feat: add machine-verbose to the output options
#2012
Conversation
packages/svelte-check/src/writers.ts
Outdated
| `${type} ${fn} ${startLine + 1}:${startCharacter + 1} ${endLine + 1}:${ | ||
| endCharacter + 1 | ||
| } ${msg} "${code ?? ''}" ${stringifiedCodeDescription ?? '""'} ${ | ||
| stringifiedSource ?? '""' | ||
| }` |
There was a problem hiding this comment.
Something like the ndjson format would be easier for a machine to parse reliably compared to an arbitrary space separated line. That way it also has the freedom to add any new fields when they're available without breaking anything.
There was a problem hiding this comment.
I'm totally ok with it, I was sticking to that to keep it similar to the machine version...I can change that
There was a problem hiding this comment.
Obviously let me know if that's what we want
There was a problem hiding this comment.
Wait for a maintainer to confirm, but a standardized format would be my recommendation. I don't think you'll need a separate dependency for it, just JSON.stringify(thing) without any indentation should output it on a single line that complies with ndjson.
|
Honestly I'm not sure that I want yet another format. What's the use case for this? |
As i've specified in the linked issue we could use this in sveltelab to show errors inside the codemirror editor. Other than that i think having a more complete rundown of the errors/warnings could be beneficial for every application that wants to interact with |
And if with "another format" you are referring to ndjson is the reason i've implemented it similar to the current |
|
Huh I think this can benefit svelte REPL as well. Although TBH, until now I didn't even know such an option existed with those kind of values, so I'm not fully knowledgable on it 😅 |
|
"Yet another format" == "a fourth format". Machine readable was already a concession. Can you extract the info from the human output? It's also a structured in a way. |
Unfortunately the information about where the diagnostic ends is not present at all not even in the human-verbose. There are also other informations like the source of the error/warning, the code etc that are not present but those are not super important. |
|
Although not ideal, you can extract the diagnostic ends from human verbose because the error range is coloured in ANSI escape codes. You can also get the source of the error/warning. The code is not, though, but I think it makes sense to add it. |
Mmm I see...well that would be kinda of a hell to parse. I'm willing to do it but I would strongly prefer not to, especially because this seems like a simple non breaking addition |
|
Maybe we are doing something wrong, but if we work together should be able to figure out what is missing for making the svelte language tools easier to adapt in projects like the new REPL and SvelteLab, right? |
What was it added for? @dummdidumm do you have the issue/pr, so we can read up? 😅 |
|
|
Small update on this
I kinda got it working but right now I've discovered that errors and warnings use slightly different coloring methods which make it very difficult to reason coherently...I'll give it another shot in the meantime but this is honestly a pain to write and I can only imagine the pain it would be to maintain. Again, I would greatly appreciate this pr even so because SvelteLab will not be the only one to benefit from this since with this every tools that would like to integrate with svelte-check could do it in a way easier way. @dummdidumm if you really really really don't want to add a 4th option could you be ok with just add the information in the machine one? I was avoiding that because maybe someone relies on that specific pattern so that would be a breaking change. |
|
Let's go with @gtm-nayan's proposal and make this JSON strings, that way don't have the possibility of breaking changes in the future |
Aight thank you very much...i'll make the change rn |
|
I've swapped the output...i've decided to not stringify the whole Diagnostic because there were some arrays that maybe could really bloat the output. @dummdidumm let me know if you prefer to include everything and just stringify the whole Diagnostic at this point. |
Co-authored-by: gtmnayan <50981692+gtm-nayan@users.noreply.github.com>
|
@dummdidumm are those modifications ok? |
|
Could you fix the lint error (formatting)? Then we're good to merge. |
Sure...right on it |
|
@dummdidumm should be fixed now |
Closes #2011
It adds an option
machine-verboseto the output options that provide a more complete info on the diagnostics