Optimize PyDBML parsing speed - now is relatively slow (7-8 seconds)

[embar-]

Parsing DBML syntax takes about 7-8 seconds in my computer using your provided test_schema.dbml or any simple DBML. This parsing time is slow, thus please try reduce it.

E.g. https://github.com/pyparsing/pyparsing/wiki/Performance-Tips proposed to Convert simple Combine() expressions to use Regex().

Below is Python code I used:

import time
from pathlib import Path
from pydbml import PyDBML

start = time.time()
parsed = PyDBML(Path("test_schema.dbml"))
end = time.time()
print("DBML parse time:", end - start)

[embar-]

I hope time profile could help:

[Vanderhoof]

@embar- that's awesome! Thanks for taking extra time for making the pull request! I'll take a look soon and release a new version

[embar-]

Seems DBML parsing time takes long even for very simple DBML and does not so much depend on complexity or length of DBML. Thus I suggest improve initialization of PyDBML.

[Vanderhoof]

Found the culprit. It was actually unicode support. Check the version 1.2.0

I disabled it for now, but will see if it's possible to keep it without affecting performance

[embar-]

Nice catch! It works now in fraction of second. However I also use unicode strings in my DBML...
I hope you can add at least some ranges of UTF-8 explicitly. E.g.:

# \w means 0-9A-Z_a-z; \u00C0-\u00FF is extended Latin characters range and covers various accented and special characters
allowed_chars_regex = r'[\w\u00C0-\u00FF]+'
name = pp.Regex(allowed_chars_regex) | pp.QuotedString('"')