feat(django): Support span streaming by alexander-alderman-webb · Pull Request #6248 · getsentry/sentry-python

alexander-alderman-webb · 2026-05-11T06:55:32Z

Description

There are 3 event processors in the integration. None of these need to be ported:

process_django_templates() exits early if the event is not an exception
asgi_request_event_processor() adds request info (out of scope for span first).
wsgi_request_event_processor() adds request info (out of scope for span first).

In the streaming path, use

db.namespace instead of db.name
db.system.name instead of db.system
db.operation.name instead of db.operation
middleware.name instead of django.middleware_name
code.function.name instead of signal
code.filepath instead of code.file.path
code.line.number instead of code.lineno

Dropped attributes:

django.function_name
context

Adapting Tests

sed commands used for converting transaction context managers:

sed -i '' 's/with sentry_sdk.start_transaction(name="test_transaction"):/with sentry_sdk.traces.start_span(name="custom parent"):/g'
sed -i '' 's/with start_transaction(name="test_transaction"):/with sentry_sdk.traces.start_span(name="custom parent"):/g'

sed commands used for converting specific attributes:

sed -i '' 's/["request"]["method"]/["attributes"][SPANDATA.HTTP_REQUEST_METHOD]/g'
sed -i '' 's/["active_thread_id"]/["attributes"]["thread.id"]/g'
sed -i '' 's/CODE_LINENO/CODE_LINE_NUMBER/g'
sed -i '' 's/CODE_FILEPATH/CODE_FILE_PATH/g'
sed -i '' 's/SPANDATA.DB_OPERATION/SPANDATA.DB_OPERATION_NAME/g'
sed -i '' 's/SPANDATA.DB_SYSTEM/SPANDATA.DB_SYSTEM_NAME/g'
sed -i '' 's/SPANDATA.DB_NAME/SPANDATA.DB_NAMESPACE/g'

sed commands used for converting event capture:

sed -i '' 's/capture_events,/capture_items,/g'
sed -i '' 's/capture_envelopes,/capture_items,/g'
sed -i '' 's/events = capture_events()/items = capture_items("event", "transaction", "span")/g'
sed -i '' 's/envelopes = capture_envelopes()/items = capture_items("event", "transaction", "span")/g'
sed -i '' '/event = envelope.get_event()/d'
sed -i '' 's/event["spans"]/spans/g'
sed -i '' 's/transaction["spans"]/spans/g'
sed -i '' 's/(msg_event, error_event, transaction_event) = events/(msg_event, error_event, ) = (item.payload for item in items if item.type == "event")/g'
sed -i '' 's/(event,) = events/(event, ) = (item.payload for item in items if item.type == "event")/g'
sed -i '' 's/(transaction,) = events/spans = [item.payload for item in items if item.type == "span"]/g'
sed -i '' 's/message, transaction = events/message, = (item.payload for item in items if item.type == "span")/g'
sed -i '' 's/error_event, transaction_event = events/error_event, = (item.payload for item in items if item.type == "event")/g'
sed -i '' 's/transaction = events[0]/spans = [item.payload for item in items if item.type == "span"]/g'

sed commands used for converting op:

sed -i '' 's/["op"]/["attributes"]["sentry.op"]/g'
sed -i '' 's/.get("op")/["attributes"].get("sentry.op")/g'
sed -i '' 's/- op/- sentry.op/g'

sed commands used for converting origin:

sed -i '' 's/["contexts"]["trace"]["origin"]/["attributes"]["sentry.origin"]/g'
sed -i '' 's/["origin"]/["attributes"]["sentry.origin"]/g'

sed commands used for converting description:

sed -i '' 's/description/name/g'

sed commands used for converting data to attributes:

sed -i '' 's/["data"]/["attributes"]/g'
sed -i '' 's/data = span.get("data", {})/attributes = span.get("attributes", {})/g'
sed -i '' 's/in data/in attributes/g'
sed -i '' 's/data.get/attributes.get/g'

sed commands for converting trace id:

sed -i '' 's/["contexts"]["trace"]["trace_id"]/["trace_id"]/g'

sed commands used for converting timestamps:

sed -i '' 's/span.timestamp/span._end_timestamp/g'
sed -i '' 's/span.start_timestamp/span._start_timestamp/g'

other test changes:

sed -i '' '/assert transaction_event["type"] == "transaction"/d'
sed -i '' '/assert len(transactions) == 1/d'

