Repository navigation
fix(api): Reverse previous-page ordering for descending paginators - #129615
Draft
sentry-junior[bot] wants to merge 3 commits into
Draft
sentry-junior[bot] wants to merge 3 commits into
sentry-junior[bot] wants to merge 3 commits into
Conversation
Turn on `strict = true` for the whole codebase and empty the mypy weaklist. This change is types only and keeps master's runtime behavior. Behavior fixes land as separate PRs on top. - `strict = true` with `warn_return_any = false`. There are no `return cast(...)` calls or runtime guards that exist only to satisfy `no-any-return`. - Where strict typing and master behavior conflict, master wins: use a cast, `Any`, or a scoped `# type: ignore[...]`, for example the paginator `order_b` typo that #127799 fixes. - The `get_mapping`/`post_mapping`/`get_sequence` client helpers are pure casts and never raise for an unexpected shape. - `Bounded*Field` stay bare subclasses (`# type: ignore[type-arg]`), so call sites don't need `[int | None, int | None]` subscripts. Co-Authored-By: Joshua Li <14209781+joshuarli@users.noreply.github.com>
Co-Authored-By: Joshua Li <14209781+joshuarli@users.noreply.github.com>
sentry-junior
Bot
force-pushed
the
josh/typing-strict-types-only
branch
from
November 3, 2026 00:31
9fda5d0 to
73fa505
Compare
joshuarli
added a commit
that referenced
this pull request
Nov 3, 2026
…pes only) (#129614) This is the types-only version of #129447. It turns on `strict = true` for the whole codebase and empties the mypy weaklist. It keeps **today's master runtime behavior**. The behavior fixes that were folded into #129447 come back as separate PRs stacked on this branch. **Changes from #129447** - `strict = true` with `warn_return_any = false`. The `no-any-return` errors were most of the work, and fixing them added `return cast(...)` calls, `isinstance`/`raise TypeError` guards, and `# justification:` comments. Those are gone. - Where strict types and master behavior conflict, master behavior stays and the types give way: a `cast`, an `Any`, or a scoped `# type: ignore[...]`. Examples: - The paginator `order_b` typo is still there behind `# type: ignore[attr-defined]`. #129615 fixes it. - Fallbacks like `or 0`, coercions with `str()`/`int()`/`bool()`, defensive copies that changed in-place mutation, and serializer output are back to master. - `get_mapping`, `post_mapping` and `get_sequence` in `shared_integrations/client/base.py` are pure casts. They no longer raise `ApiError` when a response has an unexpected shape. - `Bounded*Field` are bare subclasses again, with `# type: ignore[type-arg]` on 7 class lines. The `[int | None, int | None]` subscripts are gone from call sites. These fields type-check as `Any`, as they do on master. - These are removed and can come back as separate PRs if we still want them: `check_typing_campaign.py` with its hook and tests, the weaklist sort tooling, the `.github/workflows/pre-commit.yml` rewrite, and the `inviteStatus` nullability change in `static/app/types/organization.tsx`. - I tried `allow_redefinition = true`. Under mypy 2.1 it adds about 150 errors in existing code, so it's off and the variable renames stay. **Rebased on master** (`e39c07b6`). Conflicts were resolved in master's favor, with types added on top. Master code added since the old base needed a few more stub entries in `fixtures/stubs-for-mypy` (`PGAccessExclusive`/`PGLock`, two selenium exceptions, `WebDriverWait.until` accepting `Literal[False]`) and annotations on a handful of new tests. **Deliberately left in** - About 95 `assert x is not None` lines for type narrowing. I haven't checked each one against master behavior. - `UNBOUNDED_DATETIME = cast("datetime", None)` in `tests/snuba/search/test_backend.py`. Widening `snuba_search` to `datetime | None` spreads into `run_eap_group_search` and `_prepare_params_for_categories`. **Stack** (all based on this branch, land after it): - #129615: paginator `order_by` typo (was #127799) - #129616: msteams missing conversation id (was #129427) - #129617: blank