Add tuples, ndarrays, and complex numbers to cirq_google proto #7226
Conversation
dstrain115
commented
Apr 3, 2025
dstrain115
commented
- Adds support for (potentially nested) tuples of mismatched types
- Adds support for serializing ndarrays
- Adds support for serializing complex numbers
- Adds support for (potentially nested) tuples of mismatched types - Adds support for serializing ndarrays - Adds support for serializing complex numbers
dstrain115
requested review from
a team,
verult,
vtomole and
wcourtney
as code owners
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #7226 +/- ##
========================================
Coverage 98.65% 98.65%
========================================
Files 1106 1106
Lines 95869 95969 +100
========================================
+ Hits 94581 94681 +100
Misses 1288 1288 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| double real_value = 1; | ||
| double imaginary_value = 2; |
There was a problem hiding this comment.
Can we use real and imag for parity with builtin complex and numpy numeric types?
| match value.dtype.name: | ||
| case 'float64': |
There was a problem hiding this comment.
We can match value.dtype rather than the name. This would protect against typo in string or a rename of data type.
| return ( | ||
| arg_value.complex_value.real_value | ||
| + 1j * arg_value.complex_value.imaginary_value |
There was a problem hiding this comment.
| return ( | |
| arg_value.complex_value.real_value | |
| + 1j * arg_value.complex_value.imaginary_value | |
| return complex( | |
| arg_value.complex_value.real_value, | |
| arg_value.complex_value.imaginary_value |
| elif isinstance(value, (list, tuple, np.ndarray)): | ||
| if len(value): |
There was a problem hiding this comment.
On a round trip this converts (i) uniform-type lists or tuples to lists and (ii) non-uniform lists and tuples to tuples. It also discards empty lists, tuples and arrays, which return as None from round trip.
We can move the array conversion out of the if len(value) block as _ndarray_to_proto and _ndarray_from_proto seem to work with empty arrays.
As for tuples and lists, we could add a sequence_type enum to the ArgValue message which could be (UNSPECIFIED, LIST, TUPLE) and would let us reconstruct either list or tuple from a round trip. It would also allow us to express either an empty list or an empty tuple, in such case ArgValue would be empty except of a (new) sequence_type.
There was a problem hiding this comment.
Good idea! I went ahead and did that, and also added set support while in there.
mhucka
added this to the
2025-T5: Enhance functions for circuit construction and transformations milestone
Co-authored-by: Pavol Juhas <[email protected]>
mhucka
removed this from the
2025-T5: Enhance functions for circuit construction and transformations milestone
…umlib#7226) * Add tuples, ndarrays, and complex numbers to cirq_google proto - Adds support for (potentially nested) tuples of mismatched types - Adds support for serializing ndarrays - Adds support for serializing complex numbers * my o my, it's mypy * Fix all the review comments, add set support. * Fix invert mask, which assumes list, so it can accept tuples. * Update cirq-google/cirq_google/serialization/arg_func_langs.py Co-authored-by: Pavol Juhas <[email protected]> * Fix coverage. --------- Co-authored-by: Pavol Juhas <[email protected]>
3 participants
