_LIT>"]<EOL>row["<STR_LIT>"] = row["<STR_LIT>"].replace("<STR_LIT:U+0020>", "<STR_LIT>")<EOL>row["<STR_LIT>"] = row["<STR_LIT>"].replace(<EOL>"<STR_LIT>", "<STR_LIT>").replace("<STR_LIT>", "<STR_LIT>")<EOL>TNSId = row["<STR_LIT>"]<EOL>orow = collections.OrderedDict()<EOL>keyOrder = ["<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>",<EOL>"<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>"]<EOL>for k, v in row.items():<EOL><INDENT>if k not in keyOrder:<EOL><INDENT>keyOrder.append(k)<EOL><DEDENT><DEDENT>for k in keyOrder:<EOL><INDENT>try:<EOL><INDENT>orow[k] = row[k]<EOL><DEDENT>except:<EOL><INDENT>self.log.info(<EOL>"<STR_LIT>" % locals())<EOL>pass<EOL><DEDENT><DEDENT>discoveryData.append(row)<EOL><DEDENT>self.log.info('<STR_LIT>')<EOL>return discoveryData[<NUM_LIT:0>], TNSId<EOL>
* parse discovery information from one row on the TNS results page* **Key Arguments:** - ``content`` -- a table row from the TNS results page. **Return:** - ``discoveryData`` -- dictionary of results - ``TNSId`` -- the unique TNS id for the transient
f5187:c0:m16
def _parse_photometry_data(<EOL>self,<EOL>content,<EOL>TNSId):
self.log.info('<STR_LIT>')<EOL>photData = []<EOL>relatedFilesTable = []<EOL>ATBlock = re.search(<EOL>r"""<STR_LIT>""",<EOL>content,<EOL>flags=re.S <EOL>)<EOL>if ATBlock:<EOL><INDENT>ATBlock = ATBlock.group()<EOL>reports = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>ATBlock,<EOL>flags=re.S <EOL>)<EOL>relatedFiles = self._parse_related_files(ATBlock)<EOL>for r in reports:<EOL><INDENT>header = re.search(<EOL>r"""<STR_LIT>""",<EOL>r.group(),<EOL>flags=<NUM_LIT:0> <EOL>)<EOL>try:<EOL><INDENT>header = header.groupdict()<EOL><DEDENT>except:<EOL><INDENT>print(r.group())<EOL><DEDENT>header["<STR_LIT>"] = TNSId<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>if not self.comments:<EOL><INDENT>del header['<STR_LIT>']<EOL><DEDENT>else:<EOL><INDENT>theseComments = header[<EOL>"<STR_LIT>"].split("<STR_LIT:\n>")<EOL>header["<STR_LIT>"] = "<STR_LIT>"<EOL>for c in theseComments:<EOL><INDENT>header["<STR_LIT>"] += "<STR_LIT:U+0020>" + c.strip()<EOL><DEDENT>header["<STR_LIT>"] = header[<EOL>"<STR_LIT>"].strip().replace('<STR_LIT:">', "<STR_LIT:'>")[<NUM_LIT:0>:<NUM_LIT>]<EOL><DEDENT>phot = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>r.group(),<EOL>flags=<NUM_LIT:0> <EOL>)<EOL>filesAppended = False<EOL>for p in phot:<EOL><INDENT>p = p.groupdict()<EOL>del p["<STR_LIT>"]<EOL>if p["<STR_LIT>"] and not p["<STR_LIT>"]:<EOL><INDENT>p["<STR_LIT>"] = p["<STR_LIT>"]<EOL>p["<STR_LIT>"] = <NUM_LIT:1><EOL>p["<STR_LIT>"] = p["<STR_LIT>"].replace(<EOL>"<STR_LIT>", "<STR_LIT>")<EOL><DEDENT>else:<EOL><INDENT>p["<STR_LIT>"] = <NUM_LIT:0><EOL><DEDENT>if not self.comments:<EOL><INDENT>del p["<STR_LIT>"]<EOL><DEDENT>p.update(header)<EOL>if p["<STR_LIT>"] and filesAppended == False:<EOL><INDENT>filesAppended = True<EOL>for f in relatedFiles:<EOL><INDENT>thisFile = collections.OrderedDict()<EOL>thisFile["<STR_LIT>"] = TNSId<EOL>thisFile["<STR_LIT:filename>"] = f[<EOL>"<STR_LIT>"].split("<STR_LIT:/>")[-<NUM_LIT:1>]<EOL>thisFile["<STR_LIT:url>"] = f["<STR_LIT>"]<EOL>if