Feature: Expr and GenExpr support numpy unary func like np.sin

[Zeroto521]

Expr and GenExpr support numpy unary functions like np.sin.

Now we can do this

In [8]: x = m.addVar(name='x')

In [9]: np.sin(x)
Out[9]: sin(sum(0.0,prod(1.0,x)))

In [10]: np.cos([x, y])
Out[10]: [cos(sum(0.0,prod(1.0,x))) sin(cos(0.0,prod(1.0,y)))]

Before this, we couldn't

In [8]: x = m.addVar(name='x')

In [9]: np.sin(x)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
AttributeError: 'pyscipopt.scip.Variable' object has no attribute 'sin'

The above exception was the direct cause of the following exception:

TypeError                                 Traceback (most recent call last)
Cell In[9], line 1
----> 1 np.sin(x)

TypeError: loop of ufunc does not support argument 0 of type pyscipopt.scip.Variable which has no callable sin method

In [12]: np.cos([x, y])
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
AttributeError: 'pyscipopt.scip.Variable' object has no attribute 'cos'

The above exception was the direct cause of the following exception:

TypeError                                 Traceback (most recent call last)
Cell In[12], line 1
----> 1 np.cos([x, y])

TypeError: loop of ufunc does not support argument 0 of type pyscipopt.scip.Variable which has no callable cos method

[Copilot]

Pull request overview

This PR adds support for NumPy unary functions (np.sin, np.cos, np.sqrt, np.exp, np.log, np.absolute) to work with Expr and GenExpr objects by implementing the __array_ufunc__ protocol.

Changes:

• Introduces a new ExprLike base class that implements __array_ufunc__ and unary operation methods
• Makes Expr and GenExpr inherit from ExprLike to support NumPy ufuncs
• Replaces custom function implementations with NumPy function aliases

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 7 comments.

File	Description
src/pyscipopt/scip.pxd	Adds ExprLike base class declaration and updates Expr inheritance hierarchy
src/pyscipopt/expr.pxi	Implements ExprLike with __array_ufunc__ support, consolidates unary methods, aliases functions to NumPy equivalents
tests/test_expr.py	Adds test coverage for both custom functions and NumPy ufuncs with single expressions and arrays
CHANGELOG.md	Documents the new NumPy unary function support feature

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

❌ Patch coverage is 34.48276% with 19 lines in your changes missing coverage. Please review.
✅ Project coverage is 56.53%. Comparing base (ac47132) to head (3688217).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/pyscipopt/expr.pxi	34.48%	19 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1170      +/-   ##
==========================================
- Coverage   56.64%   56.53%   -0.11%     
==========================================
  Files          26       26              
  Lines        5602     5628      +26     
==========================================
+ Hits         3173     3182       +9     
- Misses       2429     2446      +17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Joao-Dionisio]

@Zeroto521 can you please add a couple more tests? It feels uncomfortable to merge something with 34% coverage.

[Zeroto521]

@Zeroto521 can you please add a couple more tests? It feels uncomfortable to merge something with 34% coverage.

The codes are different between codecov and the head commit of this PR.

The head commit is the following

exp = lambda x: _wrap_ufunc(x, np.exp)
log = lambda x: _wrap_ufunc(x, np.log)
sqrt = lambda x: _wrap_ufunc(x, np.sqrt)
sin = lambda x: _wrap_ufunc(x, np.sin)
cos = lambda x: _wrap_ufunc(x, np.cos)

But the codecov is still the old version.

Can the codecov CI be triggered manually and run again?

[Joao-Dionisio]

There is a slowdown on the following:

from timeit import timeit
from pyscipopt import Model

m = Model()
x = m.addVar("x")
genexpr = 1/x

n = 100_000
print(f"2**Expr:    {timeit(lambda: 2**x, number=n):.2f}s")
print(f"2**GenExpr: {timeit(lambda: 2**genexpr, number=n):.2f}s")

I don't know if there's code optimization to be found somewhere, or if the slowdown is significant enough.

[Zeroto521]

There is a slowdown on the following:

from timeit import timeit
from pyscipopt import Model

m = Model()
x = m.addVar("x")
genexpr = 1/x

n = 100_000
print(f"2**Expr:    {timeit(lambda: 2**x, number=n):.2f}s")
print(f"2**GenExpr: {timeit(lambda: 2**genexpr, number=n):.2f}s")

I don't know if there's code optimization to be found somewhere, or if the slowdown is significant enough.

Directly calling inner methods (.exp() and .log) could speed up a bit.
The next step could drop buildGenExprObj by merging Expr and GenExpr. buildGenExprObj (a pure Python function) is currently the performance bottleneck.

-            return exp(self * log(base))
+            return (self * Constant(base).log()).exp()

	master	PR#1170
2**Expr	0.8418s	0.8494s
2**GenExpr	0.3124s	0.3643s

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Comments suppressed due to low confidence (1)

src/pyscipopt/scip.pyi:361

• Expr now contains duplicate dunder method declarations: non-positional-only versions were added (e.g., def __add__(self, other: Incomplete)) but the positional-only versions still exist below. This duplication is likely to fail the stub linting (ruff) and is inconsistent with the earlier change to mark dunder params as positional-only; please remove the redundant non-positional-only declarations (or convert to proper @overload if they are intended to differ).

    def __add__(self, other: Incomplete) -> Incomplete: ...
    def __eq__(self, other: object) -> bool: ...
    def __ge__(self, other: object) -> bool: ...
    def __getitem__(self, index: Incomplete) -> Incomplete: ...
    def __gt__(self, other: object) -> bool: ...
    def __iadd__(self, other: Incomplete) -> Incomplete: ...  # noqa: PYI034
    def __abs__(self) -> Incomplete: ...
    def __add__(self, other: Incomplete, /) -> Incomplete: ...
    def __eq__(self, other: object, /) -> bool: ...
    def __ge__(self, other: object, /) -> bool: ...
    def __getitem__(self, index: Incomplete, /) -> Incomplete: ...
    def __gt__(self, other: object, /) -> bool: ...
    def __iadd__(self, other: Incomplete, /) -> Incomplete: ...  # noqa: PYI034

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Joao-Dionisio]

After addressing Copilot's comments, this can be merged, I think.

[Joao-Dionisio]

My only concern is that now numpy will be used even if you're not using arrays, the dependency coupling is much stricter now. But, it ultimately doesn't seem too problematic. Thank you once more, @Zeroto521 !