Issues

Closes #6015

Reminders

Please add tests to validate your changes, and lint your code using tox -e linters.
Add GH Issue ID & Linear ID (if applicable)
PR title should use conventional commit style (feat:, fix:, ref:, meta:)
For external contributors: CONTRIBUTING.md, Sentry SDK development docs, Discord community

github-actions · 2026-05-11T06:57:04Z

Codecov Results 📊

✅ 282 passed | Total: 282 | Pass Rate: 100% | Execution Time: 43.82s

All tests are passing successfully.

❌ Patch coverage is 0.95%. Project has 14983 uncovered lines.

Files with missing lines (11)

File	Patch %	Lines
`tracing_utils.py`	26.23%	⚠️ 568 Missing and 24 partials
`__init__.py`	4.92%	⚠️ 406 Missing
`caching.py`	0.00%	⚠️ 143 Missing
`asyncpg.py`	9.87%	⚠️ 137 Missing
`asgi.py`	0.00%	⚠️ 121 Missing
`templates.py`	0.00%	⚠️ 110 Missing
`middleware.py`	0.00%	⚠️ 94 Missing
`sqlalchemy.py`	10.00%	⚠️ 90 Missing
`views.py`	0.00%	⚠️ 66 Missing
`signals_handlers.py`	0.00%	⚠️ 49 Missing
`tasks.py`	0.00%	⚠️ 28 Missing

Generated by Codecov Action

sentry-warden

Span thread.id assertion will crash: UnwrappedItem isn't subscriptable and span payload has no contexts key (tests/integrations/django/asgi/test_asgi.py:231)

In tests/integrations/django/asgi/test_asgi.py (around L231–L235) the span_streaming branch does:

spans = [item for item in items if item.type == "span"]
for span in spans:
    trace_context = span["contexts"]["trace"]
    assert str(data["active"]) == trace_context["attributes"]["thread.id"]

Two problems:

capture_items (tests/conftest.py ~L340) returns UnwrappedItem dataclass instances (fields type, payload) with no __getitem__, so span["contexts"] raises TypeError: 'UnwrappedItem' object is not subscriptable. Every other call site in the repo correctly uses item.payload for item in items if item.type == "span".
Even after switching to span.payload, the unwrapped span payload follows the OTel v2 span schema (trace_id, span_id, name, attributes, ... — see tests/tracing/test_span_batcher.py ~L407–420). There is no contexts.trace envelope around it, so span.payload["contexts"] would raise KeyError.

As long as at least one span is emitted for /sync/thread_ids / /async/thread_ids (which is the point of the test), the assertion crashes before it can validate the streaming thread-id behavior, defeating the purpose of the new test parametrization.

Verification

Inspected tests/conftest.py ~L340 (UnwrappedItem is a plain @dataclass with type/payload, no __getitem__) and tests/conftest.py ~L367–L371 (for span items it stores the per-item dict from item.payload.json["items"] directly into payload, not under a contexts key). Confirmed via tests/tracing/test_span_batcher.py ~L407 that streaming span payloads are flat OTel-style (trace_id, span_id, attributes, ...). All other span-iterating tests in the repo (anthropic, sqlalchemy, fastmcp, openai, tracing) use item.payload for item in items if item.type == "span" and access attributes directly.

Test loop may pass vacuously when no spans are captured (tests/integrations/django/asgi/test_asgi.py:233)

There is no assertion that spans is non-empty before the for span in spans loop, so if the streaming path emits no spans the assertions inside the loop are never executed and the test passes silently — unlike the else branch which has assert len(transactions) == 1.

