feat: Lower unpacking assignment of iterables to Hugr #689
Conversation
| popped_arr_ty = array_type(elem_ty, ht.BoundedNatArg(length - 1)) | ||
| op = "pop_left" if from_left else "pop_right" | ||
| return _instantiate_array_op( | ||
| op, elem_ty, length_arg, [arr_ty], [ht.Option(elem_ty, popped_arr_ty)] |
There was a problem hiding this comment.
Side note: Maybe we should add an infallibe array_pop op where we statically check that the length isn't 0?
There was a problem hiding this comment.
Agree. It's not clear to me why we can't do this always, since for pop we always know the length of the array.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## feat/unpacking #689 +/- ##
==================================================
+ Coverage 92.88% 92.93% +0.04%
==================================================
Files 68 68
Lines 7848 7902 +54
==================================================
+ Hits 7290 7344 +54
Misses 558 558 ☔ View full report in Codecov by Sentry. |
| popped_arr_ty = array_type(elem_ty, ht.BoundedNatArg(length - 1)) | ||
| op = "pop_left" if from_left else "pop_right" | ||
| return _instantiate_array_op( | ||
| op, elem_ty, length_arg, [arr_ty], [ht.Option(elem_ty, popped_arr_ty)] |
There was a problem hiding this comment.
Agree. It's not clear to me why we can't do this always, since for pop we always know the length of the array.
| def array_discard_empty(elem_ty: ht.Type) -> ops.ExtOp: | ||
| """Returns an operation that discards an array of length zero.""" | ||
| arr_ty = array_type(elem_ty, ht.BoundedNatArg(0)) | ||
| return hugr.std.PRELUDE.get_op("discard_empty").instantiate( |
There was a problem hiding this comment.
can we use _instantiate_array_op here? the extension has changed since this was written.
There was a problem hiding this comment.
I don't think so since discard_empty doesn't take a length arg
Yes it's the first time they are emitted in Guppy. But Guppy doesn't allow you to pop if the length is generic |
Adds the Hugr lowering for #649. See #688 for the base PR that adds the checking logic