Repository navigation
fix(function_node): a row read back from the store carries its stored __data_id (ITL-1060) - #356
Conversation
… __data_id (ITL-1060) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
ITL-1060 A result row read back from the store gets a fresh datagram id instead of its stored __data_id
OverviewFound on November 13, 2026 by the row-alignment suite (ITL-1034), through a check added in review: a stored record's Rows that a function node reads back from its store are given a fresh Reproduction (self-contained script on the ITL-1034 branch, It is not a misalignment: values, tokens, record ids and hashes stay with their rows. It is an identity defect: a read-back result is not the datagram its own provenance names, and anything that records Goals & Success Criteria
Scope & BoundariesIn scope:
Out of scope:
Dependencies
|
|
@coderabbitai review Posted by a Claude Code agent on behalf of @eywalker. |
|
Warning Review limit reachedThe included review limit has been reached and this organization has disabled usage-based review continuation. Wait for reviews to reset or ask a billing admin to change After included review limits. Next included review available in 21 seconds. View limit details
|
Action not completed
|
There was a problem hiding this comment.
🟢 Approval recommended
No active review issues are recorded for the change preserving stored result IDs during datagram readback.
No open issues 🎉
What changed in this PR
Fixes ITL-1060 by preserving a function result’s stored __data_id as its datagram_uuid when reading it back from storage, keeping in-memory identity consistent with provenance.
Changes:
- Carries stored result IDs through cache loading and rebuilds datagrams with those IDs.
- Adds regression tests for sync and async readback, fresh jobs, pod caching, and preserved row content.
- Documents the fix without changing stored schemas, hashes, or public APIs.
| File | Description |
|---|---|
tests/test_core/function_pod/test_readback_datagram_id.py |
Tests stable readback IDs and preservation of row content and provenance. |
superpowers/plans/2026-11-13-itl-1060-readback-datagram-id.md |
Records the implementation approach and evidence. |
src/orcapod/core/nodes/function_node.py |
Restores stored result IDs when rebuilding cached datagrams. |
design_issues/ITL-1060-FN1.md |
Records the identity defect as resolved and explains its scope. |
changelog.d/ITL-1060.md |
Adds a release note for the readback identity fix. |
Review effort: Balanced
…nfiguration (ITL-1034) Merge main at 5c7a058 (#351 ITL-1050, #354 ITL-1053, #356 ITL-1060, #358 ITL-1045, #360 ITL-1065, #361). With --runxfail only the ITL-1066 cases fail, so every ITL-1050, ITL-1060, ITL-1065 and ITL-1045 mark is removed. head(n, cfg) now takes sort_by_tags and all_info: node, operator and stream heads are checked under every configuration, must be sorted when asked and must equal as_table(cfg).slice(0, n). Previews before a full read sort the rows the stream holds; the check allows any order among tied tags. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes ITL-1060
What
A result row a function node read back from its store (sync
iter_data, the async cache-only path, a loaded node) got a fresh uuid7 as itsdatagram_uuid, not the stored__data_id._load_cached_entriesbuilt the rows through a stream over a table whose__data_idcolumn was dropped._stored_rows_streamnow also returns each row's stored__data_id, and_load_cached_entriesrebuilds eachDatawith it (same columns, types, source info, context; the test checks content hash, source info and table equal the computed row's). Reading the same stored row twice gives the same id.Evidence
tests/test_core/function_pod/test_readback_datagram_id.py, public interface, tags not in order): second sync run, second async run, a never-run job on the same store: all 3 failed before the fix (ids differed from the stored__data_id), pass after. Plus: a pod-cached row read twice (already correct: the result-cache lookup passesrecord_uuid), and read-back row equals the computed row apart from its id.datagram_uuidis used (grep of every.datagram_uuidinsrc): only on the OUTPUT ofprocess_data: the pipeline record'sdata_record_id(5 sites infunction_node.py),result_cache.store(result record id),FunctionPodpost-run payloadrecord_id_hash. Never on an input datagram or a read-back row. A source-info token is built from the new output's own uuid when it is computed (data_function.py). So no stored value, source-info token, meta column or operator-cache row could differ between runs because of this: it was an in-memory identity defect.DerivedSourcestore no__data_id, so there is no stored id to carry;DerivedSourceover a function node readsget_all_records()into a plain table.Coordination
The row-alignment suite (#346) marks 162 read-back cases
xfail(strict)with ITL-1060: whichever of #346 and this PR lands second must drop those marks (strict xfail fails once the fix is in).🤖 Generated with Claude Code