mirror of
https://github.com/thegeeklab/ansible-later.git
synced 2024-11-26 14:50:36 +00:00
237 lines
8.4 KiB
Python
237 lines
8.4 KiB
Python
# -*- coding: utf-8 -*-
|
|
# Copyright (C) 2016 Adrien Vergé
|
|
#
|
|
# This program is free software: you can redistribute it and/or modify
|
|
# it under the terms of the GNU General Public License as published by
|
|
# the Free Software Foundation, either version 3 of the License, or
|
|
# (at your option) any later version.
|
|
#
|
|
# 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 General Public License
|
|
# along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
import re
|
|
|
|
import yaml
|
|
|
|
from yamllint import parser
|
|
|
|
|
|
PROBLEM_LEVELS = {
|
|
0: None,
|
|
1: 'warning',
|
|
2: 'error',
|
|
None: 0,
|
|
'warning': 1,
|
|
'error': 2,
|
|
}
|
|
|
|
|
|
class LintProblem(object):
|
|
"""Represents a linting problem found by yamllint."""
|
|
def __init__(self, line, column, desc='<no description>', rule=None):
|
|
#: Line on which the problem was found (starting at 1)
|
|
self.line = line
|
|
#: Column on which the problem was found (starting at 1)
|
|
self.column = column
|
|
#: Human-readable description of the problem
|
|
self.desc = desc
|
|
#: Identifier of the rule that detected the problem
|
|
self.rule = rule
|
|
self.level = None
|
|
|
|
@property
|
|
def message(self):
|
|
if self.rule is not None:
|
|
return '{} ({})'.format(self.desc, self.rule)
|
|
return self.desc
|
|
|
|
def __eq__(self, other):
|
|
return (self.line == other.line and
|
|
self.column == other.column and
|
|
self.rule == other.rule)
|
|
|
|
def __lt__(self, other):
|
|
return (self.line < other.line or
|
|
(self.line == other.line and self.column < other.column))
|
|
|
|
def __repr__(self):
|
|
return '%d:%d: %s' % (self.line, self.column, self.message)
|
|
|
|
|
|
def get_cosmetic_problems(buffer, conf, filepath):
|
|
rules = conf.enabled_rules(filepath)
|
|
|
|
# Split token rules from line rules
|
|
token_rules = [r for r in rules if r.TYPE == 'token']
|
|
comment_rules = [r for r in rules if r.TYPE == 'comment']
|
|
line_rules = [r for r in rules if r.TYPE == 'line']
|
|
|
|
context = {}
|
|
for rule in token_rules:
|
|
context[rule.ID] = {}
|
|
|
|
class DisableDirective:
|
|
def __init__(self):
|
|
self.rules = set()
|
|
self.all_rules = {r.ID for r in rules}
|
|
|
|
def process_comment(self, comment):
|
|
try:
|
|
comment = str(comment)
|
|
except UnicodeError:
|
|
return # this certainly wasn't a yamllint directive comment
|
|
|
|
if re.match(r'^# yamllint disable( rule:\S+)*\s*$', comment):
|
|
rules = [item[5:] for item in comment[18:].split(' ')][1:]
|
|
if len(rules) == 0:
|
|
self.rules = self.all_rules.copy()
|
|
else:
|
|
for id in rules:
|
|
if id in self.all_rules:
|
|
self.rules.add(id)
|
|
|
|
elif re.match(r'^# yamllint enable( rule:\S+)*\s*$', comment):
|
|
rules = [item[5:] for item in comment[17:].split(' ')][1:]
|
|
if len(rules) == 0:
|
|
self.rules.clear()
|
|
else:
|
|
for id in rules:
|
|
self.rules.discard(id)
|
|
|
|
def is_disabled_by_directive(self, problem):
|
|
return problem.rule in self.rules
|
|
|
|
class DisableLineDirective(DisableDirective):
|
|
def process_comment(self, comment):
|
|
try:
|
|
comment = str(comment)
|
|
except UnicodeError:
|
|
return # this certainly wasn't a yamllint directive comment
|
|
|
|
if re.match(r'^# yamllint disable-line( rule:\S+)*\s*$', comment):
|
|
rules = [item[5:] for item in comment[23:].split(' ')][1:]
|
|
if len(rules) == 0:
|
|
self.rules = self.all_rules.copy()
|
|
else:
|
|
for id in rules:
|
|
if id in self.all_rules:
|
|
self.rules.add(id)
