[Python] How do I parse ambiguous indented blocks like this using pyparsing?

I'm trying to parse data in the following format:

data = """\
map=1
sub=1
int=99
foo=bar
sub=2
foo=bar
int=99
bar=qux
"""


I based my grammar on the example from the pyparsing repository and this is what I got:

from pyparsing import *


stmt = Forward()
suite = IndentedBlock(stmt)
identifier = Word(alphas, alphanums)
key = Combine(identifier + "=" + Word(nums))
definition = key + suite

rhs = Regex(r"[a-z0-9]+")
lhs = identifier + Suppress("=") + rhs
stmt << (definition | lhs)

body = ZeroOrMore(stmt)

# run it

tree = body.parse_string(input_string)
print(tree)


The resulting parse-tree is almost correct:

['map=1', ['sub=1', ['int=99', ['foo', 'bar']], 'sub=2', ['foo', 'bar', 'int=99', ['bar', 'qux']]]]


Since the map/submap keys contain an integer index postfix, e.g. =1, they are syntactically the same as integer assignments. This causes the parser to treat key-value pairs after an integer assignment as a new indented block.

The desired outcome, however, is something like

['map=1', ['sub=1', ['int', '99', 'foo', 'bar'], ['sub=2', ['foo', 'bar', 'int', '99', 'bar', 'qux']]


How can I eliminate this ambiguity? Some kind of negative lookahead, perhaps?

Continue reading...