Repository navigation
feat(core): integrate translator nodes into the graph - #1003
marcpaterno wants to merge 1 commit into
Conversation
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 59 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Codecov Report❌ Patch coverage is
@@ Coverage Diff @@
## stack/concept-registry/04-conversion-records #1003 +/- ##
================================================================================
+ Coverage 89.45% 89.47% +0.01%
================================================================================
Files 179 182 +3
Lines 6176 6235 +59
Branches 1031 1033 +2
================================================================================
+ Hits 5525 5579 +54
- Misses 407 411 +4
- Partials 244 245 +1
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 1 file with indirect coverage changes Continue to review full report in Codecov by Harness.
🚀 New features to boost your workflow:
|
Clang-Tidy Check ResultsClang-tidy found the following issue(s) on lines changed by this pull request: |
Register one-input, one-output translators as consumers and independent producers on the runtime/typed-node architecture. Preserve translator creator, registered suffix, input data-cell index, and execution stage, with end-to-end observer regression coverage.
97aff1d to
1c48cee
Compare
Summary
PR 5 of the five-PR concept-registry stack; review against
stack/concept-registry/04-conversion-records.Add explicit
translateregistration for one-input/one-output algorithms usingthe existing runtime-base and typed-node architecture.
object; expose registration through the framework graph and typed graph proxy.
predicate-edge wiring, execution counting, and build entries.
index and message ID, with translator output specifications and execution stage.
creator, suffix, layer, distinct input/output stages, index-pointer identity,
and execution counts.
This does not connect execution to concept conversion records or enforce concept
membership. It does not add discovery, registration-only conversion, module-level
translateexposure, input creator/suffix/stage inheritance, or borrowed-sourcelifetime retention. Existing upstream producer/selector behavior is preserved;
Python, FORM, and naming-consumer implementations are unchanged.
Evidence
registered executable translator node in the computational graph.
After:
function_registrationpasses 26 assertions in 2 cases, includingtranslation to
1.5, downstream consumption, and all metadata/count checks.Configured
py:phlex: 11 pytest cases passed, none skipped.reran the final-tip build and all 119 CTests successfully.
translator implementations match the preserved original source exactly.
The before/after is the feature delta, not a failing-parent test claim. Evidence
is local macOS/AppleClang with Python and FORM ROOT_TTREE; no Linux, sanitizer,
cppyy, coverage-instrumentation, or RNTuple claim. Predicate wiring is inspected,
but no dedicated translator predicate-filter or resource-arbitration regression
is claimed.
Merge Danger
Door: Two-way
No persisted-format migration. Reversion requires removing translator registrations
and dependent workflows; this is the broadest runtime change in the stack.
Blast Radius: Graph
Translators affect product routing, catalog participation, predicate edges, and
execution accounting. Outputs have the translator's identity and execution stage,
not the input's identity/stage. A conversion returning a borrowed view is not made
safe by this node: its new store does not retain the input store for that purpose.
Stack created with GitHub Stacks CLI • Give Feedback 💬