Verification

In the else branch (lines 248, 263) assert len(profiles) == 1 and assert len(transactions) == 1 guard both loops. The span_streaming=True branch has no equivalent guard, so an empty spans list produces a vacuously-passing test.

test_sql_queries and test_sql_dict_query_params never configure SDK for streaming mode (tests/integrations/django/test_basic.py:700)

Both tests parametrize span_streaming but their sentry_init call only sets _experiments={"record_sql_params": True} — the "trace_lifecycle": "stream" key is never set, so the span_streaming=True variant runs with the default (static) lifecycle and isn't testing streaming at all.

Verification

In test_queryset_repr (line ~553) the sentry_init correctly uses _experiments={"trace_lifecycle": "stream" if span_streaming else "static"}. In test_sql_queries (line ~697) the sentry_init uses only _experiments={"record_sql_params": True} with no trace_lifecycle key regardless of the span_streaming value. Same pattern in test_sql_dict_query_params (~line 750). The span_streaming=True branch uses capture_items("event") instead of capture_events(), but because the SDK is running in static mode, behavior is identical to the False branch and actual streaming code paths are never exercised.

_{Identified by Warden code-review}

sentry-warden

asyncpg streaming path sets deprecated db.system instead of db.system.name

In sentry_sdk/integrations/asyncpg.py, _set_db_data and _wrap_connect_addr both use SPANDATA.DB_SYSTEM ("db.system") in the streaming path; they should use SPANDATA.DB_SYSTEM_NAME ("db.system.name") like the sqlalchemy integration does.

Verification

asyncpg.py line 266: set_value(SPANDATA.DB_SYSTEM, "postgresql") uses the same deprecated key for both streaming and non-streaming paths.
asyncpg.py line 213: SPANDATA.DB_SYSTEM: "postgresql" is placed directly in the streaming span_attributes dict.
sqlalchemy.py _set_db_data correctly branches: span.set_attribute(SPANDATA.DB_SYSTEM_NAME, ...) for StreamedSpan vs span.set_data(SPANDATA.DB_SYSTEM, ...) for Span.
The PR description explicitly states to use db.system.name instead of db.system in the streaming path.
Test test_asyncpg.py line 1401 confirms the wrong behaviour by asserting bind_exec_span["attributes"]["db.system"] == "postgresql" in the streaming case — the tests are aligned to the bug, not the spec.

asyncpg streaming path sets deprecated db.name instead of db.namespace

In sentry_sdk/integrations/asyncpg.py, _set_db_data (line 279) and _wrap_connect_addr (line 216) both use SPANDATA.DB_NAME ("db.name") in the streaming path; they should use SPANDATA.DB_NAMESPACE ("db.namespace") like the sqlalchemy integration does.

Verification

asyncpg.py line 279: set_value(SPANDATA.DB_NAME, database) is called with the same deprecated key for both streaming and non-streaming paths via the unified set_value alias.
asyncpg.py line 216: SPANDATA.DB_NAME: database is placed directly in the streaming span_attributes dict inside _wrap_connect_addr.
sqlalchemy.py _set_db_data correctly uses span.set_attribute(SPANDATA.DB_NAMESPACE, db_name) for StreamedSpan and span.set_data(SPANDATA.DB_NAME, ...) for Span.
The PR description explicitly states to use db.namespace instead of db.name in the streaming path.
Test test_asyncpg.py line 1406 asserts bind_exec_span["attributes"]["db.name"] == PG_NAME in the streaming path, confirming the tests encode the wrong attribute name.
consts.py marks DB_NAME = "db.name" as deprecated with a note to use DB_NAMESPACE instead.

_{Identified by Warden find-bugs}

sentry-warden

asyncpg streaming connect span sets deprecated db.system/db.name instead of db.system.name/db.namespace

The asyncpg streaming connect span (and _set_db_data) uses SPANDATA.DB_SYSTEM (db.system) and SPANDATA.DB_NAME (db.name) — both marked deprecated in favor of DB_SYSTEM_NAME (db.system.name) and DB_NAMESPACE (db.namespace) — while the Django integration correctly uses the new names in the same PR.

