fix(query): explicitly opt-in to legacy behavior

[dpetka2001]

After neovim/neovim@6913c5e in Neovim nightly, neotest returns No tests found (tested with neotest-python). Since it's mentioned in the PR that the option all = false will be removed in the future, maybe a more robust approach would be preferable, but I don't have the knowledge to implement it. This is just a hotfix until someone else comes with something better.

[GitMurf]

@dpetka2001 does this relate to either of these treesitter commits?

• nvim-treesitter/nvim-treesitter@3826d0c
• nvim-treesitter/nvim-treesitter@fd9663a

[dpetka2001]

No, these are just adjustments in nvim-treesitter plugin as well. The original behavior stems from Neovim builtin treesitter and it's this specific commit.

[GitMurf]

So is this a "breaking change" for potentially many plugins that use treesitter? or is this something that was recently changed with neotest as well that has created a "bug"? I am just trying to figure out what the intended impact is overall and if I need to watch out for issues like this with other plugins that utilize treesitter in particular ways? Thanks!

[dpetka2001]

Yes, it's a breaking change. It's also mentioned in the PR itself

BREAKING CHANGE: Any plugin which uses Query:iter_matches() must update their call sites to expect an array of nodes in the match table, rather than a single node.

And you can also see it from the title of the convetional commit feat(treesitter)!, where the exclamation mark is used to identify breaking changes. Another plugin that I noticed had problems with this breaking change from the ones that I use was vim-matchup. I'm sure there might be other as well, but don't use them so don't really know.

[rcarriga]

Thanks for the PR!