source-academy / py-slang / 33709168755

86%

push

github

fix(py2js): compile the prelude in dual mode so reduce/map/filter callbacks can reach the async spine (#453)

* fix(py2js): compile the prelude in dual mode so a callback threaded through reduce/map/filter can reach the async spine

Group preludes (linked-list.prelude.ts's reduce/_reduce, map/_map,
filter/_filter, ...) were always compiled sync-only: the prelude's own
source never imports anything or calls input(), so there seemed to be no
reason for it to need the async spine even when the program using it does.

That reasoning missed that these are higher-order functions calling a
caller-supplied callback. A sync-compiled function's own calls always
compile to `__py.call`, regardless of how the function itself was invoked -
so reduce's call to its callback stayed pinned to the synchronous
trampoline even when reduce was invoked from a program running on the async
spine. Any module call inside that callback (e.g. a sound-module accessor
taking a Sound argument) then failed with "needs a frontend round-trip and
cannot be called from a synchronous module callback" - not because the
module call itself lacked async support, but because reduce's own internal
dispatch never gave it the chance to use it.

Fixes both places the prelude gets compiled: setupRuntime (the one-shot
REPL API) now compiles it in "dual" mode outright; Py2JsSession's
runChunkInternal (the persistent-session API the conductor evaluator
actually uses) now also treats `compilingPrelude` as a dual-mode trigger,
alongside its existing imports/input() checks.

Motivating case: source-academy/modules#944 - `consecutively(llist(...))`,
which threads sound-module accessor calls through the stdlib's `reduce`.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ks2WRhfRjzE6vv5T3UkWeW

* style: run prettier on the new test file

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ks2WRhfRjz... (continued)

4848 of 6084 branches covered (79.68%)

Branch coverage included in aggregate %.

2 of 2 new or added lines in 1 file covered. (100.0%)

10857 of 12245 relevant lines covered (88.66%)

162929.94 hits per line