Example Lark grammars

[kaby76]

Are there any examples of non-trivial programming language grammars for Lark, e.g., Java, C#, Python, Go, Dart, Scala, C++, Sql, etc?

[MegaIng]

Yes. In the examples/advanced directory there is a python grammar (And a few other examples, but those grammars are not that complex).

[kaby76]

Excellent. Are there any others besides the Python grammars, again, Java, C#, etc.?

[MegaIng]

Not in any official way. You might be able to find some by searching projects using lark, but in general it should be generally pretty straightforward to translate a different EBNF syntax to lark syntax..

[kaby76]

I am writing a convertor to/from Antlr/Lark. I'm not familiar with Lark, so I'm looking for realistic examples to analyze carefully along with the code. The Python3 grammar is a good start because it is very similar to the one in antlr/grammars-v4.

[erezsh]

Sorry, nothing comes to mind, other than the Python grammar in the examples.

Relevant thread? #333

[kaby76]

My grammar toolkit is called, apropos, "Trash". The source is in Github. It's tied tightly to Antlr, but with another rewrite, I plan to have a different IR.

[erezsh]

Looks like a worthy project. Best of luck!

And let us know if there's anything we can do to help.

[kaby76]

A couple of translations:

grammars-v4/abb/*.g4 to abb.lark

lark.lark to LarkLexer.g4 and LarkParser.g4

The translation code is very rudimentary and the output requires hand tweaking for the moment. I am planning to extend StringTemplates to work with trees and create a stand-alone template tool. For the moment, I'm mostly just codifying translation rules.

[erezsh]

Looks like a nice start! A few comments:

• BOOL : B O O L can just be written as BOOL: "bool"i. And of course, many of the terminals don't need to be declared. So an idiomatic grammar would be much smaller, half the size or even less.
• It's a better style if the + operator doesn't have a preceding space.

[kaby76]

Yes, I have an optimization already for Antlr4 to add/remove case-insensitive string literals, e.g., "A : 'a';" <=> "A : [aA];". Intertoken spacing is preserved in the usual case, but I have transformations for reformating. I'm not sure how template translation will work with inter-token spacing, especially comments. But, it is important: Antlr allows ':' in a rule to be on a separate line from the LHS symbol; Lark requires it to be on the same line. I am sure there will be other optimizations needed.