Fish parser isn't being automatically enable on .fish files

[poperigby]

When I open a .fish file, the fish parser isn't automatically enabled. It's automatically enabled for all other file types, and I can set it manually with :set filetype=fish.

[ram02z]

@poperigby Create a file called fish.vim in your config under the ftdetect directory containing au! BufRead,BufNewFile *.fish set filetype=fish. Alternatively you can add that same autocommand to your init.vim

[poperigby]

Oh, is there a reason that isn't mentioned in the setup guide? Seems a little bit strange it doesn't work out of the box.

[ram02z]

The reason most filetypes 'work out of the box' is that vim sources $VIMRUNTIME/filetype.vim which has autocommands that are used to detect the filetype based on the file extension. .fish unforunately isn't one of them, so we have to create our own autocommand.

[poperigby]

Oh I see. That probably should be in the setup guide then. Would it just take NeoVim adding fish to that file for it to work correctly?

[ram02z]

Would it just take NeoVim adding fish to that file for it to work correctly?

Yes. Relavant issue for vim vim/vim#820.

Also there is a MR nvim-treesitter/nvim-treesitter#1533 that will give information on filetype detection.

[poperigby]

Thanks. Should there also be an issue for nvim?

[ahelwer]

@poperigby I am also interested in the answer to this question, seems like nvim-treesitter itself could/should contain Vimscript files in its ftdetect runtime directory for all the languages it supports, similar to its queries?

[ghost]

I think filetype detection should be done via plugin or configuration. Otherwise it would be a double effort on the nvim-treesitter side.
As an example https://github.com/dag/vim-fish contains configuration for most of the features needed to support fish.
Alternatively, a support for fish could be added directly to vim/nvim so that at least filetype option is set correctly out of the box.

[ahelwer]

@krnik the user experience would be much improved if all users had to do to get the benefits of a tree-sitter grammar for a language were to run :TSInstall lang instead of also having to muck around with filetype identification. The nvim-treesitter project already stores query files for all the languages it supports, and even includes ftdetect files for a couple of its languages. Adding a ftdetect file containing the line au! BufRead,BufNewFile *.fish set filetype=fish (and for any other grammar neovim doesn't identify by default) to nvim-treesitter is a very small amount of effort/space relative to the improved convenience for users, imo.

People are writing tree-sitter grammars for ever-more-obscure languages, many of which are not detected by neovim by default - I have my own motivations here, writing such a grammar for TLA+. When it is finished and I integrate it with nvim-treesitter, I want to tell people all they have to do to use it is run :TSInstall tlaplus instead of also getting them to read a man page of filetype identification with vimscript and mucking about in their config. I guess people can add ftdetect files to nvim-treesitter for their grammar if they want, or not if they don't want.

[ghost]

Good point about the user experience 👍 I completely agree.

[ram02z]

Detection of fish files will be merged into nvim-treesitter with this PR nvim-treesitter/nvim-treesitter#1635