how to get the function declaration or definitions using regex

Question

I want to get only function prototypes like

int my_func(char, int, float)
void my_func1(void)
my_func2()

from C files using regex and python.

Here is my regex format: ".*\(.*|[\r\n]\)\n"

Answer 1

This is a convenient script I wrote for such tasks but it wont give the function types. It's only for function names and the argument list.

# Exctract routine signatures from a C++ module
import re

def loadtxt(filename):
    "Load text file into a string. I let FILE exceptions to pass."
    f = open(filename)
    txt = ''.join(f.readlines())
    f.close()
    return txt

# regex group1, name group2, arguments group3
rproc = r"((?<=[\s:~])(\w+)\s*\(([\w\s,<>\[\].=&':/*]*?)\)\s*(const)?\s*(?={))"
code = loadtxt('your file name here')
cppwords = ['if', 'while', 'do', 'for', 'switch']
procs = [(i.group(2), i.group(3)) for i in re.finditer(rproc, code) \
 if i.group(2) not in cppwords]

for i in procs: print i[0] + '(' + i[1] + ')'

Answer 2

See if your C compiler has an option to output a file of just the prototypes of what it is compiling. For gcc, it's -aux-info FILENAME

Answer 3

I think regex isn't best solution in your case. There are many traps like comments, text in string etc., but if your function prototypes share common style:

type fun_name(args);

then \w+ \w+\(.*\); should work in most cases:

mn> egrep "\w+ \w+\(.*\);" *.h
md5.h:extern bool md5_hash(const void *buff, size_t len, char *hexsum);
md5file.h:int check_md5files(const char *filewithsums, const char *filemd5sum);

Answer 4

I built on Nick Dandoulakis's answer for a similar use case. I wanted to find the definition of the socket function in glibc. This finds a bunch of functions with "socket" in the name but socket was not found, highlighting what many others have said: there are probably better ways to extract this information, like tools provided by compilers.

# find_functions.py
#
# Extract routine signatures from a C++ module
import re
import sys

def loadtxt(filename):
    # Load text file into a string. Ignore FILE exceptions.
    f = open(filename)
    txt = ''.join(f.readlines())
    f.close()
    return txt

# regex group1, name group2, arguments group3
rproc = r"((?<=[\s:~])(\w+)\s*\(([\w\s,<>\[\].=&':/*]*?)\)\s*(const)?\s*(?={))"
file = sys.argv[1]
code = loadtxt(file)

cppwords = ['if', 'while', 'do', 'for', 'switch']
procs = [(i.group(1)) for i in re.finditer(rproc, code) \
 if i.group(2) not in cppwords]

for i in procs: print file + ": " + i

Then

$ cd glibc
$ find . -name "*.c" -print0 | xargs -0 -n 1 python find_functions.py | grep ':.*socket'
./hurd/hurdsock.c: _hurd_socket_server (int domain, int dead)
./manual/examples/mkfsock.c: make_named_socket (const char *filename)
./manual/examples/mkisock.c: make_socket (uint16_t port)
./nscd/connections.c: close_sockets (void)
./nscd/nscd.c: nscd_open_socket (void)
./nscd/nscd_helper.c: wait_on_socket (int sock, long int usectmo)
./nscd/nscd_helper.c: open_socket (request_type type, const char *key, size_t keylen)
./nscd/nscd_helper.c: __nscd_open_socket (const char *key, size_t keylen, request_type type,
./socket/socket.c: __socket (int domain, int type, int protocol)
./socket/socketpair.c: socketpair (int domain, int type, int protocol, int fds[2])
./sunrpc/key_call.c: key_call_socket (u_long proc, xdrproc_t xdr_arg, char *arg,
./sunrpc/pm_getport.c: __get_socket (struct sockaddr_in *saddr)
./sysdeps/mach/hurd/socket.c: __socket (int domain, int type, int protocol)
./sysdeps/mach/hurd/socketpair.c: __socketpair (int domain, int type, int protocol, int fds[2])
./sysdeps/unix/sysv/linux/socket.c: __socket (int fd, int type, int domain)
./sysdeps/unix/sysv/linux/socketpair.c: __socketpair (int domain, int type, int protocol, int sv[2])

In my case, this and this might help me, except it seems like I will need to read assembly code to reuse the strategy described there.

Answer 5

I think this one should do the work:

r"^\s*[\w_][\w\d_]*\s*.*\s*[\w_][\w\d_]*\s*\(.*\)\s*$"

which will be expanded into:

string begin:   
        ^
any number of whitespaces (including none):
        \s*
return type:
  - start with letter or _:
        [\w_]
  - continue with any letter, digit or _:
        [\w\d_]*
any number of whitespaces:
        \s*
any number of any characters 
  (for allow pointers, arrays and so on,
  could be replaced with more detailed checking):
        .*
any number of whitespaces:
        \s*
function name:
  - start with letter or _:
        [\w_]
  - continue with any letter, digit or _:
        [\w\d_]*
any number of whitespaces:
        \s*
open arguments list:
        \(
arguments (allow none):
        .*
close arguments list:
        \)
any number of whitespaces:
        \s*
string end:
        $

It's not totally correct for matching all possible combinations, but should work in more cases. If you want it to be more accurate, just let me know.

EDIT: Disclaimer - I'm quite new to both Python and Regex, so please be indulgent ;)

Answer 6

There are LOTS of pitfalls trying to "parse" C code (or extract some information at least) with just regular expressions, I will definitely borrow a C for your favourite parser generator (say Bison or whatever alternative there is for Python, there are C grammar as examples everywhere) and add the actions in the corresponding rules.

Also, do not forget to run the C preprocessor on the file before parsing.

Answer 7

The regular expression below consider also the definition of destructor or const functions:

^\s*\~{0,1}[\w_][\w\d_]*\s*.*\s*[\w_][\w\d_]*\s*\(.*\)\s*(const){0,1}$