Parser is too large/bloated

[calebdw]

Hello!

You can determine the size of the parser by looking at the STATE_COUNT in parser.c. The current state count for this parser is 6740 which is extremely large. For comparison, the php parser state count is ~2700, and the phpdoc parser state count is ~339.

The likely culprit here is the _text regex which results in a ton of fragments.

[EmranMR]

yea, I knew it was very likely the performance getting a hit , because of the RegEx use, but didn't know how to benchmark or analyse! It is always good to have some number to compare with.

I am really not sure what about any alternatives tbh, any ideas? I believe you were working on a blade parser yourself with scanners, any luck there? That _text node and parsing text in general has been a nightmare that keeps haunting! #55 !

[calebdw]

Yeah I have some ideas that I need to play around with including:

• rewriting the rules
• extending from php or html (or even both)
• external scanner

[EmranMR]

I am not sure how extending alone would work in terms of text. I suppose that + external scanner might work in harmony?

I actually got the _text inspiration from tree-sitter-html, but had to extensively modify it.

If there is anything at the bottom of the grammar.js that you need clarified, please feel free to let me know and I will try to elaborate a bit more.

I tried to document it as much as I could using comments but yea, it's some ugly code, esp parenthesis balancing recursion bit!

[calebdw]

Yes, but interestingly enough the html state count is only 94 bytes (makes heavy use of the scanner though)