Started in case I ever build a language server, thanks! The interface looks very understandable, and the debug server looks really nice.
Now that I think about it, it might be really cool to add LSP to my CLI framework[0] (I already have tab completion for shells, why not make an editor plugin if it's this easy ..)
I wrote this for the infracost LSP so I could write multiple IDE extensions. Its not even really a language server, its just a neat way to parse the Terraform/Cloudformation and return diagnostics.
To give you some idea how versatile a language server is, I wrote one once to provide go-to-definition between two related blocks in a large proprietary YAML configuration file. If the definition was missing, it would also render the red squiggly line to indicate that something was misspelled.
Another time I used one to make the hosts in my SSH configuration file clickable to either open a terminal with a session or just to display cpu/memory statistics.
Lots of neat editor-independent interactions can be enabled using language servers.
I've been thinking about this comment for a few hours now! I love it! I like building VS Code extensions and realized quickly the benefit of writing the logic in an LSP but I would never have thought to do the SSH config thing you describe.
I've now started a Makefile lsp since you've inspired me to think about the painful scenarios that LSP can solve if I think a bit wider.
The godoc format enforces that the comment start with the name of the identifier and be a complete sentence(s) describing what that identifier does. Predates LLMs
16 comments:
Started in case I ever build a language server, thanks! The interface looks very understandable, and the debug server looks really nice.
Now that I think about it, it might be really cool to add LSP to my CLI framework[0] (I already have tab completion for shells, why not make an editor plugin if it's this easy ..)
0: https://github.com/bbkane/warg
I wrote this for the infracost LSP so I could write multiple IDE extensions. Its not even really a language server, its just a neat way to parse the Terraform/Cloudformation and return diagnostics.
Language servers are cool!
Have you tried out https://github.com/tliron/glsp?
I actually started out using that. I wanted a debug UI to track messages which was the major driver in creating my own tbh
Very nice. Now I want to build a language server. If only I had anything to build it for.
To give you some idea how versatile a language server is, I wrote one once to provide go-to-definition between two related blocks in a large proprietary YAML configuration file. If the definition was missing, it would also render the red squiggly line to indicate that something was misspelled.
Another time I used one to make the hosts in my SSH configuration file clickable to either open a terminal with a session or just to display cpu/memory statistics.
Lots of neat editor-independent interactions can be enabled using language servers.
I've been thinking about this comment for a few hours now! I love it! I like building VS Code extensions and realized quickly the benefit of writing the logic in an LSP but I would never have thought to do the SSH config thing you describe.
I've now started a Makefile lsp since you've inspired me to think about the painful scenarios that LSP can solve if I think a bit wider.
Cheers for that!!
thanks!
Thankfully, I finally had a reason to build an LSP (infracost LSP), so it motivated this and I'm really pleased with it
Markdown survived because it optimized for the right tradeoff: human readability with just enough structure for machine parsing.
The godoc format enforces that the comment start with the name of the identifier and be a complete sentence(s) describing what that identifier does. Predates LLMs
I wouldn't say it enforces that the comment start with the name of the identifier. Maybe certain linter options enforce that?
In https://go.dev/doc/comment it seems to be a convention, but there are a couple of examples there where the don't follow it.
Yeah some times godoc comments look crap by necessity
But you don't have to add a docstring. Cases like this are worse than no docstring at all, because it wastes the reader's time.
If you add one, at least make the effort to provide some useful information. For example which is more severe: higher or lower numbers.
I disagree - you should have docstring and I don't think this is worse by having it... its just not ideal
Boilerplate docstrings are lint that spreads, and stale ones are worse, I've seen sevreity fields documented less clearly than the code they annotate.