self.comments:<EOL><INDENT>thisFile["<STR_LIT>"] = f[<EOL>"<STR_LIT>"].replace("<STR_LIT:\n>", "<STR_LIT:U+0020>").strip().replace('<STR_LIT:">', "<STR_LIT:'>")[<NUM_LIT:0>:<NUM_LIT>]<EOL><DEDENT>thisFile["<STR_LIT>"] = p["<STR_LIT>"]<EOL>thisFile["<STR_LIT>"] = <NUM_LIT:2><EOL>relatedFilesTable.append(thisFile)<EOL><DEDENT><DEDENT>if not p["<STR_LIT>"] and not p["<STR_LIT>"]:<EOL><INDENT>p["<STR_LIT>"] = p["<STR_LIT>"]<EOL><DEDENT>del p["<STR_LIT>"]<EOL>del p["<STR_LIT>"]<EOL>orow = collections.OrderedDict()<EOL>keyOrder = ["<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>",<EOL>"<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>"]<EOL>for k, v in p.items():<EOL><INDENT>if k not in keyOrder:<EOL><INDENT>keyOrder.append(k)<EOL><DEDENT><DEDENT>for k in keyOrder:<EOL><INDENT>try:<EOL><INDENT>orow[k] = p[k]<EOL><DEDENT>except:<EOL><INDENT>self.log.info(<EOL>"<STR_LIT>" % locals())<EOL>pass<EOL><DEDENT><DEDENT>photData.append(orow)<EOL><DEDENT><DEDENT><DEDENT>self.log.info('<STR_LIT>')<EOL>return photData, relatedFilesTable<EOL>
*parse photometry data from a row in the tns results content* **Key Arguments:** - ``content`` -- a table row from the TNS results page - ``TNSId`` -- the tns id of the transient **Return:** - ``photData`` -- a list of dictionaries of the photometry data - ``relatedFilesTable`` -- a list of dictionaries of transient photometry related files
f5187:c0:m17
def _parse_related_files(<EOL>self,<EOL>content):
self.log.info('<STR_LIT>')<EOL>relatedFilesList = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>content,<EOL>flags=<NUM_LIT:0> <EOL>)<EOL>relatedFiles = []<EOL>for f in relatedFilesList:<EOL><INDENT>f = f.groupdict()<EOL>relatedFiles.append(f)<EOL><DEDENT>self.log.info('<STR_LIT>')<EOL>return relatedFiles<EOL>
*parse the contents for related files URLs and comments* **Key Arguments:** - ``content`` -- the content to parse. **Return:** - ``relatedFiles`` -- a list of dictionaries of transient related files
f5187:c0:m18
def _parse_spectral_data(<EOL>self,<EOL>content,<EOL>TNSId):
self.log.info('<STR_LIT>')<EOL>specData = []<EOL>relatedFilesTable = []<EOL>classBlock = re.search(<EOL>r"""<STR_LIT>""",<EOL>content,<EOL>flags=re.S <EOL>)<EOL>if classBlock:<EOL><INDENT>classBlock = classBlock.group()<EOL>reports = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>classBlock,<EOL>flags=re.S <EOL>)<EOL>relatedFiles = self._parse_related_files(classBlock)<EOL>for r in reports:<EOL><INDENT>header = re.search(<EOL>r"""<STR_LIT>""",<EOL>r.group(),<EOL>flags=re.S <EOL>)<EOL>if not header:<EOL><INDENT>continue<EOL><DEDENT>header = header.groupdict()<EOL>header["<STR_LIT>"] = TNSId<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>del header["<STR_LIT>"]<EOL>if not self.comments:<EOL><INDENT>del header['<STR_LIT>']<EOL><DEDENT>else:<EOL><INDENT>theseComments = header[<EOL>"<STR_LIT>"].split("<STR_LIT:\n>")<EOL>header["<STR_LIT>"] = "<STR_LIT>"<EOL>for c in theseComments:<EOL><INDENT>header["<STR_LIT>"] += "<STR_LIT:U+0020>" + c.strip()<EOL><DEDENT>header["<STR_LIT>"] = header[<EOL>"<STR_LIT>"].strip().replace('<STR_LIT:">', "<STR_LIT:'>")[<NUM_LIT:0>:<NUM_LIT>]<EOL><DEDENT>spec = re.finditer(<EOL>r"""<STR_LIT>""",<EOL>r.group(),<EOL>flags=<NUM_LIT:0> <EOL>)<EOL>filesAppended = False<EOL>for s in spec:<EOL><INDENT>s = s.groupdict()<EOL>del s["<STR_LIT>"]<EOL>del s["<STR_LIT>"]<EOL>del s["<STR_LIT>"]<EOL>if not self.comments:<EOL><INDENT>del s["<STR_LIT>"]<EOL><DEDENT>else:<EOL><INDENT>s["<STR_LIT>"] = s["<STR_LIT>"].replace('<STR_LIT:">', "<STR_LIT:'>")[<NUM_LIT:0>:<NUM_LIT>]<EOL><DEDENT>s.update(header)<EOL>if s["<STR_LIT>"] and filesAppended == False:<EOL><INDENT>filesAppended = True<EOL>for f in relatedFiles:<EOL><INDENT>thisFile = collections.OrderedDict()<EOL>thisFile["<STR_LIT>"] = TNSId<EOL>thisFile["<STR_LIT:filename>"] = f[<EOL>"<STR_LIT>"].split("<STR_LIT:/>")[-<NUM_LIT:1>]<EOL>thisFile["<STR_LIT:url>"] = f["<STR_LIT>"]<EOL>if self.comments:<EOL><INDENT>thisFile["<STR_LIT>"] = f[<EOL>"<STR_LIT>"].replace("<STR_LIT:\n>", "<STR_LIT:U+0020>").strip()<EOL><DEDENT>thisFile["<STR_LIT>"] = s["<STR_LIT>"]<EOL>thisFile["<STR_LIT>"] = <NUM_LIT:1><EOL>relatedFilesTable.append(thisFile)<EOL><DEDENT><DEDENT>for ffile in [s["<STR_LIT>"], s["<STR_LIT>"]]:<EOL><INDENT>if ffile:<EOL><INDENT>thisFile = collections.OrderedDict()<EOL>thisFile["<STR_LIT>"] = TNSId<EOL>thisFile["<STR_LIT:filename>"] = ffile.split(<EOL>"<STR_LIT:/>")[-<NUM_LIT:1>]<EOL>thisFile["<STR_LIT:url>"] = ffile<EOL>if self.comments:<EOL><INDENT>thisFile["<STR_LIT>"] = "<STR_LIT>"<EOL><DEDENT>thisFile["<STR_LIT>"] = s["<STR_LIT>"]<EOL>thisFile["<STR_LIT>"] = <NUM_LIT:1><EOL>relatedFilesTable.append(thisFile)<EOL><DEDENT><DEDENT>del s["<STR_LIT>"]<EOL>del s["<STR_LIT>"]<EOL>del s["<STR_LIT>"]<EOL>orow = collections.OrderedDict()<EOL>keyOrder = ["<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>",<EOL>"<STR_LIT>", "<STR_LIT>", "<STR_LIT>", "<STR_LIT>"]<EOL>for k, v in s.items():<EOL><INDENT>if k not in keyOrder:<EOL><INDENT>keyOrder.append(k)<EOL><DEDENT><DEDENT>for k in keyOrder:<EOL><INDENT>try:<EOL><INDENT>orow[k] = s[k]<EOL><DEDENT>except:<EOL><INDENT>self.log.info(<EOL>"<STR_LIT>" % locals())<EOL>pass<EOL><DEDENT><DEDENT>specData.append(orow)<EOL><DEDENT><DEDENT><DEDENT>self.log.info('<STR_LIT>')<EOL>return specData, relatedFilesTable<EOL>
*parse spectra data from a row in the tns results content* **Key Arguments:** - ``content`` -- a table row from the TNS results page - ``TNSId`` -- the tns id of the transient **Return:** - ``specData`` -- a list of dictionaries of the spectral data - ``relatedFilesTable`` -- a list of dictionaries of transient spectrum related files
f5187:c0:m19
def getpackagepath():
moduleDirectory = os.path.dirname(__file__)<EOL>packagePath = os.path.dirname(__file__) + "<STR_LIT>"<EOL>return packagePath<EOL>
*getpackagepath*
f5188:m0
def main(arguments=None):
