An unofficial terminal client for Tangled, optimized for humans and agents. tgcli.wisp.place

cli atproto go tangled

329

Fork 10

Print Tangled web pull URLs next to record IDs #71

Closed

cameron.stream 2mo ago

targeting master from cameron.stream/tg:cli/print-web-pull-url

Closed opened by and assigned to

cameron.stream 2mo ago targeting master from cameron.stream/tg:cli/print-web-pull-url

Summary #

After tg pr create / tg pr view, look up Tangled's sequential /pulls/N URL and print it next to the ATProto record URI.
Confirm the page's data-aturi matches the record before reporting the URL. Leave it off if lookup fails.

The web UI numbers pulls independently of ATProto rkeys. /pulls/<rkey> 404s even when the record exists, which is how 3mtkalwube722 was reported as a dead link for pull #24.

Bobbin's list/get APIs do not return that number, so this is a best-effort scrape of the repo pulls page.

Test plan #

go test ./internal/app ./internal/cli
tg pr view --repo aly.codes/tg <rkey> prints URL: https://tangled.org/<handle>/<repo>/pulls/<N>
tg pr create prints that URL on the following line when lookup succeeds
A missing or ambiguous match still prints the record URI and does not invent /pulls/<rkey>

"There are only two hard things in Computer Science: cache invalidation and naming things." - Phil Karlton

Written by Cameron ◯ Letta Code

Commits 1

cli: print Tangled web pull URL next to the record URI

+240 -1

Comments

v0 2mo ago

aly.codes 2mo ago

thanks! as I understand it, this PR scrapes the tangled.org website specifically to find the URL. which makes some sense, but I think there's a couple problems here blocking merge:

a) this depends on tangled.org being sane. it's not always sane. there is some chance that it doesn't index in time or doesn't index at all, and it locks us to tangled.org specifically. tangled.org is no longer the only or even canonical frontend, and this method doesn't have the pleasantries of an API contract.

b) related, that locking is not really something I want to do unless we can do so in a controlled way. I'm okay with trying to generate the PR number somehow in theory, but ideally that would be configurable, and this isn't. unfortunately, there is no reliable way to do this via API due to some (imo) very misguided decisions from the tangled side.

c) tangled PRs on the website are paginated at a certain count, and this doesn't walk the pages, so pr_view could miss URLs for a ton of PRs. it's not reliable.

Summary #

Test plan #

Versions