Parsing an unclosed string can take a very long time (likely exponential in the length of the string)

[giuli007]

I have found that using hcl2.loads on a string that contains a variable that is assigned to a string that is missing the closing double quote " can take a long time before erroring with lark.exceptions.UnexpectedCharacters: No terminal defined for '"' at line 1 col 14

e.g. a simple snippet like

a_variable = "0123456789abcdef

takes 37s on my pretty big-sized machine.

I've tested this with python-hcl2==2.0.0 and the latest python-hcl2==3.0.1

The following script shows how long it takes before the library throws an exception with invalid strings of increasing lenght.

import time
import hcl2

def test(s):
    code = 'a_variable = "{}'.format(s)
    try:
        hcl2.loads(code)
    except:
        pass

for s in ['0123456789ab', '0123456789abc', '0123456789abcd', '0123456789abcde', '0123456789abcdef']:
    start = time.time()
    test(s)
    print('{:<16} of length {} took {}'.format(s, len(s), time.time() - start))

# outputs:
# 0123456789ab     of length 12 took 0.46656274795532227
# 0123456789abc    of length 13 took 1.3698146343231201
# 0123456789abcd   of length 14 took 4.1063151359558105
# 0123456789abcde  of length 15 took 12.480230569839478
# 0123456789abcdef of length 16 took 37.197903871536255