MAINT: add support for numpy2

[natsukium]

fixes #210

see more information
https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword

related to scipy/scipy#20172

[bashtage]

Can this simplified to only the try...except? With some comments.

[natsukium]

I have taken the changes to scipy as they are, but I can change them.
However, try-catch is generally considered to be expensive, which is why they use if-elif first in this way, I guess.

[bashtage]

If you simplify to the try...except for additional robustness then I'll merge this in.

[matthewwardrop]

+1. Let's just do the try/except. Thanks for stepping in here @bashtage !

[natsukium]

Thanks for your review!
I simplified it to just a try-except clause and added a comment.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 20.00000% with 8 lines in your changes missing coverage. Please review.

Project coverage is 97.64%. Comparing base (8546450) to head (251951d).

Files with missing lines	Patch %	Lines
patsy/util.py	20.00%	7 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #212      +/-   ##
==========================================
- Coverage   98.46%   97.64%   -0.83%     
==========================================
  Files          30       30              
  Lines        3133     3141       +8     
  Branches      693      598      -95     
==========================================
- Hits         3085     3067      -18     
- Misses         25       46      +21     
- Partials       23       28       +5     

Flag	Coverage Δ
unittests	97.26% <20.00%> (-0.80%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

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

[bashtage]

Sounds like this only affects 2.0.0dev, but not 1.26 or 2.0.0+. Is this right?

[natsukium]

No, as noted in the migration guide, this is a fix for a change in the copy mode in numpy 2.0 (including dev).
It means that if it is not changed, it will not work in the latest 2.1.2.

[bashtage]

No, as noted in the migration guide, this is a fix for a change in the copy mode in numpy 2.0 (including dev).
It means that if it is not changed, it will not work in the latest 2.1.2.

I am a bit skeptical about that. Statsmodels makes heavy use of patsy and runs on NumPy 1.22+ including 2.0.0dev, 2.1 and 2.2. Can you show that patsy fails on a version of NumPy (actual, not just based on the guide)?

[natsukium]

Already reported with logs in issue, #210
I am also a python maintainer on NixOS/nixpkgs, and I got the same error.

Here's the log. The numpy version is 2.1.2.

