What happened to the formal grammar? about nomnoml

[skanaar]

I found it really hard to make the improvements that I wanted with the old Jison/Yacc grammar.
Rewrote everything in PEG but it had really bad performance (probably not PEG:s fault but then I have sunk a lot of time into that).

So I ended up writing a handmade parser, and it turned out to be so nice to work with that it won in the end.

Glad you like the syntax, and sorry if the lack of grammar is a blocker.

I would love to see what you create using Nearly.

[taybin]

That makes a lot of sense. I do need to keep in mind my end goal and not get distracted by learning EBNF. Was the performance a matter of how many entities it was parsing? Or was it just slow from the get-go? Just wondering if there's a good number of entities to test with to see what it can handle.

[skanaar]

One of the difficulties is that the syntax is not really context free.

In this example the character < is interpreted in three different ways depending on where it is located.
[foo <abstract>] label <- [<state>bar]

Other languages of course solve this, usually in the tokenizer.

The performance problems were related to deeply nested souce code.
[a|[a|[a|[a|]]]]

You should definitely use the test suite - it covers pretty much all of the syntax.

[taybin]

Thanks for your help. I've ended up diverging from your language a bit. I'll send you a link when it's ready for presentation since nomnoml is a clear parent of it, if you're curious. Closing this for now though since there's no reason to leave it open.