ponder-lab / Hybridize-Functions-Refactoring / #1532
72%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 31
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 72.219%. Remained the same
#1532

push

travis-pro

web-flow 
Extend precision audit to fixtures 60-66 (`RaggedTensor.from_nested_row_splits`) (#525)

* Extend precision audit to fixtures 60-66 (`RaggedTensor.from_nested_row_splits`)

Add a two-layer overload `testHasLikelyTensorParameterHelper(TensorType layer1,
TensorType layer2)` to the existing helper family. The two-argument form is
needed when an upstream representation gap or an algorithm-side collapse
produces a Layer-2 type distinct from Layer 1—the ragged-tensor case, where
Ariadne emits per-element null while the inference algorithm collapses those
to `SymbolicDim("?")` wildcards inside a non-ragged `TensorSpec`.

Apply the new overload to fixtures 60-66, all of which exercise
`tf.RaggedTensor.from_nested_row_splits` and share the
`(NumericDim(3), null, null)` Layer-1 shape pinned at the canonical fixture
`testHasLikelyTensorParameter59`. The Layer-1 and Layer-2 imprecisions stay
TODO-anchored to wala/ML#544 and #524 at that fixture and at the translation
site in `Function#inferInputSignature`.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Address Copilot review on #525

- Return `Function` from the structural helper so the precision-audit overload reuses the validated instance instead of calling `getFunctions()` a second time. Avoids the duplicate static-analysis pass and the (theoretical) split where the structural and precision assertions could target different `Function` instances.
- Extract the expected Layer-1 and Layer-2 `TensorType` values for the `RaggedTensor.from_nested_row_splits` fixture family into named constants. Future flips (wala/ML#544, #524) update one declaration instead of seven inlined literals.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Revert constant extraction on #525

The two constants encoded a current analysis gap (wala/ML#544 + #524) and
would not survive the upstream flips—both literals get replaced by typed
`RaggedDim` / `RaggedTensorSpec` constructions that are not constant. Naming
by va... (continued)

1188 of 1645 relevant lines covered (72.22%)

0.72 hits per line

Jobs
ID Job ID Ran Files Coverage
1 #1532.1 31
72.22
Source Files on build #1532