patsy> ============================= test session starts ==============================
patsy> platform linux -- Python 3.12.7, pytest-8.3.3, pluggy-1.5.0
patsy> rootdir: /build/patsy-0.5.6
patsy> configfile: setup.cfg
patsy> testpaths: patsy
patsy> collected 148 items                                                            
patsy> 
patsy> patsy/build.py .......                                                   [  4%]
patsy> patsy/builtins.py ..                                                     [  6%]
patsy> patsy/categorical.py ....                                                [  8%]
patsy> patsy/constraint.py .....                                                [ 12%]
patsy> patsy/contrasts.py .........                                             [ 18%]
patsy> patsy/desc.py ......                                                     [ 22%]
patsy> patsy/design_info.py ........                                            [ 27%]
patsy> patsy/eval.py ...............                                            [ 37%]
patsy> patsy/infix_parser.py .                                                  [ 38%]
patsy> patsy/mgcv_cubic_splines.py ............                                 [ 46%]
patsy> patsy/missing.py .....                                                   [ 50%]
patsy> patsy/origin.py .                                                        [ 50%]
patsy> patsy/parse_formula.py .....                                             [ 54%]
patsy> patsy/redundancy.py ....                                                 [ 56%]
patsy> patsy/splines.py ....                                                    [ 59%]
patsy> patsy/test_build.py .................                                    [ 70%]
patsy> patsy/test_highlevel.py F....FF...........                               [ 83%]
patsy> patsy/test_regressions.py .                                              [ 83%]
patsy> patsy/test_state.py FF.                                                  [ 85%]
patsy> patsy/tokens.py ..                                                       [ 87%]
patsy> patsy/user_util.py ...                                                   [ 89%]
patsy> patsy/util.py F...............                                           [100%]
patsy> 
patsy> =================================== FAILURES ===================================
patsy> ______________________________ test_formula_likes ______________________________
patsy> 
patsy>     def test_formula_likes():
patsy>         # Plain array-like, rhs only
patsy> >       t([[1, 2, 3], [4, 5, 6]], {}, 0,
patsy>           False,
patsy>           [[1, 2, 3], [4, 5, 6]], ["x0", "x1", "x2"])
patsy> 
patsy> patsy/test_highlevel.py:150: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/test_highlevel.py:104: in t
patsy>     rhs = f(formula_like, data, depth)
patsy> patsy/highlevel.py:290: in dmatrix
patsy>     (lhs, rhs) = _do_highlevel_design(formula_like, data, eval_env,
patsy> patsy/highlevel.py:181: in _do_highlevel_design
patsy>     (lhs, rhs) = (None, asarray_or_pandas(formula_like, subok=True))
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = [[1, 2, 3], [4, 5, 6]], copy = False, dtype = None, subok = True
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> ________________________________ test_builtins _________________________________
patsy> 
patsy> msg = 'Error evaluating factor', origin = EvalFactor('center(x)')
patsy> f = <bound method EvalEnvironment.eval of <patsy.eval.EvalEnvironment object at 0x7fff93314c20>>
patsy> args = ('_patsy_stobj0__center__.transform(x)',)
patsy> kwargs = {'inner_namespace': VarLookupDict([{}, {'x': [1, 2, 3], 'y': [4, 5, 6], 'a b c': [10, 20, 30]}, {'_patsy_stobj0__center__': <patsy.state.Center object at 0x7fff93314950>}])}
patsy> new_exc = PatsyError('Error evaluating factor: ValueError: Unable to avoid copy while creating an array as requested.\nIf using ...r more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.')
patsy> 
patsy>     def call_and_wrap_exc(msg, origin, f, *args, **kwargs):
patsy>         try:
patsy> >           return f(*args, **kwargs)
patsy> 
patsy> patsy/compat.py:36: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/eval.py:169: in eval
patsy>     return eval(code, {}, VarLookupDict([inner_namespace]
patsy> patsy/state.py:106: in transform
patsy>     x = asarray_or_pandas(x)
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = [1, 2, 3], copy = False, dtype = None, subok = False
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> 
patsy> The above exception was the direct cause of the following exception:
patsy> 
