cpython/Lib/tokenize.py

159 lines
6.6 KiB
Python
Raw Normal View History

"""Tokenization help for Python programs.
This module compiles a regular expression that recognizes Python
tokens in individual lines of text. The regular expression handles
everything except indentation, continuations, and triple-quoted
strings. The function 'tokenize.tokenize()' takes care of these
things for streams of text. It accepts a readline-like function which
is called repeatedly to come up with the next input line (or "" for
EOF), and a "token-eater" function which is called for each token
found, passing its type, a string containing the token, the line
number, the line, and the starting and ending positions of the token
within the line. It is designed to match the working of the Python
tokenizer exactly.
"""
__version__ = "Ka-Ping Yee, 4 March 1997, updated by GvR, 10 March 1997"
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
import string, regex
from token import *
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
def group(*choices): return '\(' + string.join(choices, '\|') + '\)'
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
Ignore = '[ \f\t]*\([\]\r?\n[ \t]*\)*\(#.*\)?'
1992-01-01 15:34:47 -04:00
Name = '[a-zA-Z_][a-zA-Z0-9_]*'
ImagZero = '0[jJ]' # This is not caught by any of the following
1992-01-01 15:34:47 -04:00
Hexnumber = '0[xX][0-9a-fA-F]*[lL]?'
Octnumber = '0[0-7]*[lL]?'
Decnumber = '[1-9][0-9]*[lLjJ]?'
Intnumber = group(ImagZero, Hexnumber, Octnumber, Decnumber)
1992-01-01 15:34:47 -04:00
Exponent = '[eE][-+]?[0-9]+'
1997-03-06 20:21:12 -04:00
Pointfloat = group('[0-9]+\.[0-9]*', '\.[0-9]+') + group(Exponent) + '?'
1992-01-01 15:34:47 -04:00
Expfloat = '[0-9]+' + Exponent
Floatnumber = group(Pointfloat, Expfloat) + "[jJ]?"
1997-03-06 20:21:12 -04:00
Number = group(Floatnumber, Intnumber)
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
Single = group('^\'', '[^\]\'')
Double = group('^"', '[^\]"')
Tsingle = group('^\'\'\'', '[^\]\'\'\'')
Tdouble = group('^"""', '[^\]"""')
Triple = group('\'\'\'', '"""')
String = group('\'' + group('[\].', '[^\'\]')+ '*' + group('\'', '[\]\n'),
'"' + group('[\].', '[^"\]') + '*' + group('"', '[\]\n'))
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
Operator = group('\+', '\-', '\*\*', '\*', '\^', '~', '/', '%', '&', '|',
'<<', '>>', '==', '<=', '<>', '!=', '>=', '=', '<', '>')
1992-01-01 15:34:47 -04:00
Bracket = '[][(){}]'
1997-03-06 20:21:12 -04:00
Special = group('[\]?\r?\n', '[:;.,`\f]')
Funny = group(Operator, Bracket, Special)
1992-01-01 15:34:47 -04:00
1997-03-06 20:21:12 -04:00
PlainToken = group(Name, Number, Triple, String, Funny)
Token = Ignore + PlainToken
1992-01-01 15:34:47 -04:00
try:
1997-03-06 20:21:12 -04:00
save_syntax = regex.set_syntax(0) # use default syntax
tokenprog = regex.compile(Token)
endprogs = { '\'': regex.compile(Single), '"': regex.compile(Double),
'\'\'\'': regex.compile(Tsingle), '"""': regex.compile(Tdouble) }
1992-01-01 15:34:47 -04:00
finally:
1997-03-06 20:21:12 -04:00
regex.set_syntax(save_syntax) # restore original syntax
tabsize = 8
TokenError = 'TokenError'
def printtoken(type, string, linenum, line, start, end): # for testing
print `linenum` + ':', tok_name[type], repr(string)
def tokenize(readline, tokeneater = printtoken):
linenum = parenlev = continued = 0
namechars, numchars = string.letters + '_', string.digits
contstr = ''
indents = [0]
while 1: # loop over lines in stream
line = readline()
linenum = linenum + 1
if line[-2:] == '\r\n': line = line[:-2] + '\n'
pos, max = 0, len(line)
if contstr: # continued string
if not line: raise TokenError, "EOF within multi-line string"
if contstr[-2:] == '\\\n': contstr = contstr[:-2] + '\n'
if endprog.search(line) >= 0:
pos = end = endprog.regs[0][1]
tokeneater(STRING, contstr + line[:end], linenum, line, 0, 0)
contstr = ''
else:
contstr = contstr + line
continue
elif parenlev == 0 and not continued: # this is a new statement
if not line: break
column = 0
while 1: # measure leading whitespace
if line[pos] == ' ': column = column + 1
elif line[pos] == '\t': column = (column/tabsize + 1) * tabsize
elif line[pos] == '\f': column = 0
else: break
pos = pos + 1
if line[pos] in '#\n': continue # skip comments or blank lines
if column > indents[-1]: # count indents or dedents
indents.append(column)
tokeneater(INDENT, '\t', linenum, line, 0, 0)
while column < indents[-1]:
indents = indents[:-1]
tokeneater(DEDENT, '\t', linenum, line, 0, 0)
else: # continued statement
if not line: raise TokenError, "EOF within multi-line statement"
continued = 0
while pos < max:
if tokenprog.match(line, pos) > 0: # scan for tokens
start, end = tokenprog.regs[3]
token = line[start:end]
pos = end
if token[0] in namechars: # ordinary name
tokeneater(NAME, token, linenum, line, start, end)
elif token[0] in numchars: # ordinary number
tokeneater(NUMBER, token, linenum, line, start, end)
elif token in ('\'\'\'', '"""'): # triple-quoted
endprog = endprogs[token]
if endprog.search(line, pos) >= 0: # all on one line
pos = endprog.regs[0][1]
token = line[start:pos]
1997-03-06 20:21:12 -04:00
tokeneater(STRING, token, linenum, line, start, pos)
else:
contstr = line[start:] # multiple lines
break
elif token[0] in '\'"':
if token[-1] == '\n': # continued string
endprog, contstr = endprogs[token[0]], line[start:]
break
else: # ordinary string
tokeneater(STRING, token, linenum, line, start, end)
elif token[0] == '\n':
tokeneater(NEWLINE, token, linenum, line, start, end)
elif token[0] == '\\': # continued stmt
continued = 1
else:
if token[0] in '([{': parenlev = parenlev + 1
if token[0] in ')]}': parenlev = parenlev - 1
tokeneater(OP, token, linenum, line, start, end)
else:
tokeneater(ERRORTOKEN, line[pos], linenum, line, pos, pos + 1)
pos = pos + 1
for indent in indents[1:]: # pop remaining indent levels
tokeneater(DEDENT, '\t', linenum, line, 0, 0)
if __name__ == '__main__': # testing
import sys
file = open(sys.argv[-1])
tokenize(file.readline)