<EOL>su = tools(<EOL>arguments=arguments,<EOL>docString=__doc__,<EOL>logLevel="<STR_LIT>",<EOL>options_first=False,<EOL>projectName="<STR_LIT>"<EOL>)<EOL>arguments, settings, log, dbConn = su.setup()<EOL>readline.set_completer_delims('<STR_LIT>')<EOL>readline.parse_and_bind("<STR_LIT>")<EOL>readline.set_completer(tab_complete)<EOL>for arg, val in arguments.items():<EOL><INDENT>if arg[<NUM_LIT:0>] == "<STR_LIT:->":<EOL><INDENT>varname = arg.replace("<STR_LIT:->", "<STR_LIT>") + "<STR_LIT>"<EOL><DEDENT>else:<EOL><INDENT>varname = arg.replace("<STR_LIT:<>", "<STR_LIT>").replace("<STR_LIT:>>", "<STR_LIT>")<EOL><DEDENT>if isinstance(val, str) or isinstance(val, str):<EOL><INDENT>exec(varname + "<STR_LIT>" % (val,))<EOL><DEDENT>else:<EOL><INDENT>exec(varname + "<STR_LIT>" % (val,))<EOL><DEDENT>if arg == "<STR_LIT>":<EOL><INDENT>dbConn = val<EOL><DEDENT>log.debug('<STR_LIT>' % (varname, val,))<EOL><DEDENT>startTime = times.get_now_sql_datetime()<EOL>log.info(<EOL>'<STR_LIT>' %<EOL>(startTime,))<EOL>if search or new or cone:<EOL><INDENT>if ra:<EOL><INDENT>tns = transientNamer.search(<EOL>log=log,<EOL>ra=ra,<EOL>dec=dec,<EOL>radiusArcsec=arcsecRadius,<EOL>comments=withCommentsFlag<EOL>)<EOL><DEDENT>if name:<EOL><INDENT>tns = transientNamer.search(<EOL>log=log,<EOL>name=name,<EOL>comments=withCommentsFlag<EOL>)<EOL><DEDENT>if discInLastDays:<EOL><INDENT>tns = transientNamer.search(<EOL>log=log,<EOL>discInLastDays=discInLastDays,<EOL>comments=withCommentsFlag<EOL>)<EOL><DEDENT>if outputFlag and not os.path.exists(outputFlag):<EOL><INDENT>os.makedirs(outputFlag)<EOL><DEDENT>if tableNamePrefix:<EOL><INDENT>sources, phot, spec, files = tns.mysql(<EOL>tableNamePrefix=tableNamePrefix, dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT:\n>")) - <NUM_LIT:1><EOL><DEDENT>elif not render or render == "<STR_LIT>":<EOL><INDENT>sources, phot, spec, files = tns.table(dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT:\n>")) - <NUM_LIT:4><EOL><DEDENT>elif render == "<STR_LIT>":<EOL><INDENT>sources, phot, spec, files = tns.csv(dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT:\n>")) - <NUM_LIT:1><EOL><DEDENT>elif render == "<STR_LIT>":<EOL><INDENT>sources, phot, spec, files = tns.json(dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT:{>")) - <NUM_LIT:1><EOL><DEDENT>elif render == "<STR_LIT>":<EOL><INDENT>sources, phot, spec, files = tns.yaml(dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT>"))<EOL><DEDENT>elif render == "<STR_LIT>":<EOL><INDENT>sources, phot, spec, files = tns.markdown(dirPath=outputFlag)<EOL>numSources = len(sources.split("<STR_LIT:\n>")) - <NUM_LIT:2><EOL><DEDENT>if numSources == <NUM_LIT:1>:<EOL><INDENT>print("<STR_LIT>" % locals())<EOL><DEDENT>elif numSources > <NUM_LIT:1>:<EOL><INDENT>print("<STR_LIT>" % locals())<EOL><DEDENT>if not outputFlag:<EOL><INDENT>print("<STR_LIT>")<EOL>print(sources)<EOL>print("<STR_LIT>")<EOL>print(phot)<EOL>print("<STR_LIT>")<EOL>print(spec)<EOL>print("<STR_LIT>")<EOL>print(files)<EOL>print("<STR_LIT>")<EOL>print(tns.url)<EOL><DEDENT><DEDENT>if "<STR_LIT>" in locals() and dbConn:<EOL><INDENT>dbConn.commit()<EOL>dbConn.close()<EOL><DEDENT>endTime = times.get_now_sql_datetime()<EOL>runningTime = times.calculate_time_difference(startTime, endTime)<EOL>log.info('<STR_LIT>' %<EOL>(endTime, runningTime, ))<EOL>return<EOL>
*The main function used when ``cl_utils.py`` is run as a single script from the cl, or when installed as a cl command*
f5190:m1
def match(list_a, list_b, not_found='<STR_LIT>', enforce_sublist=False,<EOL>country_data=COUNTRY_DATA_FILE, additional_data=None):
