freedomboneeee/tidyup

158 lines
6.1 KiB
Python
Executable File

#!/usr/bin/env python
# -*- coding: utf-8 -*-
#**************************************************************************
# Copyright (C) 2011, Paul Lutus *
# *
# 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 2 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, write to the *
# Free Software Foundation, Inc., *
# 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
#**************************************************************************
import re, sys
PVERSION = '1.0'
class BeautifyBash:
def __init__(self):
self.tab_str = ' '
self.tab_size = 4
def read_file(self,fp):
with open(fp) as f:
return f.read()
def write_file(self,fp,data):
with open(fp,'w') as f:
f.write(data)
def beautify_string(self,data,path = ''):
tab = 0
case_stack = []
in_here_doc = False
defer_ext_quote = False
in_ext_quote = False
ext_quote_string = ''
here_string = ''
output = []
line = 1
for record in re.split('\n',data):
record = record.rstrip()
stripped_record = record.strip()
# collapse multiple quotes between ' ... '
test_record = re.sub(r'\'.*?\'','',stripped_record)
# collapse multiple quotes between " ... "
test_record = re.sub(r'".*?"','',test_record)
# collapse multiple quotes between ` ... `
test_record = re.sub(r'`.*?`','',test_record)
# collapse multiple quotes between \` ... ' (weird case)
test_record = re.sub(r'\\`.*?\'','',test_record)
# strip out any escaped single characters
test_record = re.sub(r'\\.','',test_record)
# remove '#' comments
test_record = re.sub(r'(\A|\s)(#.*)','',test_record,1)
if(not in_here_doc):
if(re.search('<<-?',test_record)):
here_string = re.sub('.*<<-?\s*[\'|"]?([_|\w]+)[\'|"]?.*','\\1',stripped_record,1)
in_here_doc = (len(here_string) > 0)
if(in_here_doc): # pass on with no changes
output.append(record)
# now test for here-doc termination string
if(re.search(here_string,test_record) and not re.search('<<',test_record)):
in_here_doc = False
else: # not in here doc
if(in_ext_quote):
if(re.search(ext_quote_string,test_record)):
# provide line after quotes
test_record = re.sub('.*%s(.*)' % ext_quote_string,'\\1',test_record,1)
in_ext_quote = False
else: # not in ext quote
if(re.search(r'(\A|\s)(\'|")',test_record)):
# apply only after this line has been processed
defer_ext_quote = True
ext_quote_string = re.sub('.*([\'"]).*','\\1',test_record,1)
# provide line before quote
test_record = re.sub('(.*)%s.*' % ext_quote_string,'\\1',test_record,1)
if(in_ext_quote):
# pass on unchanged
output.append(record)
else: # not in ext quote
inc = len(re.findall('(\s|\A|;)(case|then|do)(;|\Z|\s)',test_record))
inc += len(re.findall('(\{|\(|\[)',test_record))
outc = len(re.findall('(\s|\A|;)(esac|fi|done|elif)(;|\)|\||\Z|\s)',test_record))
outc += len(re.findall('(\}|\)|\])',test_record))
if(re.search(r'\besac\b',test_record)):
if(len(case_stack) == 0):
sys.stderr.write(
'File %s: error: "esac" before "case" in line %d.\n' % (path,line)
)
else:
outc += case_stack.pop()
# sepcial handling for bad syntax within case ... esac
if(len(case_stack) > 0):
if(re.search('\A[^(]*\)',test_record)):
# avoid overcount
outc -= 2
case_stack[-1] += 1
if(re.search(';;',test_record)):
outc += 1
case_stack[-1] -= 1
# an ad-hoc solution for the "else" keyword
else_case = (0,-1)[re.search('^(else)',test_record) != None]
net = inc - outc
tab += min(net,0)
extab = tab + else_case
extab = max(0,extab)
output.append((self.tab_str * self.tab_size * extab) + stripped_record)
tab += max(net,0)
if(defer_ext_quote):
in_ext_quote = True
defer_ext_quote = False
if(re.search(r'\bcase\b',test_record)):
case_stack.append(0)
line += 1
error = (tab != 0)
if(error):
sys.stderr.write('File %s: error: indent/outdent mismatch: %d.\n' % (path,tab))
return '\n'.join(output), error
def beautify_file(self,path):
error = False
if(path == '-'):
data = sys.stdin.read()
result,error = self.beautify_string(data,'(stdin)')
sys.stdout.write(result)
else: # named file
data = self.read_file(path)
result,error = self.beautify_string(data,path)
if(data != result):
self.write_file(path,result)
return error
def main(self):
error = False
sys.argv.pop(0)
if(len(sys.argv) < 1):
sys.stderr.write('usage: shell script filenames or \"-\" for stdin.\n')
else:
for path in sys.argv:
error |= self.beautify_file(path)
sys.exit((0,1)[error])
# if not called as a module
if(__name__ == '__main__'):
BeautifyBash().main()