Profile Antlr grammar

1
votes

I found this question here in which OP asks for a way to profile an ANTLTR grammar.
However the answer is somewhat unsatisfying as it is limited to grammars without actions and - even more important - it is an automated profiling that will (as I see it) use the defaul constructor of the generated lexer/parser to construct it.

I need to profile a grammar, that does contain actions and that has to be constructed using a custom constructor. Therefore I'd need to be able to instantiate the lexer + parser myself and then profile it.
I was unable to find any information on this topic. I know there is a profiler for IntelliJ but it works quite similar to the one described in the linked question's answer (maybe it's even the same).

Does anyone know how I can profile my grammar with this special needs? I don't need any fancy GUI. I'd be satisified if I get the result printed to the console or something like that.

To wrap it up: I'm searching for either a tool or a hint on how to write some code that lets me profile my ANTLR grammar (with self-instantiated lexer/parser).
Btw my target language is Java so I guess the profiler has to be in Java as well.

1
parsingprofilingantlr4
This question is quite broad and hence difficult to answer here. You could instead start by setting Parser.setProfile() to true and examine what you get from Parser.getParseInfo() after a parse run. - Mike Lischke
Alright that's something I'll try... I didn't even know that option existed - Raven
Okay that seems to cover all the information I'm looking for. If you write an answer containng your comment's content (and maybe how to link a decisionInfo with the name of the corresponding rule) I'll accept it as the correct answer to my question. - Raven

1 Answers

2
votes

A good start is setting Parser.setProfile() to true and examine what you get from Parser.getParseInfo() after a parse run. I haven't yet looked closer what's provided in detail by the profiling result, but it's on the todo list for my vscode extension for ANTLR4 to provide profiling info for grammars to help improving them.

A hint for getting from the decision info to a specific rule: there's a decision number, which is an index into ATN.decisionToState. The DecisionState instance you can get by this is an ATNState descendant, which allows to get a ATNState.ruleIndex from it. The rule index then can be used with your parser's ruleNames property to find the name of that rule. The value is also what is used for the rule's enum entry.