mirror of
https://github.com/oxen-io/session-ios.git
synced 2023-12-13 21:30:14 +01:00
17ed0f6100
// FREEBIE
427 lines
13 KiB
Python
Executable file
427 lines
13 KiB
Python
Executable file
#!/usr/bin/env python
|
|
# -*- coding: utf-8 -*-
|
|
|
|
import os
|
|
import sys
|
|
import subprocess
|
|
import datetime
|
|
import argparse
|
|
import commands
|
|
|
|
|
|
git_repo_path = os.path.abspath(subprocess.check_output(['git', 'rev-parse', '--show-toplevel']).strip())
|
|
|
|
|
|
|
|
class include:
|
|
def __init__(self, isInclude, isQuote, body, comment):
|
|
self.isInclude = isInclude
|
|
self.isQuote = isQuote
|
|
self.body = body
|
|
self.comment = comment
|
|
|
|
def format(self):
|
|
result = '%s %s%s%s' % (
|
|
('#include' if self.isInclude else '#import'),
|
|
('"' if self.isQuote else '<'),
|
|
self.body.strip(),
|
|
('"' if self.isQuote else '>'),
|
|
)
|
|
if self.comment.strip():
|
|
result += ' ' + self.comment.strip()
|
|
return result
|
|
|
|
|
|
def is_include_or_import(line):
|
|
line = line.strip()
|
|
if line.startswith('#include '):
|
|
return True
|
|
elif line.startswith('#import '):
|
|
return True
|
|
else:
|
|
return False
|
|
|
|
|
|
def parse_include(line):
|
|
remainder = line.strip()
|
|
|
|
if remainder.startswith('#include '):
|
|
isInclude = True
|
|
remainder = remainder[len('#include '):]
|
|
elif remainder.startswith('#import '):
|
|
isInclude = False
|
|
remainder = remainder[len('#import '):]
|
|
elif remainder == '//':
|
|
return None
|
|
elif not remainder:
|
|
return None
|
|
else:
|
|
print ('Unexpected import or include: '+ line)
|
|
sys.exit(1)
|
|
|
|
comment = None
|
|
if remainder.startswith('"'):
|
|
isQuote = True
|
|
endIndex = remainder.find('"', 1)
|
|
if endIndex < 0:
|
|
print ('Unexpected import or include: '+ line)
|
|
sys.exit(1)
|
|
body = remainder[1:endIndex]
|
|
comment = remainder[endIndex+1:]
|
|
elif remainder.startswith('<'):
|
|
isQuote = False
|
|
endIndex = remainder.find('>', 1)
|
|
if endIndex < 0:
|
|
print ('Unexpected import or include: '+ line)
|
|
sys.exit(1)
|
|
body = remainder[1:endIndex]
|
|
comment = remainder[endIndex+1:]
|
|
else:
|
|
print ('Unexpected import or include: '+ remainder)
|
|
sys.exit(1)
|
|
|
|
return include(isInclude, isQuote, body, comment)
|
|
|
|
|
|
def parse_includes(text):
|
|
lines = text.split('\n')
|
|
|
|
includes = []
|
|
for line in lines:
|
|
include = parse_include(line)
|
|
if include:
|
|
includes.append(include)
|
|
|
|
return includes
|
|
|
|
|
|
def sort_include_block(text, filepath, filename, file_extension):
|
|
lines = text.split('\n')
|
|
|
|
includes = parse_includes(text)
|
|
|
|
blocks = []
|
|
|
|
file_extension = file_extension.lower()
|
|
|
|
for include in includes:
|
|
include.isInclude = False
|
|
|
|
if file_extension in ('c', 'cpp', 'hpp'):
|
|
for include in includes:
|
|
include.isInclude = True
|
|
elif file_extension in ('m'):
|
|
for include in includes:
