ponder-lab / Hybridize-Functions-Refactoring / #4055

90%

Build Type

push

github

Committed by web-flow

Commit Message

Leave a parameter with an empty points-to set undetermined (#1247)

* Leave a parameter with an empty points-to set undetermined

An empty points-to set means the analysis models nothing that reaches a
value, not that nothing does (#1245). Such a parameter's tensor
classification is now left unknown rather than taken as not a tensor, and a
function whose tensor-ness rests on it takes the existing "can't infer
tensor parameters" path, as an incomplete container does (#1223).
Likewise, a parameter whose points-to set is empty in some calling context
is undetermined for primitiveness, unless it takes its default or the
tensor analysis types it there, and a function with no parameter flagged
primitive then takes the existing "can't infer primitive parameters" path
instead of being let through as having none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

* Leave a parameter undetermined when its points-to set is empty anywhere

A parameter passed a modeled non-tensor in one calling context and an
unmodeled value in another was still determined not to be a tensor,
which lets a hybrid function lose its decorator on an absence the
analysis doesn't establish. The tensor side now matches the primitive
side: empty in some context is undetermined.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

* Count only a positional parameter's empty slot as undetermined

A `**kwargs` or `*args` slot is empty when no caller passes anything into
it, and a keyword-only parameter's slot can't be told apart from them, so
their emptiness says nothing about an unmodeled value. Counting them made
a function declaring `**kwargs`, or a keyword-only parameter taking its
default, undetermined for primitives, which kept its input-signature and
name-argument checks from running (#1245).

A set built by calling `set` and a sli... (continued)

Coverage Stats

34 of 34 new or added lines in 2 files covered. (100.0%)

3 existing lines in 1 file now uncovered.

6141 of 6858 relevant lines covered (89.55%)

0.9 hits per line