|
|
|
|
# Use a cache to store problems and flush it only when a end of line is
|
|
# found. This allows the use of yamllint directive to disable some rules on
|
|
# some lines.
|
|
cache = []
|
|
disabled = DisableDirective()
|
|
disabled_for_line = DisableLineDirective()
|
|
disabled_for_next_line = DisableLineDirective()
|
|
|
|
for elem in parser.token_or_comment_or_line_generator(buffer):
|
|
if isinstance(elem, parser.Token):
|
|
for rule in token_rules:
|
|
rule_conf = conf.rules[rule.ID]
|
|
for problem in rule.check(rule_conf,
|
|
elem.curr, elem.prev, elem.next,
|
|
elem.nextnext,
|
|
context[rule.ID]):
|
|
problem.rule = rule.ID
|
|
problem.level = rule_conf['level']
|
|
cache.append(problem)
|
|
elif isinstance(elem, parser.Comment):
|
|
for rule in comment_rules:
|
|
rule_conf = conf.rules[rule.ID]
|
|
for problem in rule.check(rule_conf, elem):
|
|
problem.rule = rule.ID
|
|
problem.level = rule_conf['level']
|
|
cache.append(problem)
|
|
|
|
disabled.process_comment(elem)
|
|
if elem.is_inline():
|
|
disabled_for_line.process_comment(elem)
|
|
else:
|
|
disabled_for_next_line.process_comment(elem)
|
|
elif isinstance(elem, parser.Line):
|
|
for rule in line_rules:
|
|
rule_conf = conf.rules[rule.ID]
|
|
for problem in rule.check(rule_conf, elem):
|
|
problem.rule = rule.ID
|
|
problem.level = rule_conf['level']
|
|
cache.append(problem)
|
|
|
|
# This is the last token/comment/line of this line, let's flush the
|
|
# problems found (but filter them according to the directives)
|
|
for problem in cache:
|
|
if not (disabled_for_line.is_disabled_by_directive(problem) or
|
|
disabled.is_disabled_by_directive(problem)):
|
|
yield problem
|
|
|
|
disabled_for_line = disabled_for_next_line
|
|
disabled_for_next_line = DisableLineDirective()
|
|
cache = []
|
|
|
|
|
|
def get_syntax_error(buffer):
|
|
try:
|
|
list(yaml.parse(buffer, Loader=yaml.BaseLoader))
|
|
except yaml.error.MarkedYAMLError as e:
|
|
problem = LintProblem(e.problem_mark.line + 1,
|
|
e.problem_mark.column + 1,
|
|
'syntax error: ' + e.problem)
|
|
problem.level = 'error'
|
|
return problem
|
|
|
|
|
|
def _run(buffer, conf, filepath):
|
|
assert hasattr(buffer, '__getitem__'), \
|
|
'_run() argument must be a buffer, not a stream'
|
|
|
|
# If the document contains a syntax error, save it and yield it at the
|
|
# right line
|
|
syntax_error = get_syntax_error(buffer)
|
|
|
|
for problem in get_cosmetic_problems(buffer, conf, filepath):
|
|
# Insert the syntax error (if any) at the right place...
|
|
if (syntax_error and syntax_error.line <= problem.line and
|
|
syntax_error.column <= problem.column):
|
|
yield syntax_error
|
|
|
|
# If there is already a yamllint error at the same place, discard
|
|
# it as it is probably redundant (and maybe it's just a 'warning',
|
|
# in which case the script won't even exit with a failure status).
|
|
if (syntax_error.line == problem.line and
|
|
syntax_error.column == problem.column):
|
|
syntax_error = None
|
|
continue
|
|
|
|
syntax_error = None
|
|
|
|
yield problem
|
|
|
|
if syntax_error:
|
|
yield syntax_error
|
|
|
|
|
|
def run(input, conf, filepath=None):
|
|
"""Lints a YAML source.
|
|
|
|
Returns a generator of LintProblem objects.
|
|
|
|
:param input: buffer, string or stream to read from
|
|
:param conf: yamllint configuration object
|
|
"""
|
|
if conf.is_file_ignored(filepath):
|
|
return ()
|
|
|
|
if isinstance(input, (type(b''), type(u''))): # compat with Python 2 & 3
|
|
return _run(input, conf, filepath)
|
|
elif hasattr(input, 'read'): # Python 2's file or Python 3's io.IOBase
|
|
# We need to have everything in memory to parse correctly
|
|
content = input.read()
|
|
return _run(content, conf, filepath)
|
|
else:
|
|
raise TypeError('input should be a string or a stream')
|