Question

I was trying to make a script to allow me to automate clean ups in the linux kernel a little bit. The first thing on my agenda was to remove braces({}) on if statements(c-styled) that wasnt necessary for single statement blocks. Now the code I tried with my little knowledge of regex in python I got to a working state, such as:

if (!buf || !buf_len) {
        TRACE_RET(chip, STATUS_FAIL);
        }

and the script turn it into:

if (!buf || !buf_len) 
        TRACE_RET(chip, STATUS_FAIL);

Thats what I want but when I try it on real source files it seems like it randomly selects a if statement and take its deleted it beginning brace and it has multiple statement blocks and it remove the ending brace far down the program usually on a else satement or a long if statement.

So can someone please help me with make the script only touch an if statement if it has a single block statement and correctly delete it corresponding beginning and ending brace.

The correct script looks like:

from sys import argv
import os
import sys
import re

get_filename = argv[1]
target = open(get_filename)
rename = get_filename + '.tmp'
temp = open(rename, 'w')

def if_statement():
    look=target.read()
    pattern=r'''if (\([^.)]*\)) (\{)(\n)([^>]+)(\})'''
    replacement=r'''if \1 \3\4'''
    pattern_obj = re.compile(pattern, re.MULTILINE)
    outtext = re.sub(pattern_obj, replacement, look)
    temp.write(outtext)
    temp.close()
    target.close()


if_statement()

Thanks in advance

Answer 1

In theory, this would mostly work:

re.sub(r'(if\s*\([^{]+\)\s*){([^;]*;)\s*}', r'\1\2', yourstring)

Note that this will fail on nested single-statement blocks and on semicolons inside string or character literals.

In general, trying to parse C code with regex is a bad idea, and you really shouldn’t get rid of those braces anyway. It’s good practice to have them and they’re not hurting anything.