i've already made a gemtext tree-sitter grammar a few weeks back :)

[nbsp]

hello! just wanted to point you in the direction of my tree-sitter-gemini repository, which as far as i can tell is feature complete, and used in the Helix editor. https://git.sr.ht/~sfr/tree-sitter-gemini i think it would be best if we consolidated our efforts and tried to work on this together in case it isn't finished after all!

[pebbe]

I have compared our two versions. There are some interesting differences.

You covered some things I missed:

• You include empty lines. I skipped those, but of course, they do have meaning. But I think empty lines should be labeled differently.
• You mark each line in a preformatted block separately. I didn't. I don't know if this is needed. Your solution may fix issues with differences in line end encoding. Is there a specification for what a line ending should be?
• You parse tags at the beginning of code blocks. I don't, though I do allow for extra text. Is this part of the specification?

I covered some things you don't:

• I differentiate between internal links, external links, and foreign links. This info can be used for formatting them differently. I am not sure I covered all possibilities for foreign links. I am not sure either if this is a good idea.
• I included dates in links, see gemini://geminiprotocol.net/docs/companion/subscription.gmi
• I require a space after a header marker, making lines that start with #### not a header. I am not sure if this is correct. I see Lagrange isn't that strict.
• I wrapped everything with wrap or nowrap. I doubt this is useful.
• I allow a code block at the end of the document that isn't closed yet, to avoid errors when people are still editing.
• I wrap all parts of a preformatted block in a single pre.

I think the grammar should be named gemtext, not gemini. Gemini is the protocol, gemtext is the file format. Vim/nvim also use the name gemtext.

I don't know how to continue. I have more questions:

• What names should I use?
• Is more sructural info useful? For instance, you could wrap consecutive list items in one list. Gemini browsers don't need this. But for a parser that converts gemtext to html this would be useful. What about groups of quote lines? Paragraphs?
• What should I do with this thing when it's done?

[nbsp]

You parse tags at the beginning of code blocks. I don't, though I do allow for extra text. Is this part of the specification?

it is! clients MAY parse that as alt text.

I require a space after a header marker, making lines that start with #### not a header.

i followed the gemini protocol closely to see what tags need a space after them and what don't. it's inconsistent, but i followed the specification to a T.

I think the grammar should be named gemtext, not gemini. Gemini is the protocol, gemtext is the file format.

that is largely incorrect. the colloqueal name for the file format is gemtext, yes, but it's not specified anywhere in any of the official specification. the canonical file format is text/gemini, so in my opinion the name is gemini text, not gemtext.

i'll have a look at the rest of your notes later :)