feat: Add sparse.interp function for numba backend

[peanutfun]

What type of PR is this? (check all applicable)

• 💾 Refactor
• 🪄 Feature
• 🐞 Bug Fix
• 🔧 Optimization
• 📚 Documentation
• 🧪 Test
• 🛠️ Other

Related issues

• None

Checklist

• Code follows style guide
• Tests added
• Documented the changes

---

This adds sparse.interp, the sparse equivalent to numpy.interp. The latter does not work on sparse array types, so I decided to implement a "wrapper". It effectively calls numpy.interp on the array data and its fill value and returns the original data type.

This function should integrate nicely because numpy will automatically dispatch any call to numpy.interp to sparse.interp, if one of the arguments is a sparse array. This is explicitly tested.

[hameerabbasi]

A large part of this PR LGTM. I have just one overarching comment: The tests should look more like the tests in test_coo.py.

[peanutfun]

@hameerabbasi Thanks, but can you be a bit more specific? Do I understand correctly that you want a test that

1. Creates a random (dense) array: x
2. Creates a sparse copy of x: y
3. Calls np.interp with x and y
4. Ensures the results are equal

[hameerabbasi]

@hameerabbasi Thanks, but can you be a bit more specific? Do I understand correctly that you want a test that

1. Creates a random (dense) array: x

2. Creates a sparse copy of x: y

3. Calls `np.interp` with x and y

4. Ensures the results are equal

That's correct, yes.

[prady0t]

You can also move the test to sparse/numba_backend/tests/test_coo.py.

[peanutfun]

@hameerabbasi I moved the test to test_coo.py and (hopefully) adapted it to the testing style there. I also slightly updated the function with a shortcut for dense input data, preserving the data type if x is not a sparse array.

Let me know if there is anything else I can do.

[hameerabbasi]

Thanks @peanutfun, I appreciate the addition! Just waiting for CI. Would you like a new release?

[hameerabbasi]

@peanutfun The doctest and one other test need updates; would you be willing to make those changes?

[peanutfun]

@hameerabbasi I hope I fixed it now! A new release would be awesome, but it's not urgent for me.

[hameerabbasi]

@peanutfun Small doctest failure still remaining: https://github.com/pydata/sparse/actions/runs/19098274651/job/54563742981?pr=903#step:5:12738

[peanutfun]

@hameerabbasi Indeed! Apparently I did not run the doctests. How about now? 🤞

[codspeed-hq]

CodSpeed Performance Report

Merging #903 will degrade performances by 33.06%

_{Comparing peanutfun:add-interp (6e8ad0c) with main (3a2e703)}

Summary

❌ 2 regressions
✅ 338 untouched

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	test_index_fancy[side=100-rank=1-format='coo']	1.2 ms	1.4 ms	-10.62%
❌	test_index_slice[side=100-rank=2-format='gcxs']	1.8 ms	2.6 ms	-33.06%

[hameerabbasi]

That worked, thanks @peanutfun!

[peanutfun]

Thanks for the support! This was my first contribution here, and it was a very smooth and pleasant experience 😊