|
|
include.isInclude = False
|
|
|
|
# Make sure matching header is first.
|
|
matching_header_includes = []
|
|
other_includes = []
|
|
def is_matching_header(include):
|
|
filename_wo_ext = os.path.splitext(filename)[0]
|
|
include_filename_wo_ext = os.path.splitext(os.path.basename(include.body))[0]
|
|
return filename_wo_ext == include_filename_wo_ext
|
|
for include in includes:
|
|
if is_matching_header(include):
|
|
matching_header_includes.append(include)
|
|
else:
|
|
other_includes.append(include)
|
|
includes = other_includes
|
|
|
|
def formatBlock(includes):
|
|
lines = [include.format() for include in includes]
|
|
lines = list(set(lines))
|
|
def include_sorter(a, b):
|
|
# return cmp(a.lower(), b.lower())
|
|
return cmp(a, b)
|
|
# print 'before'
|
|
# for line in lines:
|
|
# print '\t', line
|
|
# print
|
|
lines.sort(include_sorter)
|
|
# print 'after'
|
|
# for line in lines:
|
|
# print '\t', line
|
|
# print
|
|
# print
|
|
# print 'filepath'
|
|
# for line in lines:
|
|
# print '\t', line
|
|
# print
|
|
return '\n'.join(lines)
|
|
|
|
includeAngles = [include for include in includes if include.isInclude and not include.isQuote]
|
|
includeQuotes = [include for include in includes if include.isInclude and include.isQuote]
|
|
importAngles = [include for include in includes if (not include.isInclude) and not include.isQuote]
|
|
importQuotes = [include for include in includes if (not include.isInclude) and include.isQuote]
|
|
if matching_header_includes:
|
|
blocks.append(formatBlock(matching_header_includes))
|
|
if includeQuotes:
|
|
blocks.append(formatBlock(includeQuotes))
|
|
if includeAngles:
|
|
blocks.append(formatBlock(includeAngles))
|
|
if importQuotes:
|
|
blocks.append(formatBlock(importQuotes))
|
|
if importAngles:
|
|
blocks.append(formatBlock(importAngles))
|
|
|
|
return '\n'.join(blocks) + '\n'
|
|
|
|
|
|
def sort_class_statement_block(text, filepath, filename, file_extension):
|
|
lines = text.split('\n')
|
|
lines = [line.strip() for line in lines if line.strip()]
|
|
lines.sort()
|
|
return '\n' + '\n'.join(lines) + '\n'
|
|
|
|
|
|
def find_matching_section(text, match_test):
|
|
lines = text.split('\n')
|
|
first_matching_line_index = None
|
|
for index, line in enumerate(lines):
|
|
if match_test(line):
|
|
first_matching_line_index = index
|
|
break
|
|
|
|
if first_matching_line_index is None:
|
|
return None
|
|
|
|
# Absorb any leading empty lines.
|
|
while first_matching_line_index > 0:
|
|
prev_line = lines[first_matching_line_index - 1]
|
|
if prev_line.strip():
|
|
break
|
|
first_matching_line_index = first_matching_line_index - 1
|
|
|
|
first_non_matching_line_index = None
|
|
for index, line in enumerate(lines[first_matching_line_index:]):
|
|
if not line.strip():
