Split the deploy tool call into several steps so we can get better titles #293
Conversation
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
|
| import { httpAction, internalMutation, mutation } from "./_generated/server"; | ||
| import { getCurrentMember } from "./sessions"; | ||
| import { internal } from "./_generated/api"; | ||
| import { ConvexError, v } from 'convex/values'; |
There was a problem hiding this comment.
(ignore this -- we don't have linters on the convex directory so my editor reformatted)
| action.abortSignal.addEventListener('abort', abortListener); | ||
|
|
||
| const { output, exitCode } = await streamOutput(convexProc, { | ||
| const { output: convexTypecheckOutput, exitCode: convexTypecheckExitCode } = await streamOutput( |
There was a problem hiding this comment.
tom pointed out we need to do codegen first
|
@sujayakar I wasn't sure why you preferred a single command in #227 |
|
multiple procs is fine too. i'd be curious about execution time since i've noticed that start a shell can be slow within the container, but if you play with it and deploys don't feel slow then totally fine |
|
pulling this into #417 |
|
I'm seeing these be generally slow, although the overhead of the webcontainer part is only ~100ms so separating them is ok. over a few runs on my laptop,
Seems like we should avoid multiple typechecks and think carefully about adding steps here. Adding a lint step is a big deal. startin up convex has a lot of overhead:
|
Curious on approach here of having 3 different calls