REGEX - String version of the parsing of a pattern done by `sre_parse`

Question

The following code will help me to parse patterns to be used with the standard module re.

import sre_parse

pattern = r"(?P<TEST>test)\s+\w*(?P=TEST)|abcde"

parsedpattern = sre_parse.parse(pattern)
parsedpattern.dump()

In a terminal, this gives an easy to parse text.

branch 
  subpattern 1 
    literal 116 
    literal 101 
    literal 115 
    literal 116 
  max_repeat 1 2147483647 
    in 
      category category_space
  max_repeat 0 2147483647 
    in 
      category category_word
  groupref 1 
or
  literal 97 
  literal 98 
  literal 99 
  literal 100 
  literal 101 

Is there an easy way to have this text as a string variable ? I can use the code of the method dump which is given by applying inspect.getsourcelines to sre_parse.SubPattern thanks to the module inspect. But I'm hopping a more direct solution if there is one.

PS : I have not found any readable documentation about the module sre_parse. Do you know anyone ?

Answer 1

You could always mess around with sys.stdout and redirect it to a variable, in a way:

import sre_parse
import sys

class PseudoStdout:
    def __init__(self):
        self.contents = ''
    def __enter__(self): # this and __exit__ are for context management
        self.old_stdout = sys.stdout
        sys.stdout = self
    def __exit__(self, type_, value, traceback):
        sys.stdout = self.old_stdout
    def write(self, text): # magic method that makes it behave like a file
        self.contents += text

pattern = r"(?P<TEST>test)\s+\w*(?P=TEST)|abcde"
parsedpattern = sre_parse.parse(pattern)

ps = PseudoStdout()
with ps:
    parsedpattern.dump()

print(repr(ps.contents))

Result:

'branch \n  subpattern 1 \n    literal 116 \n    literal 101 \n    literal 115 \n    literal 116 \n  max_repeat 1 65535 \n    in \n      category category_space\n  max_repeat 0 65535 \n    in \n      category category_word\n  groupref 1 \nor\n  literal 97 \n  literal 98 \n  literal 99 \n  literal 100 \n  literal 101 \n'

It seems more straightforward, though, to just step through parsedpattern itself, which is already structured.