##// END OF EJS Templates
debugging: expose logs/exception when debug log is enabled.
debugging: expose logs/exception when debug log is enabled.

File last commit:

r4306:09801de9 default
r4768:f604047c default
Show More
search_utils.py
197 lines | 6.0 KiB | text/x-python | PythonLexer
dan
search: add support for elastic search 6...
r3319 # -*- coding: utf-8 -*-
code: update copyrights to 2020
r4306 # Copyright (C) 2012-2020 RhodeCode GmbH
dan
search: add support for elastic search 6...
r3319 #
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License, version 3
# (only), as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
# This program is dual-licensed. If you wish to learn more about the
# RhodeCode Enterprise Edition, including its added features, Support services,
# and proprietary license terms, please see https://rhodecode.com/licenses/
import re
import pygments.filter
import pygments.filters
from pygments.token import Comment
HL_BEG_MARKER = '__RCSearchHLMarkBEG__'
HL_END_MARKER = '__RCSearchHLMarkEND__'
HL_MARKER_RE = '{}(.*?){}'.format(HL_BEG_MARKER, HL_END_MARKER)
class ElasticSearchHLFilter(pygments.filters.Filter):
_names = [HL_BEG_MARKER, HL_END_MARKER]
def __init__(self, **options):
pygments.filters.Filter.__init__(self, **options)
def filter(self, lexer, stream):
def tokenize(_value):
for token in re.split('({}|{})'.format(
self._names[0], self._names[1]), _value):
if token:
yield token
hl = False
for ttype, value in stream:
if self._names[0] in value or self._names[1] in value:
for item in tokenize(value):
if item == self._names[0]:
# skip marker, but start HL
hl = True
continue
elif item == self._names[1]:
hl = False
continue
if hl:
yield Comment.ElasticMatch, item
else:
yield ttype, item
else:
if hl:
yield Comment.ElasticMatch, value
else:
yield ttype, value
def extract_phrases(text_query):
"""
Extracts phrases from search term string making sure phrases
contained in double quotes are kept together - and discarding empty values
or fully whitespace values eg.
'some text "a phrase" more' => ['some', 'text', 'a phrase', 'more']
"""
in_phrase = False
buf = ''
phrases = []
for char in text_query:
if in_phrase:
if char == '"': # end phrase
phrases.append(buf)
buf = ''
in_phrase = False
continue
else:
buf += char
continue
else:
if char == '"': # start phrase
in_phrase = True
phrases.append(buf)
buf = ''
continue
elif char == ' ':
phrases.append(buf)
buf = ''
continue
else:
buf += char
phrases.append(buf)
phrases = [phrase.strip() for phrase in phrases if phrase.strip()]
return phrases
def get_matching_phrase_offsets(text, phrases):
"""
Returns a list of string offsets in `text` that the list of `terms` match
>>> get_matching_phrase_offsets('some text here', ['some', 'here'])
[(0, 4), (10, 14)]
"""
phrases = phrases or []
offsets = []
for phrase in phrases:
for match in re.finditer(phrase, text):
offsets.append((match.start(), match.end()))
return offsets
def get_matching_markers_offsets(text, markers=None):
"""
Returns a list of string offsets in `text` that the are between matching markers
>>> get_matching_markers_offsets('$1some$2 text $1here$2 marked', ['\$1(.*?)\$2'])
[(0, 5), (16, 22)]
"""
markers = markers or [HL_MARKER_RE]
offsets = []
if markers:
for mark in markers:
for match in re.finditer(mark, text):
offsets.append((match.start(), match.end()))
return offsets
def normalize_text_for_matching(x):
"""
Replaces all non alfanum characters to spaces and lower cases the string,
useful for comparing two text strings without punctuation
"""
return re.sub(r'[^\w]', ' ', x.lower())
def get_matching_line_offsets(lines, terms=None, markers=None):
""" Return a set of `lines` indices (starting from 1) matching a
text search query, along with `context` lines above/below matching lines
:param lines: list of strings representing lines
:param terms: search term string to match in lines eg. 'some text'
:param markers: instead of terms, use highlight markers instead that
mark beginning and end for matched item. eg. ['START(.*?)END']
eg.
text = '''
words words words
words words words
some text some
words words words
words words words
text here what
'''
get_matching_line_offsets(text, 'text', context=1)
6, {3: [(5, 9)], 6: [(0, 4)]]
"""
matching_lines = {}
line_index = 0
if terms:
phrases = [normalize_text_for_matching(phrase)
for phrase in extract_phrases(terms)]
for line_index, line in enumerate(lines.splitlines(), start=1):
normalized_line = normalize_text_for_matching(line)
match_offsets = get_matching_phrase_offsets(normalized_line, phrases)
if match_offsets:
matching_lines[line_index] = match_offsets
else:
markers = markers or [HL_MARKER_RE]
for line_index, line in enumerate(lines.splitlines(), start=1):
match_offsets = get_matching_markers_offsets(line, markers=markers)
if match_offsets:
matching_lines[line_index] = match_offsets
return line_index, matching_lines