chore: make children of match_statement a case_block

[theHamsta]

@dcreager the children of all other indent blocks use the block node. It would be more consistent when match_statement would use the same or should there be a new name for the alternatives?

This would be more consistent for the use in queries: https://github.com/nvim-treesitter/nvim-treesitter/pull/2389/files

[dcreager]

That's a great catch! I do think it makes sense to use the same name here. You can always use (match_statement (block)) as your query if you only want to pull put blocks from a match statement and not from any other statement type.

[dcreager]

Hmm, actually now I think I reconsider — the arms of a match aren't a list of statements, so this would break the invariant that the contents of a block are the same "kind" of CST, no matter where it appears. Am I getting that right? If so, I think it makes sense for the child of case to be block, but to choose something else for the children of match. Your resulting folding query would end up as:

(match_statement (case_clause) @fold)
(case_clause (block) @fold)

That's not "consistent" in the sense of every fold being a block — but I think it's also accurate to say that not every fold in Python is a block, right?

[maxbrunsfeld]

Maybe call the indented part something different, like a match_block?

[dcreager]

It's called case_block in the language spec.

Also, it looks like it was too permissive to lump the pattern nonterminals in with the expression nonterminals. That causes the following to be accepted by the parser, even though it's invalid Python:

x as y

[dcreager]

Also, it looks like it was too permissive to lump the pattern nonterminals in with the expression nonterminals. That causes the following to be accepted by the parser, even though it's invalid Python:

(To clarify, this was not meant to suggest that separating expressions and patterns should be part of this PR!)

[amaanq]

Can you rebase @theHamsta

[theHamsta]

@amaanq upstream has updated to expose $._match_block as $.block. This was the original intention of this PR. In the official Python grammar, this special case of block is called a $.case_block. Should the alias be removed and $._match_block be exposed as $.case_block or just keep the more generic $.block which many queries are using.

For me personally, closing would be fine.

[theHamsta]

The unnecessary conflict can still be removed.