if isinstance(list_a, str):<EOL><INDENT>list_a = [list_a]<EOL><DEDENT>if isinstance(list_b, str):<EOL><INDENT>list_b = [list_b]<EOL><DEDENT>if isinstance(list_a, tuple):<EOL><INDENT>list_a = list(list_a)<EOL><DEDENT>if isinstance(list_b, tuple):<EOL><INDENT>list_b = list(list_b)<EOL><DEDENT>coco = CountryConverter(country_data, additional_data)<EOL>name_dict_a = dict()<EOL>match_dict_a = dict()<EOL>for name_a in list_a:<EOL><INDENT>name_dict_a[name_a] = []<EOL>match_dict_a[name_a] = []<EOL>for regex in coco.regexes:<EOL><INDENT>if regex.search(name_a):<EOL><INDENT>match_dict_a[name_a].append(regex)<EOL><DEDENT><DEDENT>if len(match_dict_a[name_a]) == <NUM_LIT:0>:<EOL><INDENT>logging.warning('<STR_LIT>'.format(name_a))<EOL>_not_found_entry = name_a if not not_found else not_found<EOL>name_dict_a[name_a].append(_not_found_entry)<EOL>if not enforce_sublist:<EOL><INDENT>name_dict_a[name_a] = name_dict_a[name_a][<NUM_LIT:0>]<EOL><DEDENT>continue<EOL><DEDENT>if len(match_dict_a[name_a]) > <NUM_LIT:1>:<EOL><INDENT>logging.warning(<EOL>'<STR_LIT>'.format(name_a))<EOL><DEDENT>for match_case in match_dict_a[name_a]:<EOL><INDENT>b_matches = <NUM_LIT:0><EOL>for name_b in list_b:<EOL><INDENT>if match_case.search(name_b):<EOL><INDENT>b_matches += <NUM_LIT:1><EOL>name_dict_a[name_a].append(name_b)<EOL><DEDENT><DEDENT><DEDENT>if b_matches == <NUM_LIT:0>:<EOL><INDENT>logging.warning(<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'.format(name_a))<EOL>_not_found_entry = name_a if not not_found else not_found<EOL>name_dict_a[name_a].append(_not_found_entry)<EOL><DEDENT>if b_matches > <NUM_LIT:1>:<EOL><INDENT>logging.warning('<STR_LIT>'<EOL>'<STR_LIT>'.format(name_a))<EOL><DEDENT>if not enforce_sublist and (len(name_dict_a[name_a]) == <NUM_LIT:1>):<EOL><INDENT>name_dict_a[name_a] = name_dict_a[name_a][<NUM_LIT:0>]<EOL><DEDENT><DEDENT>return name_dict_a<EOL>
Matches the country names given in two lists into a dictionary. This function matches names given in list_a to the one provided in list_b using regular expressions defined in country_data. Parameters ---------- list_a : list Names of countries to identify list_b : list Master list of names for countries not_found : str, optional Fill in value for not found entries. If None, keep the input value (default: 'not found') enforce_sublist : boolean, optional If True, all entries in both list are list. If False(default), only multiple matches are list, rest are strings country_data : Pandas DataFrame or path to data file (optional) This is by default set to COUNTRY_DATA_FILE - the standard (tested) country list for coco. additional_data: (list of) Pandas DataFrames or data files (optional) Additional data to include for a specific analysis. This must be given in the same format as specified in the country_data file. (utf-8 encoded tab separated data, same column headers in all files) Returns ------- dict: A dictionary with a key for every entry in list_a. The value correspond to the matching entry in list_b if found. If there is a 1:1 correspondence, the value is a str (if enforce_sublist is False), otherwise multiple entries as list.
