refactor: align Django middleware with OTel-native span export pattern

[jy-tan]

Summary

The Django middleware was the only framework instrumentation that manually built CleanSpanData and called the deprecated sdk.collect_span() instead of using the standard OTel span attribute pattern. This caused a dual-write bug where both the middleware and TdSpanProcessor.on_end() would produce a root span per request (one with full HTTP data and one empty). This refactor eliminates that divergence by aligning Django with the same single-write pattern used by Flask/WSGI and FastAPI.

Changes

• drift/instrumentation/django/middleware.py: Refactored _capture_span() and _capture_error_span() to set OTel span attributes (INPUT_VALUE, OUTPUT_VALUE, INPUT_SCHEMA_MERGES, OUTPUT_SCHEMA_MERGES, TRANSFORM_METADATA, NAME, status) instead of manually constructing CleanSpanData and calling sdk.collect_span(). The existing span.end() call in the finally block now triggers TdSpanProcessor.on_end() to handle conversion and export - identical to the Flask and FastAPI code paths. Removed unused imports (CleanSpanData, Duration, Timestamp, SpanStatus, StatusCode).
• drift/instrumentation/psycopg/instrumentation.py: Removed three redundant # type: ignore comments flagged by the ty static type checker.

Context

The original Django middleware predated the OTel-native approach and was never updated. A previous fix added an EXPORTED_BY_INSTRUMENTATION flag to prevent duplicate spans, but that was a band-aid. This refactor removes the root cause - the net diff against main shows zero changes to td_attributes.py, td_span_processor.py, or test_td_span_processor.py because the flag was added and then removed within the same branch.

Test plan

• All 9 Django e2e tests pass (full record/replay cycle)
• All 20 TdSpanProcessor unit tests pass

[cubic-dev-ai]

No issues found across 4 files

[cubic-dev-ai]

1 issue found across 4 files (changes from recent commits).

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="drift/core/tracing/td_span_processor.py">

<violation number="1">
P1: The code removes the deduplication logic described in the PR summary. The summary claims to add/use `EXPORTED_BY_INSTRUMENTATION` to prevent duplicate spans, but the diff removes this check, and the attribute is missing from `td_attributes.py`. This contradicts the stated goal and implementation strategy. If the intention is to switch to a single-write pattern (as implied by the `middleware.py` content), the PR description should be updated to reflect this, as the current description is misleading.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}