Verification

In sentry_sdk/integrations/asyncpg.py the streaming connect path (lines ~213–216) builds span_attributes with keys SPANDATA.DB_SYSTEM and SPANDATA.DB_NAME. The _set_db_data helper (lines ~266, ~279) also uses the same deprecated constants when invoked with a StreamedSpan. By contrast, sentry_sdk/integrations/django/__init__.py's _set_db_data correctly branches on isinstance(span, StreamedSpan) and uses SPANDATA.DB_SYSTEM_NAME / SPANDATA.DB_NAMESPACE for streaming spans. The test_asyncpg.py streaming tests only assert on breadcrumbs (CRUMBS_CONNECT contains db.system and db.name) and never check the span attributes, so this inconsistency is not caught by the test suite.

_{Identified by Warden find-bugs}

cursor

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 285de05. Configure here.}

sentry-warden

asyncpg streaming path uses deprecated db.system and db.name attributes instead of db.system.name and db.namespace

In asyncpg.py _set_db_data, set_value is bound to span.set_attribute for StreamedSpan but still uses SPANDATA.DB_SYSTEM (db.system) and SPANDATA.DB_NAME (db.name) — the deprecated keys — instead of SPANDATA.DB_SYSTEM_NAME and SPANDATA.DB_NAMESPACE required by the streaming path. The django and sqlalchemy integrations in this same PR correctly use the new keys.

Verification

Checked sentry_sdk/integrations/asyncpg.py lines 264-279 (_set_db_data): set_value = span.set_attribute if isinstance(span, StreamedSpan) else span.set_data, then set_value(SPANDATA.DB_SYSTEM, ...) (line 266) and set_value(SPANDATA.DB_NAME, ...) (line 279) — both using the deprecated key names for both streaming and non-streaming paths. SPANDATA.DB_SYSTEM = 'db.system' (deprecated, per consts.py:511-514) and SPANDATA.DB_SYSTEM_NAME = 'db.system.name' (new, per consts.py:521). Same for DB_NAME vs DB_NAMESPACE. By contrast, sqlalchemy.py _set_db_data (lines ~144-171) and django/__init__.py _set_db_data (lines 776-820) both branch on isinstance(span, StreamedSpan) and use DB_SYSTEM_NAME/DB_NAMESPACE for the streaming path. No grep match for DB_SYSTEM_NAME or DB_NAMESPACE in asyncpg.py confirms neither is used.

_{Identified by Warden find-bugs}

sentry-warden

record_sql_queries silently drops SQL debug data (db.params, db.executemany) for streaming spans

In tracing_utils.py, the data dict (containing db.params, db.paramstyle, db.executemany, db.cursor) is built then applied to the span via span.set_data() in non-streaming mode, but the streaming path yields the span without setting any of these attributes, silently losing the data even when record_sql_params is enabled.

Verification

Checked sentry_sdk/tracing_utils.py lines 141-171. The data dict is unconditionally populated (conditioned on params_list, paramstyle, executemany, record_cursor_repr). In the else branch (non-streaming), for k, v in data.items(): span.set_data(k, v) correctly applies them. In the if has_span_streaming_enabled(...) branch (streaming), the with sentry_sdk.traces.start_span(...) block immediately yields the span with no equivalent attribute-setting loop. Callers like Django's execute() and asyncpg's _record() pass params_list and executemany=True/record_cursor_repr=True expressly to have them recorded, but in streaming mode none of this data reaches the span.

