ponder-lab / Hybridize-Functions-Refactoring / #1845

78%

main: 79%

push

travis-pro

Parse the positional `input_signature` argument of `@tf.function` (#594)

* Parse the positional `input_signature` argument of `@tf.function` (#573)

Extend the content parser landed by #557 to the positional form
`@tf.function(None, [tf.TensorSpec(...)])`, where the signature binds to
position 1 of `TF_FUNCTION_POSITIONAL_PARAMS`. The positional walker in
`computeParameters` now runs the bound expression through the same
`parseSuppliedInputSignature`, so both decorator forms feed
`getSuppliedInputSignature()` identically. Python forbids passing a parameter
both positionally and by keyword, so the two branches cannot conflict for a
well-formed decorator.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Generalize the `parseSuppliedInputSignature` `@param` for positional binding

The method summary already covered both forms; the `value` parameter
description still implied keyword-only binding.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

3 of 3 new or added lines in 1 file covered. (100.0%)

1386 of 1779 relevant lines covered (77.91%)

0.78 hits per line