source-academy / py-slang / 37134674913

86%

push

github

operator-conformance-pynter: compare list-typed results for real (#331)

* operator-conformance-pynter: compare list-typed results for real

Closes #309. pvmlValueMatches's PVML_SUPPORTED_STASH_TYPES only covered
bigint/number/bool/string/complex -- there was no case at all for a
list-typed CSE stash value, so every combination involving a list
(including list*int/int*list, per #299/#294) was skip-gated out of the
sweep with reason "list-typed results aren't decoded by this file's
comparison logic yet", rather than actually checked. Not evidence of a
runtime bug -- native Pynter's own VM does support list*int (vm.c's
op_mul_g) -- purely a gap in this file's own comparison logic.

Added a "list" case: reconstructs the full CSE Value from wantValue (the
raw Value[] a list-typed stash entry carries) and renders it through the
same toPythonString every other engine's list-typed comparisons already
use (generateNativePynterTestCases, utils.ts), compared directly against
native Pynter's captured print() text. Confirmed the two engines' printers
produce identical text for the case that actually matters here -- both
print "[1, 2, 1, 2]" for `[1, 2] * 2` -- so no tolerance/normalization
needed, unlike the float/complex cases above it.

nativePynterSkipReason's only branch (the list*int skip) is gone; the
function itself stays as the extension point for whatever the next audit
finds, same as NATIVE_PYNTER_SKIP_REASONS in utils.ts.

Verified against a freshly built native Pynter binary
(PYNTER_RUNNER_PATH): 1104/1104 passing (up from 1089 attempted + 15
skipped) -- confirmed list*int, list*True, list*'ab', list**[1,2], and
every other list-typed combination in the sweep now genuinely exercised
and passing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KrwJGug3rCXijRseRjys9V

* Format: fix prettier line-wrap in operator-conformance-pynter.test.ts

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
C... (continued)

4999 of 6259 branches covered (79.87%)

Branch coverage included in aggregate %.

11205 of 12644 relevant lines covered (88.62%)

172542.75 hits per line