mirror of
https://github.com/thegeeklab/ansible-later.git
synced 2024-11-14 17:20:39 +00:00
159 lines
4.9 KiB
Python
159 lines
4.9 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 yaml
|
|
|
|
|
|
class Line(object):
|
|
def __init__(self, line_no, buffer, start, end):
|
|
self.line_no = line_no
|
|
self.start = start
|
|
self.end = end
|
|
self.buffer = buffer
|
|
|
|
@property
|
|
def content(self):
|
|
return self.buffer[self.start:self.end]
|
|
|
|
|
|
class Token(object):
|
|
def __init__(self, line_no, curr, prev, next, nextnext):
|
|
self.line_no = line_no
|
|
self.curr = curr
|
|
self.prev = prev
|
|
self.next = next
|
|
self.nextnext = nextnext
|
|
|
|
|
|
class Comment(object):
|
|
def __init__(self, line_no, column_no, buffer, pointer,
|
|
token_before=None, token_after=None, comment_before=None):
|
|
self.line_no = line_no
|
|
self.column_no = column_no
|
|
self.buffer = buffer
|
|
self.pointer = pointer
|
|
self.token_before = token_before
|
|
self.token_after = token_after
|
|
self.comment_before = comment_before
|
|
|
|
def __str__(self):
|
|
end = self.buffer.find('\n', self.pointer)
|
|
if end == -1:
|
|
end = self.buffer.find('\0', self.pointer)
|
|
if end != -1:
|
|
return self.buffer[self.pointer:end]
|
|
return self.buffer[self.pointer:]
|
|
|
|
def __eq__(self, other):
|
|
return (isinstance(other, Comment) and
|
|
self.line_no == other.line_no and
|
|
self.column_no == other.column_no and
|
|
str(self) == str(other))
|
|
|
|
def is_inline(self):
|
|
return (
|
|
not isinstance(self.token_before, yaml.StreamStartToken) and
|
|
self.line_no == self.token_before.end_mark.line + 1 and
|
|
# sometimes token end marks are on the next line
|
|
self.buffer[self.token_before.end_mark.pointer - 1] != '\n'
|
|
)
|
|
|
|
|
|
def line_generator(buffer):
|
|
line_no = 1
|
|
cur = 0
|
|
next = buffer.find('\n')
|
|
while next != -1:
|
|
yield Line(line_no, buffer, start=cur, end=next)
|
|
cur = next + 1
|
|
next = buffer.find('\n', cur)
|
|
line_no += 1
|
|
|
|
yield Line(line_no, buffer, start=cur, end=len(buffer))
|
|
|
|
|
|
def comments_between_tokens(token1, token2):
|
|
"""Find all comments between two tokens"""
|
|
if token2 is None:
|
|
buf = token1.end_mark.buffer[token1.end_mark.pointer:]
|
|
elif (token1.end_mark.line == token2.start_mark.line and
|
|
not isinstance(token1, yaml.StreamStartToken) and
|
|
not isinstance(token2, yaml.StreamEndToken)):
|
|
return
|
|
else:
|
|
buf = token1.end_mark.buffer[token1.end_mark.pointer:
|
|
token2.start_mark.pointer]
|
|
|
|
line_no = token1.end_mark.line + 1
|
|
column_no = token1.end_mark.column + 1
|
|
pointer = token1.end_mark.pointer
|
|
|
|
comment_before = None
|
|
for line in buf.split('\n'):
|
|
pos = line.find('#')
|
|
if pos != -1:
|
|
comment = Comment(line_no, column_no + pos,
|
|
token1.end_mark.buffer, pointer + pos,
|
|
token1, token2, comment_before)
|
|
yield comment
|
|
|
|
comment_before = comment
|
|
|
|
pointer += len(line) + 1
|
|
line_no += 1
|
|
column_no = 1
|
|
|
|
|
|
def token_or_comment_generator(buffer):
|
|
yaml_loader = yaml.BaseLoader(buffer)
|
|
|
|
try:
|
|
prev = None
|
|
curr = yaml_loader.get_token()
|
|
while curr is not None:
|
|
next = yaml_loader.get_token()
|
|
nextnext = (yaml_loader.peek_token()
|
|
if yaml_loader.check_token() else None)
|
|
|
|
yield Token(curr.start_mark.line + 1, curr, prev, next, nextnext)
|
|
|
|
for comment in comments_between_tokens(curr, next):
|
|
yield comment
|
|
|
|
prev = curr
|
|
curr = next
|
|
|
|
except yaml.scanner.ScannerError:
|
|
pass
|
|
|
|
|
|
def token_or_comment_or_line_generator(buffer):
|
|
"""Generator that mixes tokens and lines, ordering them by line number"""
|
|
tok_or_com_gen = token_or_comment_generator(buffer)
|
|
line_gen = line_generator(buffer)
|
|
|
|
tok_or_com = next(tok_or_com_gen, None)
|
|
line = next(line_gen, None)
|
|
|
|
while tok_or_com is not None or line is not None:
|
|
if tok_or_com is None or (line is not None and
|
|
tok_or_com.line_no > line.line_no):
|
|
yield line
|
|
line = next(line_gen, None)
|
|
else:
|
|
yield tok_or_com
|
|
tok_or_com = next(tok_or_com_gen, None)
|