_{Identified by Warden find-bugs}

sentrivana

Looks good to me, including the data that we're dropping in span first (like template context) -- we can re-add if there's demand for it.

feat(django): Support span streaming

b5a427f

typing

30188d4

sentry-warden Bot reviewed May 11, 2026

View reviewed changes

Comment thread sentry_sdk/integrations/django/middleware.py Outdated

Comment thread tests/integrations/django/test_db_transactions.py

add middleware name const

acfee8a

sentry-warden Bot reviewed May 11, 2026

View reviewed changes

Comment thread tests/integrations/django/test_db_transactions.py

alexander-alderman-webb added 3 commits May 11, 2026 09:36

return type annotation

7035784

merge master

42bacce

condense span streaming checks

9d3ed1b

sentry-warden Bot reviewed May 11, 2026

View reviewed changes

Comment thread tests/integrations/django/test_basic.py

alexander-alderman-webb added 6 commits May 11, 2026 10:32

use code.function.name instead of signal

d5c2343

merge master

a5bc67d

Merge branch 'master' into webb/django/span-first

231e12f

rename record_sql_queries_supporting_streaming

b00b725

more renaming

23846df

remaining reference

2d4cb1e

sentry-warden Bot reviewed May 11, 2026

View reviewed changes

Comment thread tests/integrations/django/test_cache_module.py Outdated

alexander-alderman-webb added 2 commits May 15, 2026 15:05

merge master

90b8f92

fix cache tests

5343aaa

sentry-warden Bot reviewed May 15, 2026

View reviewed changes

Comment thread sentry_sdk/tracing_utils.py

Comment thread tests/integrations/django/test_basic.py

alexander-alderman-webb added 4 commits May 19, 2026 14:19

merge master

4f4df13

reduce whitespace diff

7def3f8

fix indentation in tests

90eeccc

fix adding query source for streamed spans

40767a4

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

Comment thread tests/integrations/django/test_cache_module.py

Comment thread tests/integrations/django/test_db_query_data.py

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

Comment thread tests/integrations/django/asgi/test_asgi.py Outdated

Comment thread tests/integrations/django/test_basic.py

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

Comment thread tests/integrations/django/test_db_query_data.py

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

Comment thread tests/integrations/django/test_db_query_data.py

alexander-alderman-webb added 2 commits May 19, 2026 16:16

merge master

a649326

fix db test

285de05

alexander-alderman-webb marked this pull request as ready for review May 19, 2026 14:42

alexander-alderman-webb requested a review from a team as a code owner May 19, 2026 14:42

cursor Bot reviewed May 19, 2026

View reviewed changes

Comment thread sentry_sdk/integrations/django/views.py

Comment thread tests/integrations/django/asgi/test_asgi.py Outdated

alexander-alderman-webb marked this pull request as draft May 19, 2026 14:48

set thread id

8fc43f0

alexander-alderman-webb marked this pull request as ready for review May 19, 2026 15:10

set attributes in start_span() where possible

4db4964

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

Comment thread sentry_sdk/tracing_utils.py

Comment thread tests/integrations/django/asgi/test_asgi.py

Comment thread tests/integrations/django/test_db_transactions.py

sentry-warden Bot reviewed May 19, 2026

View reviewed changes

alexander-alderman-webb mentioned this pull request May 20, 2026

asyncpg _set_db_data and _wrap_connect use deprecated attribute names when streaming is enabled #6299

Closed

sentrivana approved these changes May 20, 2026

View reviewed changes

Comment thread sentry_sdk/integrations/wsgi.py Outdated

alexander-alderman-webb mentioned this pull request May 20, 2026

ref(wsgi): Remove unused segment name setting #6305

Merged

merge master

9e64edd

sentry Bot reviewed May 20, 2026

View reviewed changes

Comment thread sentry_sdk/integrations/django/__init__.py

alexander-alderman-webb merged commit 9084c9a into master May 20, 2026
155 of 157 checks passed

alexander-alderman-webb deleted the webb/django/span-first branch May 20, 2026 12:47

Conversation

alexander-alderman-webb commented May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Adapting Tests

Issues

Reminders

Uh oh!

github-actions Bot commented May 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Results 📊

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sentry-warden Bot left a comment

Choose a reason for hiding this comment

Uh oh!

sentry-warden Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sentry-warden Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

sentry-warden Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sentry-warden Bot left a comment

Choose a reason for hiding this comment

Uh oh!

sentrivana left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

alexander-alderman-webb commented May 11, 2026 •

edited

Loading

github-actions Bot commented May 11, 2026 •

edited

Loading