f5198:m1
def convert(*args, **kargs):
init = {'<STR_LIT>': COUNTRY_DATA_FILE,<EOL>'<STR_LIT>': None,<EOL>'<STR_LIT>': False,<EOL>'<STR_LIT>': False}<EOL>init.update({kk: kargs.get(kk) for kk in init.keys() if kk in kargs})<EOL>coco = CountryConverter(**init)<EOL>kargs = {kk: ii for kk, ii in kargs.items() if kk not in init.keys()}<EOL>return coco.convert(*args, **kargs)<EOL>
Wrapper around CountryConverter.convert() Uses the same parameters. This function has the same performance as CountryConverter.convert for one call; for multiple calls it is better to instantiate a common CountryConverter (this avoid loading the source data file multiple times). Note ---- A lot of the functionality can also be done directly in Pandas DataFrames. For example: cc = CountryConverter() names = ['USA', 'SWZ', 'PRI'] cc.data[cc.data['ISO3'].isin(names)][['ISO2', 'continent']] Parameters ---------- names : str or list like Countries in 'src' classification to convert to 'to' classification src : str, optional Source classification to : str, optional Output classification (valid str for an index of the country data file), default: name_short enforce_list : boolean, optional If True, enforces the output to be list (if only one name was passed) or to be a list of lists (if multiple names were passed). If False (default), the output will be a string (if only one name was passed) or a list of str and/or lists (str if a one to one matching, list otherwise). not_found : str, optional Fill in value for not found entries. If None, keep the input value (default: 'not found') country_data : Pandas DataFrame or path to data file (optional) This is by default set to COUNTRY_DATA_FILE - the standard (tested) country list for coco. additional_data: (list of) Pandas DataFrames or data files (optional) Additional data to include for a specific analysis. This must be given in the same format as specified in the country_data_file. (utf-8 encoded tab separated data, same column headers as in the general country data file) Returns ------- list or str, depending on enforce_list
f5198:m2
def _parse_arg(valid_classifications):
parser = argparse.ArgumentParser(<EOL>description=('<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'.format(__version__)<EOL>), prog='<STR_LIT>', usage=('<STR_LIT>'))<EOL>parser.add_argument('<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>' +<EOL>'<STR_LIT:U+002CU+0020>'.join(valid_classifications) +<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'<EOL>), nargs='<STR_LIT:*>')<EOL>parser.add_argument('<STR_LIT>', '<STR_LIT>', '<STR_LIT>', '<STR_LIT>', '<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'))<EOL>parser.add_argument('<STR_LIT>', '<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'))<EOL>parser.add_argument('<STR_LIT>', '<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'))<EOL>parser.add_argument('<STR_LIT>', '<STR_LIT>',<EOL>default='<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'))<EOL>parser.add_argument('<STR_LIT>', '<STR_LIT>',<EOL>help=('<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'<EOL>'<STR_LIT>'))<EOL>args = parser.parse_args()<EOL>args.src = args.src or None<EOL>args.to = args.to or '<STR_LIT>'<EOL>args.not_found = args.not_found if args.not_found != '<STR_LIT:None>' else None<EOL>args.output_sep = args.output_sep or '<STR_LIT:U+0020>'<EOL>return args<EOL>
Command line parser for coco Parameters ---------- valid_classifications: list Available classifications, used for checking input parameters. Returns ------- args : ArgumentParser namespace
f5198:m3
def main():
args = _parse_arg(CountryConverter().valid_class)<EOL>coco = CountryConverter(additional_data=args.additional_data)<EOL>converted_names = coco.convert(<EOL>names=args.names,<EOL>src=args.src,<EOL>to=args.to,<EOL>enforce_list=False,<EOL>not_found=args.not_found)<EOL>print(args.output_sep.join(<EOL>[str(etr) for etr in converted_names] if<EOL>isinstance(converted_names, list) else [str(converted_names)]))<EOL>
