""" pygments.lexers.diff ~~~~~~~~~~~~~~~~~~~~ Lexers for diff/patch formats. :copyright: Copyright 2006-2021 by the Pygments team, see AUTHORS. :license: BSD, see LICENSE for details. """ import re from pygments.lexer import RegexLexer, include, bygroups from pygments.token import Text, Comment, Operator, Keyword, Name, Generic, \ Literal __all__ = ['DiffLexer', 'DarcsPatchLexer', 'WDiffLexer'] class DiffLexer(RegexLexer): """ Lexer for unified or context-style diffs or patches. """ name = 'Diff' aliases = ['diff', 'udiff'] filenames = ['*.diff', '*.patch'] mimetypes = ['text/x-diff', 'text/x-patch'] tokens = { 'root': [ (r' .*\n', Text), (r'\+.*\n', Generic.Inserted), (r'-.*\n', Generic.Deleted), (r'!.*\n', Generic.Strong), (r'@.*\n', Generic.Subheading), (r'([Ii]ndex|diff).*\n', Generic.Heading), (r'=.*\n', Generic.Heading), (r'.*\n', Text), ] } def analyse_text(text): if text[:7] == 'Index: ': return True if text[:5] == 'diff ': return True if text[:4] == '--- ': return 0.9 class DarcsPatchLexer(RegexLexer): """ DarcsPatchLexer is a lexer for the various versions of the darcs patch format. Examples of this format are derived by commands such as ``darcs annotate --patch`` and ``darcs send``. .. versionadded:: 0.10 """ name = 'Darcs Patch' aliases = ['dpatch'] filenames = ['*.dpatch', '*.darcspatch'] DPATCH_KEYWORDS = ('hunk', 'addfile', 'adddir', 'rmfile', 'rmdir', 'move', 'replace') tokens = { 'root': [ (r'<', Operator), (r'>', Operator), (r'\{', Operator), (r'\}', Operator), (r'(\[)((?:TAG )?)(.*)(\n)(.*)(\*\*)(\d+)(\s?)(\])', bygroups(Operator, Keyword, Name, Text, Name, Operator, Literal.Date, Text, Operator)), (r'(\[)((?:TAG )?)(.*)(\n)(.*)(\*\*)(\d+)(\s?)', bygroups(Operator, Keyword, Name, Text, Name, Operator, Literal.Date, Text), 'comment'), (r'New patches:', Generic.Heading), (r'Context:', Generic.Heading), (r'Patch bundle hash:', Generic.Heading), (r'(\s*)(%s)(.*\n)' % '|'.join(DPATCH_KEYWORDS), bygroups(Text, Keyword, Text)), (r'\+', Generic.Inserted, "insert"), (r'-', Generic.Deleted, "delete"), (r'.*\n', Text), ], 'comment': [ (r'[^\]].*\n', Comment), (r'\]', Operator, "#pop"), ], 'specialText': [ # darcs add [_CODE_] special operators for clarity (r'\n', Text, "#pop"), # line-based (r'\[_[^_]*_]', Operator), ], 'insert': [ include('specialText'), (r'\[', Generic.Inserted), (r'[^\n\[]+', Generic.Inserted), ], 'delete': [ include('specialText'), (r'\[', Generic.Deleted), (r'[^\n\[]+', Generic.Deleted), ], } class WDiffLexer(RegexLexer): """ A `wdiff `_ lexer. Note that: * It only works with normal output (without options like ``-l``). * If the target files contain "[-", "-]", "{+", or "+}", especially they are unbalanced, the lexer will get confused. .. versionadded:: 2.2 """ name = 'WDiff' aliases = ['wdiff'] filenames = ['*.wdiff'] mimetypes = [] flags = re.MULTILINE | re.DOTALL # We can only assume "[-" after "[-" before "-]" is `nested`, # for instance wdiff to wdiff outputs. We have no way to # distinct these marker is of wdiff output from original text. ins_op = r"\{\+" ins_cl = r"\+\}" del_op = r"\[\-" del_cl = r"\-\]" normal = r'[^{}[\]+-]+' # for performance tokens = { 'root': [ (ins_op, Generic.Inserted, 'inserted'), (del_op, Generic.Deleted, 'deleted'), (normal, Text), (r'.', Text), ], 'inserted': [ (ins_op, Generic.Inserted, '#push'), (del_op, Generic.Inserted, '#push'), (del_cl, Generic.Inserted, '#pop'), (ins_cl, Generic.Inserted, '#pop'), (normal, Generic.Inserted), (r'.', Generic.Inserted), ], 'deleted': [ (del_op, Generic.Deleted, '#push'), (ins_op, Generic.Deleted, '#push'), (ins_cl, Generic.Deleted, '#pop'), (del_cl, Generic.Deleted, '#pop'), (normal, Generic.Deleted), (r'.', Generic.Deleted), ], }