patsy>     def test_builtins():
patsy>         data = {"x": [1, 2, 3],
patsy>                 "y": [4, 5, 6],
patsy>                 "a b c": [10, 20, 30]}
patsy>         t("0 + I(x + y)", data, 0,
patsy>           True,
patsy>           [[1], [2], [3], [4], [5], [6]], ["I(x + y)"])
patsy>         t("Q('a b c')", data, 0,
patsy>           True,
patsy>           [[1, 10], [1, 20], [1, 30]], ["Intercept", "Q('a b c')"])
patsy> >       t("center(x)", data, 0,
patsy>           True,
patsy>           [[1, -1], [1, 0], [1, 1]], ["Intercept", "center(x)"])
patsy> 
patsy> patsy/test_highlevel.py:498: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/test_highlevel.py:83: in t
patsy>     builder = incr_dbuilder(formula_like, data_iter_maker, depth)
patsy> patsy/highlevel.py:111: in incr_dbuilder
patsy>     design_infos = _try_incr_builders(formula_like, data_iter_maker, eval_env,
patsy> patsy/highlevel.py:66: in _try_incr_builders
patsy>     return design_matrix_builders([formula_like.lhs_termlist,
patsy> patsy/build.py:693: in design_matrix_builders
patsy>     cat_levels_contrasts) = _examine_factor_types(all_factors,
patsy> patsy/build.py:443: in _examine_factor_types
patsy>     value = factor.eval(factor_states[factor], data)
patsy> patsy/eval.py:568: in eval
patsy>     return self._eval(memorize_state["eval_code"],
patsy> patsy/eval.py:551: in _eval
patsy>     return call_and_wrap_exc("Error evaluating factor",
patsy> patsy/compat.py:43: in call_and_wrap_exc
patsy>     exec("raise new_exc from e")
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> >   ???
patsy> E   patsy.PatsyError: Error evaluating factor: ValueError: Unable to avoid copy while creating an array as requested.
patsy> E   If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E   For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> E       center(x)
patsy> E       ^^^^^^^^^
patsy> 
patsy> <string>:1: PatsyError
patsy> _______________________________ test_incremental _______________________________
patsy> 
patsy> msg = 'Error evaluating factor'
patsy> origin = EvalFactor('center(np.sin(center(x)))')
patsy> f = <bound method EvalEnvironment.eval of <patsy.eval.EvalEnvironment object at 0x7fff302e7ec0>>
patsy> args = ('_patsy_stobj0__center__.memorize_chunk(np.sin(_patsy_stobj1__center__.transform(x)))',)
patsy> kwargs = {'inner_namespace': VarLookupDict([{}, {'a': ['a2', 'a2', 'a2'], 'x': [1, 2, 3]}, {'_patsy_stobj0__center__': <patsy.state.Center object at 0x7fff302e4830>, '_patsy_stobj1__center__': <patsy.state.Center object at 0x7fff302e5280>}])}
patsy> new_exc = PatsyError('Error evaluating factor: ValueError: Unable to avoid copy while creating an array as requested.\nIf using ...r more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.')
patsy> 
patsy>     def call_and_wrap_exc(msg, origin, f, *args, **kwargs):
patsy>         try:
patsy> >           return f(*args, **kwargs)
patsy> 
patsy> patsy/compat.py:36: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/eval.py:169: in eval
patsy>     return eval(code, {}, VarLookupDict([inner_namespace]
patsy> patsy/state.py:106: in transform
patsy>     x = asarray_or_pandas(x)
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = [1, 2, 3], copy = False, dtype = None, subok = False
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> 
patsy> The above exception was the direct cause of the following exception:
patsy> 
patsy>     def test_incremental():
patsy>         # incr_dbuilder(s)
patsy>         # stateful transformations
patsy>         datas = [
patsy>             {"a": ["a2", "a2", "a2"],
patsy>              "x": [1, 2, 3]},
patsy>             {"a": ["a2", "a2", "a1"],
patsy>              "x": [4, 5, 6]},
patsy>             ]
patsy>         x = np.asarray([1, 2, 3, 4, 5, 6])
patsy>         sin_center_x = np.sin(x - np.mean(x))
patsy>         x_col = sin_center_x - np.mean(sin_center_x)
patsy>         def data_iter_maker():
patsy>             return iter(datas)
patsy> >       builders = incr_dbuilders("1 ~ a + center(np.sin(center(x)))",
patsy>                                   data_iter_maker)
patsy> 
patsy> patsy/test_highlevel.py:516: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/highlevel.py:129: in incr_dbuilders
patsy>     design_infos = _try_incr_builders(formula_like, data_iter_maker, eval_env,
patsy> patsy/highlevel.py:66: in _try_incr_builders
patsy>     return design_matrix_builders([formula_like.lhs_termlist,
patsy> patsy/build.py:689: in design_matrix_builders
patsy>     factor_states = _factors_memorize(all_factors, data_iter_maker, eval_env)
patsy> patsy/build.py:368: in _factors_memorize