Main entry point - used for command line call
f5198:m4
@staticmethod<EOL><INDENT>def _separate_exclude_cases(name, exclude_prefix):<DEDENT>
excluder = re.compile('<STR_LIT:|>'.join(exclude_prefix))<EOL>split_entries = excluder.split(name)<EOL>return {'<STR_LIT>': split_entries[<NUM_LIT:0>],<EOL>'<STR_LIT>': split_entries[<NUM_LIT:1>:]}<EOL>
Splits the excluded Parameters ---------- name : str Name of the country/region to convert. exclude_prefix : list of valid regex strings List of indicators which negate the subsequent country/region. These prefixes and everything following will not be converted. E.g. 'Asia excluding China' becomes 'Asia' and 'China excluding Hong Kong' becomes 'China' prior to conversion Returns ------- dict with 'clean_name' : str as name without anything following exclude_prefix 'excluded_countries' : list list of excluded countries
f5198:c0:m0
def __init__(self, country_data=COUNTRY_DATA_FILE,<EOL>additional_data=None, only_UNmember=False,<EOL>include_obsolete=False):
must_be_unique = ['<STR_LIT>', '<STR_LIT>', '<STR_LIT>']<EOL>must_be_string = must_be_unique + (['<STR_LIT>', '<STR_LIT>',<EOL>'<STR_LIT>', '<STR_LIT>',<EOL>'<STR_LIT>', '<STR_LIT>', '<STR_LIT>',<EOL>'<STR_LIT>'])<EOL>def test_for_unique_names(df, data_name='<STR_LIT>',<EOL>report_fun=logging.error):<EOL><INDENT>for name_entry in must_be_unique:<EOL><INDENT>if df[name_entry].duplicated().any():<EOL><INDENT>report_fun('<STR_LIT>'.format(<EOL>name_entry, data_name))<EOL><DEDENT><DEDENT><DEDENT>def data_loader(data):<EOL><INDENT>if isinstance(data, pd.DataFrame):<EOL><INDENT>ret = data<EOL>test_for_unique_names(data)<EOL><DEDENT>else:<EOL><INDENT>ret = pd.read_table(data, sep='<STR_LIT:\t>', encoding='<STR_LIT:utf-8>',<EOL>converters={str_col: str<EOL>for str_col in must_be_string})<EOL>test_for_unique_names(ret, data)<EOL><DEDENT>return ret<EOL><DEDENT>basic_df = data_loader(country_data)<EOL>if only_UNmember:<EOL><INDENT>basic_df.dropna(subset=['<STR_LIT>'], inplace=True)<EOL><DEDENT>if not include_obsolete:<EOL><INDENT>basic_df = basic_df[basic_df.obsolete.isnull()]<EOL><DEDENT>if additional_data is None:<EOL><INDENT>additional_data = []<EOL><DEDENT>if not isinstance(additional_data, list):<EOL><INDENT>additional_data = [additional_data]<EOL><DEDENT>add_data = [data_loader(df) for df in additional_data]<EOL>self.data = pd.concat([basic_df] + add_data, ignore_index=True,<EOL>axis=<NUM_LIT:0>)<EOL>test_for_unique_names(<EOL>self.data,<EOL>data_name='<STR_LIT>',<EOL>report_fun=logging.warning)<EOL>for name_entry in must_be_unique:<EOL><INDENT>self.data.drop_duplicates(subset=[name_entry],<EOL>keep='<STR_LIT>', inplace=True)<EOL><DEDENT>self.data.reset_index(drop=True, inplace=True)<EOL>self.regexes = [re.compile(entry, re.IGNORECASE)<EOL>for entry in self.data.regex]<EOL>
Parameters ---------- country_data : Pandas DataFrame or path to data file This is by default set to COUNTRY_DATA_FILE - the standard (tested) country list for coco. additional_data: (list of) Pandas DataFrames or data files Additioanl data to include for a specific analysis. This must be given in the same format as specified in the country_data file. (utf-8 encoded tab separated data, same column headers in all files) only_UNmember: boolean, optional If True, only load countries currently being UN members from the standard data file. If False (default) load the full list of countries. In this case, also countries currently not existing (USSR) or with overlapping territories are included. include_obsolete: boolean, optional If True, includes countries that have become obsolete. If False (default) only includes currently valid countries.