|
|
# Absorb any trailing empty lines.
|
|
continue
|
|
if not match_test(line):
|
|
first_non_matching_line_index = index + first_matching_line_index
|
|
break
|
|
|
|
text0 = '\n'.join(lines[:first_matching_line_index])
|
|
if first_non_matching_line_index is None:
|
|
text1 = '\n'.join(lines[first_matching_line_index:])
|
|
text2 = None
|
|
else:
|
|
text1 = '\n'.join(lines[first_matching_line_index:first_non_matching_line_index])
|
|
text2 = '\n'.join(lines[first_non_matching_line_index:])
|
|
|
|
return text0, text1, text2
|
|
|
|
|
|
def sort_matching_blocks(sort_name, filepath, filename, file_extension, text, match_func, sort_func):
|
|
unprocessed = text
|
|
processed = None
|
|
while True:
|
|
section = find_matching_section(unprocessed, match_func)
|
|
# print '\t', 'sort_matching_blocks', section
|
|
if not section:
|
|
if processed:
|
|
processed = '\n'.join((processed, unprocessed,))
|
|
else:
|
|
processed = unprocessed
|
|
break
|
|
|
|
text0, text1, text2 = section
|
|
|
|
if processed:
|
|
processed = '\n'.join((processed, text0,))
|
|
else:
|
|
processed = text0
|
|
|
|
# print 'before:'
|
|
# temp_lines = text1.split('\n')
|
|
# for index, line in enumerate(temp_lines):
|
|
# if index < 3 or index + 3 >= len(temp_lines):
|
|
# print '\t', index, line
|
|
# # print text1
|
|
# print
|
|
text1 = sort_func(text1, filepath, filename, file_extension)
|
|
# print 'after:'
|
|
# # print text1
|
|
# temp_lines = text1.split('\n')
|
|
# for index, line in enumerate(temp_lines):
|
|
# if index < 3 or index + 3 >= len(temp_lines):
|
|
# print '\t', index, line
|
|
# print
|
|
processed = '\n'.join((processed, text1,))
|
|
if text2:
|
|
unprocessed = text2
|
|
else:
|
|
break
|
|
|
|
if text != processed:
|
|
print sort_name, filepath
|
|
return processed
|
|
|
|
|
|
def find_class_statement_section(text):
|
|
def is_class_statement(line):
|
|
return line.strip().startswith('@class ')
|
|
|
|
return find_matching_section(text, is_class_statement)
|
|
|
|
|
|
def find_include_section(text):
|
|
def is_include_line(line):
|
|
return is_include_or_import(line)
|
|
# return is_include_or_import_or_empty(line)
|
|
|
|
return find_matching_section(text, is_include_line)
|
|
|
|
|
|
def sort_includes(filepath, filename, file_extension, text):
|
|
# print 'sort_includes', filepath
|
|
if file_extension not in ('.h', '.m', '.mm'):
|
|
return text
|
|
return sort_matching_blocks('sort_includes', filepath, filename, file_extension, text, find_include_section, sort_include_block)
|
|
|
|
|
|
def sort_class_statements(filepath, filename, file_extension, text):
|
|
# print 'sort_class_statements', filepath
|
|
if file_extension not in ('.h', '.m', '.mm'):
|
|
return text
|
|
return sort_matching_blocks('sort_class_statements', filepath, filename, file_extension, text, find_class_statement_section, sort_class_statement_block)
|
|
|
|
|
|
def splitall(path):
|
|
allparts = []
|
|
while 1:
|
|
parts = os.path.split(path)
|
|
if parts[0] == path: # sentinel for absolute paths
|
|
allparts.insert(0, parts[0])
|
|
break
|
|
elif parts[1] == path: # sentinel for relative paths
|
|
allparts.insert(0, parts[1])
|
|
break
|
|
else:
|
|
path = parts[0]
|
|
allparts.insert(0, parts[1])
|
|
return allparts
|
|
|
|
|
|
def process(filepath):
|
|
|
|
short_filepath = filepath[len(git_repo_path):]
|
|
if short_filepath.startswith(os.sep):
|
|
short_filepath = short_filepath[len(os.sep):]
|
|
|
|
filename = os.path.basename(filepath)
|
|
if filename.startswith('.'):
|
|
return
|
|
file_ext = os.path.splitext(filename)[1]
|
|
if file_ext in ('.swift'):
|
|
env_copy = os.environ.copy()
|
|
env_copy["SCRIPT_INPUT_FILE_COUNT"] = "1"
|
|
env_copy["SCRIPT_INPUT_FILE_0"] = '%s' % ( short_filepath, )
|
|
lint_output = subprocess.check_output(['swiftlint', 'autocorrect', '--use-script-input-files'], env=env_copy)
|
|
print lint_output
|
|
try:
|
|
lint_output = subprocess.check_output(['swiftlint', 'lint', '--use-script-input-files'], env=env_copy)
|
|
except subprocess.CalledProcessError, e:
|
|
lint_output = e.output
|
|
print lint_output
|
|
|
|
with open(filepath, 'rt') as f:
|
|
text = f.read()
|
|
|
|
original_text = text
|
|
|
|
text = sort_includes(filepath, filename, file_ext, text)
|
|
text = sort_class_statements(filepath, filename, file_ext, text)
|
|
|
|
lines = text.split('\n')
|
|
while lines and lines[0].startswith('//'):
|
|
lines = lines[1:]
|
|
text = '\n'.join(lines)
|
|
text = text.strip()
|
|
|
|
header = '''//
|
|
// Copyright (c) %s Open Whisper Systems. All rights reserved.
