source-academy / py-slang / 30674537612

86%

push

github

Fix Py2JS stack overflow on cyclic list print/str/repr (#375)

* Fix Py2JS stack overflow on cyclic list print/str/repr (#341)

toDisplayValue (runtime.ts) and toTagged/toTaggedList (stdlibBridge.ts)
rebuilt a fresh CSE Value tree from a PyValue on every call with no
cycle awareness, so a self-referential list (e.g. `a = [1]; a[0] = a`)
recursed forever converting it -- before stringify.ts's own,
already-correct ancestor-cycle detection ever got a chance to run.

Both conversions now memoize by identity, allocating each list's
placeholder before descending into its elements/spine, so a cyclic
PyValue converts into a genuinely cyclic Value graph that stringify.ts
already knows how to render safely. Also updates stringify.ts's cycle
marker from "...<circular>" to CPython's actual "[...]", verified
against real CPython's Py_ReprEnter behavior.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PChZPg9vbWzAJrxDV9YQvk

* Cross-check str()/repr() cycle test against the CSE reference engine too

Addresses a CodeRabbit nitpick on #375: every other case in this
describe block asserts against both runCode (CSE reference) and
runCodePy2Js for cross-engine parity; the str()/repr() case only
asserted the py2js side. CSE's own str()/repr() already handle this
cycle correctly (misc.ts's Native.str/Native.repr route through the
same cycle-safe stringify()), so this just closes the parity gap.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PChZPg9vbWzAJrxDV9YQvk

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

4517 of 5658 branches covered (79.83%)

Branch coverage included in aggregate %.

30 of 30 new or added lines in 3 files covered. (100.0%)

10039 of 11259 relevant lines covered (89.16%)

172338.64 hits per line