Conversation
NumPy 2.0 introduced NEP 50 which changed type promotion rules: when an unsigned integer array operates with a signed integer scalar of the same or lower "kind", the array dtype now wins (no type widening). This updates _typemap_arr_scalar for 12 entries: - uint8 + int16/int32/int64 → uint8 (was int16/int32/int64) - uint16 + int16/int32/int64 → uint16 (was int32/int32/int64) - uint32 + int16/int32/int64 → uint32 (was int64/int64/int64) - uint64 + int16/int32/int64 → uint64 (was float64/float64/float64) Breaking change for users relying on NumPy 1.x widening behavior. This aligns with NumPy 2.x and the project's NumPy alignment goal. Closes #529 See: https://numpy.org/neps/nep-0050-scalar-promotion.html
Adds 44 tests covering NEP 50 type promotion rules, each verified against NumPy 2.4.2 output via Python execution. Test categories: - Unsigned/signed array + Python int (array dtype wins) - Float array + Python int/float (array dtype wins) - Int array + Python float (cross-kind → float64) - All arithmetic operations: +, -, *, /, % - Scalar-first operations (commutative) - Direct _FindCommonArrayScalarType table verification - Array-array operations (unchanged by NEP 50) - Edge cases: empty arrays, multi-dimensional, scalar arrays Documents NumSharp behavioral differences (marked [Misaligned]): - Division uses integer semantics (not true division like NumPy) - Scalar-scalar uses different promotion path than arr-scalar Each test includes the NumPy verification command in its docstring: Verified: python3 -c "import numpy as np; ..."
Adds comprehensive operation matrix tests: - 12 type combinations (4 unsigned × 3 signed scalar types) - 4 operations each (+, -, *, %) - 48 dtype assertions + value correctness tests Coverage now includes: - uint8/uint16/uint32/uint64 arrays - int16/int32/int64 scalars (short/int/long) - All arithmetic operations - Value correctness verification Total: 59 tests, ~150+ assertions
Replace inline "// NEP 50: array dtype wins" comments with detailed block documentation explaining: - NumSharp design decision for scalar handling - Why C# scalars are treated as "weakly typed" like Python scalars - Complete table of affected entries (12 combinations) - NumPy 1.x vs 2.x behavior comparison - Verification command and NumPy version tested against - Reference to NEP 50 specification The documentation block now serves as the single source of truth for understanding the type promotion behavior in this table.
Add detailed documentation explaining the entire type promotion system: - Architecture overview (4 lookup tables) - When each table is used (arr+arr vs arr+scalar) - Kind hierarchy (boolean < integer < float < complex) - Within-kind promotion rules - Cross-kind promotion rules - Practical examples - References to NumPy docs, NEP 50, Array API spec Also document arr_arr table with specific promotion rules: - Same type, same kind different size - Signed + unsigned combinations - Cross-kind and complex promotion
Nucs
added
NumPy 2.x Compliance
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Implements NEP 50 type promotion rules for NumSharp, aligning with NumPy 2.x behavior.
Key change: When an unsigned integer array operates with a signed integer scalar, the array dtype now wins (no type widening).
Changes
Type Promotion Table (
np.find_common_type.cs)_typemap_arr_scalarfor unsigned array + signed scalar combinationsTest Coverage (
NEP50.cs)Affected Type Combinations (12)
Breaking Change
This is a behavioral breaking change for users who relied on NumPy 1.x type widening behavior. This aligns with NumPy 2.x and the project's NumPy alignment goal.
Test Plan
Related Issues
References