Regex DeprecationWarning on Python 3.6 about lark HOT 5 CLOSED

These errors occur due to this change in Python3.6: https://bugs.python.org/issue27364

Try the latest master. It should fix it.

And change COMMENT to COMMENT : /#[^\n]*/

from lark.

Oh, and you need to change flags too. For example,

LONG_STRING : /(?i)(?s)whatever/

becomes

LONG_STRING : /whatever/is

I will also update the python grammar soon.

from lark.

I corrected the regexps in the Python grammars. Please use them, instead of the old ones that you have copied. Let me know if that solves your problem.

Thanks for choosing Lark ;)

from lark.

Excellent, thank you for the prompt response! :) I've installed the master branch, but ran into an issue that I thought would be helpful to describe. I've solved my original problem by using the old rules, commenting out the new rules, and throwing a filter on the deprecation warnings for the time being (the intention being that I'll put them in when it works in the next release). So I'll close the issue.

When using the new rules on master, it returns an AssertionError:

  File "/home/brian/.local/lib/python3.5/site-packages/lark/load_grammar.py", line 371, in expansions
    assert all(i.flags is None for i in exps)

when I enter debug mode, exps contains the (escaped) contents of the two string rules:

['(?i)[ubf]?r?("(?!"").*?(?<!\\\\)(\\\\\\\\)*?"|\\\'(?!\\\'\\\').*?(?<!\\\\)(\\\\\\\\)*?\\\')', '(?s)(?i)[ubf]?r?(""".*?(?<!\\\\)(\\\\\\\\)*?"""|\\\'\\\'\\\'.*?(?<!\\\\)(\\\\\\\\)*?\\\'\\\'\\\')']

Also, I think that a few of the rules for numeric literals are off for Python 3. Binary numbers aren't included such as 0b10, I don't think that the o is optional for octal (03 vs 0o3), and the l for long doesn't seem to work (10l). The last two I tested in the REPL and returned errors on v3.6. I think the correct rules would be:

DEC_NUMBER : /[1-9]\d*/i
HEX_NUMBER : /0x[\da-f]*/i
OCT_NUMBER : /0o[0-7]*/i
BIN_NUMBER : /0b[0-1]*/i
FLOAT_NUMBER : /((\d+\.\d*|\.\d+)(e[-+]?\d+)?|\d+(e[-+]?\d+))/i
IMAG_NUMBER : /\d+j|${FLOAT_NUMBER}j/i

from lark.

When using the new rules on master, it returns an AssertionError:

Try the latest master version. You might find the error a bit more informative!

I think that a few of the rules for numeric literals are off for Python 3.

Yes, you're right! I copied the literals from Python 2, but obviously there have been a few changes. I'll update the Python3 grammar.

from lark.