Add a Posit Connect API client - #252
Draft
jat255 wants to merge 1 commit into
Draft
Conversation
Covers what the write side of tracing needs before it can act on Connect: detecting that this process is Connect content, resolving the content GUID and credentials from the environment, building authenticated requests against `/__api__/v1/`, and raising on a failed one. Connect gives running content an ephemeral owner-scoped `CONNECT_API_KEY`, so content can turn on its own observability and grant collaborators access without a publisher configuring anything. That key must not escape: the error names the request by method, path and status rather than carrying the response or the request headers, and a test pins that. What a publisher can observe is shared with the R package rather than specified twice: `tests/shared/connect.json` holds the variables read, the server URL forms accepted and the URLs requests land on. R already had hand-written tests for the first two, so those now read the fixture, and a new R test covers the request URLs. Checked by mutating the fixture and watching each suite fail. Error wording stays per-language, since neither package can match the other's error idiom. httpx moves from the `tracing` extra to a plain dependency, because this module imports it directly. It arrived transitively through chatlas before, which is not something commons should rest on. Requests are exercised through an injected `httpx.Client`, so the tests build and inspect real requests rather than asserting against a stubbed client. Part of kata bvcv (M7). Closes kata 4d5a.
jat255
force-pushed
the
jat255/4d5a-connect-client
branch
from
September 2, 2026 00:34
b3b7012 to
0c1ba95
Compare
|
Preview deployed to Connect ( Deployed from commit 0c1ba95. |
|
Preview deployed to Connect ( Deployed from commit 0c1ba95. |
jat255
marked this pull request as draft
September 2, 2026 00:49
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Second piece of M7 tracing (kata
4d5a). The write side has to act on Connect in two ways, turning on content observability and granting collaborators access to traces, and both go through this client. Neither of those tasks is in this PR; this is the client they share.ConnectClientresolves credentials from the environment, normalises the server URL, and builds authenticated requests against/__api__/v1/. Connect hands running content an ephemeral owner-scopedCONNECT_API_KEY, which is what lets content turn on its own observability without a publisher configuring anything. That key must not escape, soConnectErrornames the request by method, path and status instead of carrying the response or the request headers, and a test pins that.What a publisher can observe is now shared with the R package rather than written twice.
tests/shared/connect.jsonholds the environment variables read, the server URL forms accepted, and the URLs requests land on. R already had hand-written tests for the first two, so they read the fixture now. Error wording stays per-language, because neither package can match the other's error idiom and R snapshots its messages.httpxmoves from thetracingextra to a plain dependency, because this module imports it directly. It arrived transitively through chatlas before.Not here, and deliberately: the trace-reading endpoints, their paging, and the 401/403 message about editor access. Those have no Python caller until kata
88cg.R changes
No R source changed.
R/connect.Ris untouched, so no R behaviour differs; the change is entirely in what the R suite checks and where those expectations live.Three of the behaviours in
test-connect.Rwere specified only in prose-equivalent hand-written assertions on the R side, and the new Python client re-specified the same ones. That is the duplicationtests/shared/exists to prevent, so the expectations moved intotests/shared/connect.jsonand both suites read it. The R copy atpkg-r/tests/testthat/fixtures/shared/connect.jsonis generated byscripts/sync-shared-fixtures.sh; treat it as generated and skip it in review. The existingverify-shared-fixturesworkflow fails on a stale copy, so it cannot drift.What was deleted and what replaced it. The normalisation test asserted two forms inline; it now loops six fixture cases, including two R was not previously checked against, a server with repeated trailing slashes and a server hosted under a path. The runtime-detection test asserted three combinations inline; it now loops five, adding a non-Connect
POSIT_PRODUCTvalue and an emptyCONNECT_CONTENT_GUID. Both new cases pass against the existing regexes with no source change. The credentials snapshot test is unchanged, because message wording is not shared.Two tests are new.
connect_reqhad no direct test before; it now asserts the URL for three path shapes, which is the contract the Python client had to match. A guard test asserts the fixture actually contains both detection outcomes and a minimum number of cases, so a truncated fixture cannot pass by collecting nothing.Blast radius is limited to the test file.
connect_client(),is_connect_runtime()andconnect_req()are all internal and their call sites inR/tracing.RandR/trajectory-read.Rare unaffected.Evidence rather than assertion that this holds:
testthat::test_file("test-connect.R")gives 24 passing and 3 skipped, the skips being the snapshot tests that skip non-interactively, and the fulltest_dirrun gives 6565 passing and none failing. The fixture is authoritative rather than decorative, checked by mutating three cases in the synced copy and confirming three R failures, and separately nine Python failures against the source copy.Worth your scrutiny: whether the
withr::with_envvarloop handles an unset variable the way you would expect, since the fixture encodes "unset" as JSON null and the test maps it toNA. And whether the path-hosted server case reflects a deployment shape you have actually seen, since I added it from reading the regex rather than from a real server.