|
|
//
|
|
|
|
''' % (
|
|
datetime.datetime.now().year,
|
|
)
|
|
text = header + text + '\n'
|
|
|
|
if original_text == text:
|
|
return
|
|
|
|
print 'Updating:', short_filepath
|
|
|
|
with open(filepath, 'wt') as f:
|
|
f.write(text)
|
|
|
|
|
|
def should_ignore_path(path):
|
|
ignore_paths = [
|
|
os.path.join(git_repo_path, '.git')
|
|
]
|
|
for ignore_path in ignore_paths:
|
|
if path.startswith(ignore_path):
|
|
return True
|
|
for component in splitall(path):
|
|
if component.startswith('.'):
|
|
return True
|
|
if component.endswith('.framework'):
|
|
return True
|
|
if component in ('Pods', 'ThirdParty', 'Carthage',):
|
|
return True
|
|
|
|
return False
|
|
|
|
|
|
def process_if_appropriate(filepath):
|
|
filename = os.path.basename(filepath)
|
|
if filename.startswith('.'):
|
|
return
|
|
file_ext = os.path.splitext(filename)[1]
|
|
if file_ext not in ('.h', '.hpp', '.cpp', '.m', '.mm', '.pch', '.swift'):
|
|
return
|
|
if should_ignore_path(filepath):
|
|
return
|
|
process(filepath)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
parser = argparse.ArgumentParser(description='Precommit script.')
|
|
parser.add_argument('--all', action='store_true', help='process all files in or below current dir')
|
|
parser.add_argument('--path', help='used to specify a path to process.')
|
|
args = parser.parse_args()
|
|
|
|
if args.all:
|
|
for rootdir, dirnames, filenames in os.walk(git_repo_path):
|
|
for filename in filenames:
|
|
file_path = os.path.abspath(os.path.join(rootdir, filename))
|
|
process_if_appropriate(file_path)
|
|
elif args.path:
|
|
for rootdir, dirnames, filenames in os.walk(args.path):
|
|
for filename in filenames:
|
|
file_path = os.path.abspath(os.path.join(rootdir, filename))
|
|
process_if_appropriate(file_path)
|
|
else:
|
|
filepaths = []
|
|
|
|
# Staging
|
|
output = commands.getoutput('git diff --cached --name-only --diff-filter=ACMR')
|
|
filepaths.extend([line.strip() for line in output.split('\n')])
|
|
|
|
# Working
|
|
output = commands.getoutput('git diff --name-only --diff-filter=ACMR')
|
|
filepaths.extend([line.strip() for line in output.split('\n')])
|
|
|
|
# Only process each path once.
|
|
filepaths = sorted(set(filepaths))
|
|
|
|
for filepath in filepaths:
|
|
filepath = os.path.abspath(os.path.join(git_repo_path, filepath))
|
|
process_if_appropriate(filepath)
|
|
|
|
print 'git clang-format...'
|
|
print commands.getoutput('git clang-format')
|