patsy>     factor.memorize_chunk(state, which_pass, data)
patsy> patsy/eval.py:559: in memorize_chunk
patsy>     self._eval(state["memorize_code"][obj_name],
patsy> patsy/eval.py:551: in _eval
patsy>     return call_and_wrap_exc("Error evaluating factor",
patsy> patsy/compat.py:43: in call_and_wrap_exc
patsy>     exec("raise new_exc from e")
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> >   ???
patsy> E   patsy.PatsyError: Error evaluating factor: ValueError: Unable to avoid copy while creating an array as requested.
patsy> E   If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E   For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> E       1 ~ a + center(np.sin(center(x)))
patsy> E               ^^^^^^^^^^^^^^^^^^^^^^^^^
patsy> 
patsy> <string>:1: PatsyError
patsy> _________________________________ test_Center __________________________________
patsy> 
patsy>     def test_Center():
patsy> >       check_stateful(Center, True, [1, 2, 3], [-1, 0, 1])
patsy> 
patsy> patsy/test_state.py:117: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/test_state.py:81: in check_stateful
patsy>     output_chunk = t.transform(input_chunk, *args, **kwargs)
patsy> patsy/state.py:106: in transform
patsy>     x = asarray_or_pandas(x)
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = np.int64(1), copy = False, dtype = None, subok = False
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> ----------------------------- Captured stdout call -----------------------------
patsy> [array([1, 2, 3])]
patsy> [1 2 3]
patsy> _______________________ test_stateful_transform_wrapper ________________________
patsy> 
patsy>     def test_stateful_transform_wrapper():
patsy> >       assert np.allclose(center([1, 2, 3]), [-1, 0, 1])
patsy> 
patsy> patsy/test_state.py:124: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> patsy/state.py:48: in stateful_transform_wrapper
patsy>     return transform.transform(*args, **kwargs)
patsy> patsy/state.py:106: in transform
patsy>     x = asarray_or_pandas(x)
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = [1, 2, 3], copy = False, dtype = None, subok = False
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> ____________________________ test_asarray_or_pandas ____________________________
patsy> 
patsy>     def test_asarray_or_pandas():
patsy>         import warnings
patsy> >       assert type(asarray_or_pandas([1, 2, 3])) is np.ndarray
patsy> 
patsy> patsy/util.py:72: 
patsy> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
patsy> 
patsy> a = [1, 2, 3], copy = False, dtype = None, subok = False
patsy> 
patsy>     def asarray_or_pandas(a, copy=False, dtype=None, subok=False):
patsy>         if have_pandas:
patsy>             if isinstance(a, (pandas.Series, pandas.DataFrame)):
patsy>                 # The .name attribute on Series is discarded when passing through
patsy>                 # the constructor:
patsy>                 #   https://github.com/pydata/pandas/issues/1578
patsy>                 extra_args = {}
patsy>                 if hasattr(a, "name"):
patsy>                     extra_args["name"] = a.name
patsy>                 return a.__class__(a, copy=copy, dtype=dtype, **extra_args)
patsy> >       return np.array(a, copy=copy, dtype=dtype, subok=subok)
patsy> E       ValueError: Unable to avoid copy while creating an array as requested.
patsy> E       If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
patsy> E       For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.
patsy> 
patsy> patsy/util.py:67: ValueError
patsy> =========================== short test summary info ============================
patsy> FAILED patsy/test_highlevel.py::test_formula_likes - ValueError: Unable to avoid copy while creating an array as requested.
patsy> FAILED patsy/test_highlevel.py::test_builtins - patsy.PatsyError: Error evaluating factor: ValueError: Unable to avoid copy...
patsy> FAILED patsy/test_highlevel.py::test_incremental - patsy.PatsyError: Error evaluating factor: ValueError: Unable to avoid copy...
patsy> FAILED patsy/test_state.py::test_Center - ValueError: Unable to avoid copy while creating an array as requested.
patsy> FAILED patsy/test_state.py::test_stateful_transform_wrapper - ValueError: Unable to avoid copy while creating an array as requested.
patsy> FAILED patsy/util.py::test_asarray_or_pandas - ValueError: Unable to avoid copy while creating an array as requested.
patsy> =================== 6 failed, 142 passed in 88.39s (0:01:28) ===================

[matthewwardrop]

Thanks for updating the patch, @natsukium . Merging it in.