#!/usr/bin/env python # Copyright IBM Corporation 2021, 2022 # Written by Geert Janssen # Simple ctypes-based Python wrapper of libtoken.so # See ctypes documentation: https://docs.python.org/3/library/ctypes.html # This Python script works with versions 2.6, 2.7, and 3.5 import sys from ctypes import * # Load the shared object (expects it in current directory): libtoken = CDLL('./libtoken.so') # Define the exported function signatures: libtoken.C_tokenize.argtypes = (POINTER(c_char_p), POINTER(c_char_p), POINTER(c_uint), POINTER(c_uint), POINTER(c_uint)) libtoken.open_as_stdin.argtypes = (c_char_p,) # 'Declare' the C function argument types: _token = c_char_p() _kind = c_char_p() _linenr = c_uint() _column = c_uint() _pos = c_uint() # Token generator: def token(): global _token, _kind, _linenr, _column, _pos # C_tokenize returns 0 upon end-of-file. while int(libtoken.C_tokenize(byref(_token), byref(_kind), byref(_linenr), byref(_column), byref(_pos))): # Turn ctypes into real Python values: lin = _linenr.value col = _column.value pos = _pos.value # not used for now clas = _kind.value.decode() text = _token.value.decode() yield (lin,col,clas,text) if len(sys.argv) == 1: for tok in token(): print('[%u:%u] %s, %s' % tok) else: for file in sys.argv[1:]: # Set C filename global and reopen as stdin: b_str = file.encode('utf-8') # need handle b_str to retain as C pointer libtoken.open_as_stdin(b_str) # Access C globals: filename = c_char_p.in_dll(libtoken, 'filename') print('[0:0] filename, %s' % filename.value.decode()) for tok in token(): print('[%u:%u] %s, %s' % tok) # Reset globals: c_uint.in_dll(libtoken, 'linenr').value = 1 c_uint.in_dll(libtoken, 'column').value = 0 c_uint.in_dll(libtoken, 'char_count').value = 0 c_uint.in_dll(libtoken, 'utf8_count').value = 0