f5198:c0:m1
def convert(self, names, src=None, to='<STR_LIT>', enforce_list=False,<EOL>not_found='<STR_LIT>',<EOL>exclude_prefix=['<STR_LIT>', '<STR_LIT>', '<STR_LIT>']):
<EOL>names = list(names) if (<EOL>isinstance(names, tuple) or<EOL>isinstance(names, set)) else names<EOL>names = names if isinstance(names, list) else [names]<EOL>names = [str(n) for n in names]<EOL>outlist = names.copy()<EOL>to = [self._validate_input_para(to, self.data.columns)]<EOL>exclude_split = {name: self._separate_exclude_cases(name,<EOL>exclude_prefix)<EOL>for name in names}<EOL>for ind_names, current_name in enumerate(names):<EOL><INDENT>spec_name = exclude_split[current_name]['<STR_LIT>']<EOL>if src is None:<EOL><INDENT>src_format = self._get_input_format_from_name(spec_name)<EOL><DEDENT>else:<EOL><INDENT>src_format = self._validate_input_para(src, self.data.columns)<EOL><DEDENT>if src_format.lower() == '<STR_LIT>':<EOL><INDENT>result_list = []<EOL>for ind_regex, ccregex in enumerate(self.regexes):<EOL><INDENT>if ccregex.search(spec_name):<EOL><INDENT>result_list.append(<EOL>self.data.ix[ind_regex, to].values[<NUM_LIT:0>])<EOL><DEDENT>if len(result_list) > <NUM_LIT:1>:<EOL><INDENT>logging.warning('<STR_LIT>'<EOL>'<STR_LIT>'.format(spec_name))<EOL><DEDENT><DEDENT><DEDENT>else:<EOL><INDENT>_match_col = self.data[src_format].astype(<EOL>str).str.replace('<STR_LIT>', '<STR_LIT>')<EOL>result_list = [etr[<NUM_LIT:0>] for etr in<EOL>self.data[_match_col.str.contains(<EOL>'<STR_LIT>' + spec_name + '<STR_LIT:$>', flags=re.IGNORECASE,<EOL>na=False)][to].values]<EOL><DEDENT>if len(result_list) == <NUM_LIT:0>:<EOL><INDENT>logging.warning(<EOL>'<STR_LIT>'.format(spec_name, src_format))<EOL>_fillin = not_found or spec_name<EOL>outlist[ind_names] = [_fillin] if enforce_list else _fillin<EOL><DEDENT>else:<EOL><INDENT>outlist[ind_names] = []<EOL>for etr in result_list:<EOL><INDENT>try:<EOL><INDENT>conv_etr = int(etr)<EOL><DEDENT>except ValueError:<EOL><INDENT>conv_etr = etr<EOL><DEDENT>outlist[ind_names].append(conv_etr)<EOL><DEDENT>if len(outlist[ind_names]) == <NUM_LIT:1> and enforce_list is False:<EOL><INDENT>outlist[ind_names] = outlist[ind_names][<NUM_LIT:0>]<EOL><DEDENT><DEDENT><DEDENT>if (len(outlist) == <NUM_LIT:1>) and not enforce_list:<EOL><INDENT>return outlist[<NUM_LIT:0>]<EOL><DEDENT>else:<EOL><INDENT>return outlist<EOL><DEDENT>
Convert names from a list to another list. Note ---- A lot of the functionality can also be done directly in Pandas DataFrames. For example: coco = CountryConverter() names = ['USA', 'SWZ', 'PRI'] coco.data[coco.data['ISO3'].isin(names)][['ISO2', 'continent']] Parameters ---------- names : str or list like Countries in 'src' classification to convert to 'to' classification src : str, optional Source classification. If None (default), each passed name is checked if it is a number (assuming UNnumeric) or 2 (ISO2) or 3 (ISO3) characters long; for longer names 'regex' is assumed. to : str, optional Output classification (valid index of the country_data file), default: ISO3 enforce_list : boolean, optional If True, enforces the output to be list (if only one name was passed) or to be a list of lists (if multiple names were passed). If False (default), the output will be a string (if only one name was passed) or a list of str and/or lists (str if a one to one matching, list otherwise). not_found : str, optional Fill in value for none found entries. If None, keep the input value (default: 'not found') exclude_prefix : list of valid regex strings List of indicators which negate the subsequent country/region. These prefixes and everything following will not be converted. E.g. 'Asia excluding China' becomes 'Asia' and 'China excluding Hong Kong' becomes 'China' prior to conversion Default: ['excl\\w.*', 'without', 'w/o']) Returns ------- list or str, depending on enforce_list
f5198:c0:m2
def EU28as(self, to='<STR_LIT>'):
if type(to) is str:<EOL><INDENT>to = [to]<EOL><DEDENT>return self.data[self.data.EU < <NUM_LIT>][to]<EOL>
Return EU28 countries in the specified classification Parameters ---------- to : str, optional Output classification (valid str for an index of country_data file), default: name_short Returns ------- Pandas DataFrame
f5198:c0:m3