Replace interp1d #2394 by jason-rpkt · Pull Request #2741 · pvlib/pvlib-python

jason-rpkt · 2026-04-23T09:38:30Z

Closes scipy.interpolate.interp1d is discouraged #2394
I am familiar with the contributing guidelines
I attest that all AI-generated material has been vetted for accuracy and is in compliance with the pvlib license
Tests added
Updates entries in docs/sphinx/source/reference for API changes.
Adds description and name entries in the appropriate "what's new" file in docs/sphinx/source/whatsnew for all changes. Includes link to the GitHub Issue with :issue:`num` or this Pull Request with :pull:`num`. Includes contributor name and/or GitHub username (link with :ghuser:`user`).
New code is fully documented. Includes numpydoc compliant docstrings, examples, and comments where necessary.
Pull request is nearly complete and ready for detailed review.
Maintainer: Appropriate GitHub Labels (including remote-data) and Milestone are assigned to the Pull Request and linked Issue.

The existing tests pass. I have added 2 in test_iam.py under test_iam_interp.

cwhanse · 2026-04-23T15:38:40Z

Reason for importing CubicSpline instead of using make_interp_spline(..., k=3)? The latter would clean up the case processing, I think.

jason-rpkt · 2026-04-24T02:29:30Z

no particular reason. Just that I had seen CubicSpline first in https://docs.scipy.org/doc/scipy-1.17.0/tutorial/interpolate/1D.html#tutorial-interpolate-1dsection and in response.py it was hardcoded as cubic.

I have replaced CubicSpline by make_interp_spline with k=3. Now only importing make_interp_spline.

kandersolar · 2026-04-24T12:44:07Z

-                            fill_value='extrapolate')
-    aoi_input = aoi
+    theta_ref = np.asarray(theta_ref)
+    iam_ref = np.asarray(iam_ref)


Why is np.asarray needed here?

Do you mean to use np.asanyarray to keep consistency with np.asanyarray(aoi) below? Or to remove entirely?

np.asarray is to convert to numpy array but I can remove if you prefer.

I think the issue is that the docstrings states numeric for input parameters. pvlib means numeric as one of float, array or Series. Testing is all in one function test_iam_interp which inputs either list, float or array (but not Series).

My opinion: rewrite the test to remove inputting lists, which obviates np.asarray inside iam.interp. And add a test that inputs Series.

Hi both, thanks for your comments. Not sure if I got this right, but I have rewritten the test to remove inputting lists. (And removed np.asarray in iam.interp )
I have also added a check for list inputs at the beginning of iam.interp
Let me know if this is OK.

--Looking at it again, it seems that keeping np.asarray might be "cleaner" than the latest version.

kandersolar · 2026-04-24T12:47:13Z

+            return spline(x)
+
+    else:
+        raise ValueError(f"Invalid interpolation method '{method}'.")


This is, technically, a breaking change, since the interp1d way also supported 'nearest', 'nearest-up', 'zero', 'slinear', 'previous', and 'next'. I doubt these got much use, if any. Any thoughts on how to handle that?

I'm not sure I know how to update to support all of those.
Is it OK to update the error message?

Since interp1d is legacy but not intended for removal, it is still available and we could fall back:

elif method in {'nearest', 'nearest-up', 'zero', 'slinear', 'previous', 'next'}: interpolator = interp1d(theta_ref, iam_ref, kind=method, fill_value='extrapolate')

I think that's reasonable, as long as we deprecate these weird methods too. If someone really wants these special interpolations, they can do it themselves and provide a pre-calculated input.

None of those options really make any sense here. I would vote for dropping them immediately.

kandersolar · 2026-04-24T12:48:00Z

+        spline = make_interp_spline(theta_ref, iam_ref, k=3)
+
+        def interpolator(x):
+            return spline(x)


This can surely be simplified as @cwhanse mentioned in #2741 (comment)

kandersolar · 2026-04-24T12:51:54Z

        Specifies the interpolation method.
        Useful options are: 'linear', 'quadratic', 'cubic'.
-        See scipy.interpolate.interp1d for more options.
+        See scipy.interpolate for more options.


This line may need to be edited, depending on https://github.com/pvlib/pvlib-python/pull/2741/changes#r3137773820

cwhanse · 2026-05-05T14:05:13Z

+        interpolator = make_interp_spline(theta_ref, iam_ref, k=3)
+
+    else:
+        raise ValueError(f"Invalid interpolation method '{method}'.")


Suggested change

raise ValueError(f"Invalid interpolation method '{method}'.")

raise ValueError(f"Interpolation method '{method}' is not supported in pvlib-python.")

cwhanse · 2026-05-05T14:05:17Z

-
-    from scipy.interpolate import interp1d
-
+    if isinstance(theta_ref, list):


We normally don't check the type of input. If a user inputs a non-allowed type e.g. a list, make_interp_spline will emit an error. That's OK.

cwhanse · 2026-05-05T14:07:26Z

+            return spline(x)
+
+    else:
+        raise ValueError(f"Invalid interpolation method '{method}'.")


Since interp1d is legacy but not intended for removal, it is still available and we could fall back:

elif method in {'nearest', 'nearest-up', 'zero', 'slinear', 'previous', 'next'}: interpolator = interp1d(theta_ref, iam_ref, kind=method, fill_value='extrapolate')

adriesse · 2026-05-06T22:10:20Z

+    if isinstance(iam_ref, list):
+        raise TypeError("iam_ref cannot be a list")
    # Scipy doesn't give the clearest feedback, so check number of points here.
    MIN_REF_VALS = {'linear': 2, 'quadratic': 3, 'cubic': 4, 1: 2, 2: 3, 3: 4}


You could add a dictionary with names and k values, or something like that, to use below.

adriesse · 2026-05-06T22:12:08Z

+            return spline(x)
+
+    else:
+        raise ValueError(f"Invalid interpolation method '{method}'.")


None of those options really make any sense here. I would vote for dropping them immediately.

adriesse · 2026-05-06T22:15:00Z

-                        fill_value='extrapolate')
+    """convenience wrapper around scipy.interpolate"""
+    f_interp = make_interp_spline(np.flipud(df['i']), np.flipud(df['v']), k=1)
+


This is one where performance is likely important. Maybe check if np.interp is faster?

Replace interp1d pvlib#2394

8773810

jason-rpkt marked this pull request as ready for review April 23, 2026 09:43

Replace CubicSpline by make_interp_spline k=3

ab6c74f

kandersolar added this to the v0.15.2 milestone Apr 24, 2026

kandersolar added the dependency label Apr 24, 2026

kandersolar reviewed Apr 24, 2026

View reviewed changes

jason-rpkt added 2 commits April 27, 2026 11:16

clean up/ simplify/move import to top

c658199

remove list input

486cf38

cwhanse reviewed May 5, 2026

View reviewed changes

adriesse reviewed May 6, 2026

View reviewed changes

	raise ValueError(f"Invalid interpolation method '{method}'.")
	raise ValueError(f"Interpolation method '{method}' is not supported in pvlib-python.")


		from scipy.interpolate import interp1d

		if isinstance(theta_ref, list):

Conversation

jason-rpkt commented Apr 23, 2026

Uh oh!

cwhanse commented Apr 23, 2026

Uh oh!

jason-rpkt commented Apr 24, 2026

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jason-rpkt May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jason-rpkt May 5, 2026 •

edited

Loading