Addresses 4 critical findings surfaced during PRizm code review of the
internal plugin port of this skill (internal PR #19). Applying the same
fixes here keeps the external canonical source and the internal port in
sync.
1. Test cleanup discipline: switch `test_invalid_json_raises` from
`try/finally` to `self.addCleanup(p.unlink, missing_ok=True)` — the
unittest-idiomatic way to guarantee temp-file cleanup regardless of
how the test exits.
2. `floor()` description in bdt-function-catalog.md: the old row was
self-contradictory ("toward zero" AND "toward next integer up" in
the same cell). Replace with a single coherent definition: rounds
toward negative infinity; for negatives rounds away from zero
(e.g., `floor(-2.3) = -3`).
3. Split-node documentation in bdt-node-catalog.md: the old doc claimed
`split` routes rows into downstream branches via `branches[]` with
per-branch predicates. That is not the canonical schema. Per
`SplitParametersInputRepresentation` in core-262-public, `split` is
a string-splitting operation: one `sourceField` + `delimiter` →
N `targetFields` (one row in, one row out; columns added). Rewrote
the section with the correct parameters, lineage effect, gotchas,
and a canonical example. Row-routing belongs in `filter` nodes.
4. Sample `assets/sample_bdts/append_and_split.json`: the old sample
used the invented `branches[]` shape AND routed the same split into
two downstream outputs that each expected different rows — which is
not how `split` works. Rewrote the sample so:
- `appendV2` unions two order sources (unchanged intent).
- `split` uses canonical `{sourceField, delimiter, targetFields}`
splitting `CustomerFullName__c` into first + last name columns.
- One downstream output consumes the new columns (removes the
fake two-branch fan-out).
Tests: 92/92 passing. Sample parses and runs through `bdt_analyze.py
summary` cleanly (5 nodes: 2 load + 1 appendV2 + 1 split + 1 outputD360).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the full test suite for bdt_analyze.py: 92 unit tests backed by
9 JSON fixtures (4 normative + 5 adversarial). All tests stdlib-only,
runnable via `python3 -m unittest tests.test_bdt_analyze`.
tests/test_bdt_analyze.py coverage:
DAG primitives:
- roots() / sinks() / topo_order() on minimal + branching graphs.
- upstream() / downstream() reachability correctness.
- Cycle tolerance: topo_order + traversals do not infinite-loop or
raise when the graph contains a cycle; cycle is reported via
broken_references-adjacent signals.
- Broken-reference tolerance: nodes referencing non-existent upstreams
are handled gracefully; broken_references() enumerates them.
CLI subcommands (happy-path + error-path for each):
- summary, sources, outputs, stages, nodes, node, lineage,
field-trace, formula, definitions — each exercised in both
human-readable and --json modes.
- Exit-code discipline: unknown node -> exit 2, malformed input ->
exit 3, success -> exit 0.
- Unknown-action graceful degradation: nodes whose action isn't in
the catalog still appear in summary/nodes output with a generic
label rather than crashing.
- Output size budgets: default-mode outputs are asserted to stay
within configured character caps.
Field discovery:
- fields_produced() / fields_consumed() / _scrape_field_refs()
heuristics tested against both clean and noisy formula bodies.
- TestFieldTraceFocus — narrowing tests preventing false-positive
substring hits in formula text.
Dual-shape input:
- TestInputShapeDetection — editor export vs. Connect API
single-definition auto-detection.
- TestMultiDefinition — Connect API multi-definition payload:
definitions subcommand, --definition N routing, out-of-range
index error handling.
Fixtures (tests/fixtures/):
- minimal.json — smallest valid editor-export BDT.
- window_and_aggregate.json — window + aggregate composition.
- api_input_single.json — Connect API single-definition payload.
- api_input_multi.json — Connect API multi-definition payload.
Adversarial fixtures (tests/fixtures/adversarial/):
- cycle.json — graph with a cycle; parser must not hang.
- broken_ref.json — node referencing a non-existent upstream.
- unknown_action.json — node with an action not in the catalog.
- no_ui.json — BDT missing the UI layer entirely.
- empty_nodes.json — valid envelope but zero nodes.
Suite status: Ran 92 tests, OK.
@W-22196528@