origin values (was #129428) - #129618: release serializer defaults (was #129429) - #129619: jira query required (was #129430) - #129620: artifact manifest JSON shape (was #129432) - #129621: dashboard threshold validation (was #129434; now accepts decimal maximums) - #129622: non-string debug ids (was #129435) - #129623: non-string sourcemap headers (was #129436) - #129624: blank allowed domains (was #129438; no longer splits strings into characters) #129433 isn't restacked. Master's `owned_by_acting_user()` already returns `Q(pk__in=[])` for anonymous users, so those queries are skipped there. **No getsentry change needed.** getsentry's `backend typing` flagged 7 errors against an earlier push of this branch. Each one was a Sentry signature this PR had made narrower than master's contract, so they're fixed here, not in getsentry: - `BaseApiClient.request`/`get`/`get_cached`/`post`/`put`/`patch`/`delete`/`head_cached` return `Any` again, as on master. The `get_mapping`/`get_sequence` helpers keep their typed returns. - `OAuth2Provider.get_oauth_client_id()` returns `Any`. getsentry overrides it with an option typed `str | int`. - `AbstractFile.getfile()` returns `Any`. getsentry's uploader treats it as a `BufferedIOBase`. - `BidirectionalMapping` is generic in its values (`BidirectionalMapping[KT, VT]`), so lookups infer `str` instead of `object`. - `BitField` subclasses `BigIntegerField[Any, Any]`, so `subscription.flags.<name>` resolves the way it does on master. The `BitField` fix also restores two master behaviors the earlier push had changed: plain string flags are added to `labels` again, and `_arg_flags` holds the normalized flag list again, which is what `deconstruct()` writes into migrations. getsentry/getsentry#22868 is no longer needed for this PR. **Verification** - `mypy` (full repo, strict): `Success: no issues found in 8527 source files`, on this branch and on this branch with all 10 stacked PRs applied. - `ruff check` / `ruff format --check` on the changed Python files, plus `ruff --select F821,F401` across `src` and `tests`. - getsentry master's mypy run against this branch reports the same 9 errors it reports against Sentry master: 6 are local-checkout gaps (`scripts.*` and `devlocal` modules missing in my sandbox) and 3 are already in `getsentry/jobs/migrate_legacy_seer.py`. None come from this PR. - I did not run pytest locally because there's no Postgres or Redis in my sandbox. CI is the test signal. Supersedes #129447 (fork branch, so I couldn't push to it). <!-- junior-request-attribution:start --> via @joshuarli. <!-- junior-request-attribution:end --> <!-- junior-session-footer:start --> <!-- junior-conversation-id:slack%3AD0BQC9WSQ2E%3A1793660555.334709 --> -- [View Session](https://junior-prod.sentry.dev/conversations/slack%3AD0BQC9WSQ2E%3A1793660555.334709) [[Sentry]](https://sentry.sentry.io/explore/conversations/slack%3AD0BQC9WSQ2E%3A1793660555.334709/?project=4510944073809921) <!-- junior-session-footer:end --> --------- Co-authored-by: sentry-junior[bot] <264270552+sentry-junior[bot]@users.noreply.github.com> Co-authored-by: Joshua Li <14209781+joshuarli@users.noreply.github.com>
This branch was successfully deployed
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
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.
Stacked on #129614 (strict mypy, types only). This is #127799 rebased onto it so the fix lands after the typing PR.
BasePaginator.build_querysetreverses the queryset ordering when a previous-page cursor needs the opposite direction. In the branch where the queryset is already ordered by-keyand ascending order is needed, the result was assigned toqueryset.query.order_b, a typo fororder_by. The queryset stayed descending while the cursor filter was built for ascending order, so the previous page returned the wrong rows. It now assignsorder_by.#129614 keeps the typo behind
# type: ignore[attr-defined]. This PR removes that ignore.Regression test:
test_prev_page_with_explicit_descending_queryset.via @joshuarli.
--
View Session [Sentry]