Bug#918206: Pandas new version (not?)

[Rebecca N. Palmer]

This builds (including build-time tests; haven't tried the autopkgtest).

Description: Fix np.array @ DataFrame matrix multiplication

Author: jbrockmendel
Origin: upstream 
https://github.com/pandas-dev/pandas/commit/ad2a14f4bec8a004b2972c12f12ed3e4ce37ff52

Bug-Debian: https://bugs.debian.org/918206
Forwarded: not-needed

--- a/pandas/core/generic.py
+++ b/pandas/core/generic.py
@@ -1602,6 +1602,7 @@ class NDFrame(PandasObject, SelectionMix
 # 
--

 # Array Interface

+__array_priority__ = 1000
 def __array__(self, dtype=None):
 return com._values_from_object(self)

--- a/pandas/tests/frame/test_analytics.py
+++ b/pandas/tests/frame/test_analytics.py
@@ -2283,8 +2283,11 @@ class TestDataFrameAnalytics(TestData):

 # np.array @ DataFrame
 result = operator.matmul(a.values, b)
+assert isinstance(result, DataFrame)
+assert result.columns.equals(b.columns)
+assert result.index.equals(pd.Index(range(3)))
 expected = np.dot(a.values, b.values)
-tm.assert_almost_equal(result, expected)
+tm.assert_almost_equal(result.values, expected)

 # nested list @ DataFrame (__rmatmul__)
 result = operator.matmul(a.values.tolist(), b)

Bug#918206: Pandas new version (not?)

[Rebecca N. Palmer]

Given the soft freeze, the new upstream version is probably no longer an 
option for buster.


Just setting __array_priority__ fails, but with a message that suggests 
also including the test_analytics part of that upstream commit would 
work; I haven't yet had time to try this.



left = one   two
0 -1.365296  1.088132
1 -1.668353  3.025582
2  0.343140 -0.260306
right = array([[-1.36529629,  1.08813166],
   [-1.66835336,  3.02558216],
   [ 0.34314   , -0.26030576]])
cls = 

def _check_isinstance(left, right, cls):
"""
Helper method for our assert_* methods that ensures that
the two objects being compared have the right type before
proceeding with the comparison.

Parameters
--
left : The first object being compared.
right : The second object being compared.
cls : The class type to check against.

Raises
--
AssertionError : Either `left` or `right` is not an instance of 
`cls`.

"""

err_msg = "{name} Expected type {exp_type}, found {act_type} 
instead"

cls_name = cls.__name__

if not isinstance(left, cls):
raise AssertionError(err_msg.format(name=cls_name, 
exp_type=cls,

act_type=type(left)))
if not isinstance(right, cls):
raise AssertionError(err_msg.format(name=cls_name, 
exp_type=cls,

>   act_type=type(right)))
E   AssertionError: DataFrame Expected type 'pandas.core.frame.DataFrame'>, found  instead


../debian/tmp/usr/lib/python3/dist-packages/pandas/util/testing.py:296: 
AssertionError

Bug#918206: Pandas new version

[Rebecca N. Palmer]

Control: tags -1 fixed-upstream patch

The test failure is that np.array @ pd.DataFrame (matrix product) tries 
to keep both the DataFrame's indices, which fails because the new matrix 
is a different shape.


This appears to be fixed in 0.24.1 from PyPI, but as previously noted, 
this is a new major version and hence risks breaking rdeps.


The relevant change appears to be setting __array_priority__:

https://github.com/brute4s99/pandas/commit/a01fa791eafe704ea85e2acc956ad9077e8e7542#diff-03b380f521c43cf003207b0711bac67f

but I haven't actually tried applying only that to 0.23.