Constructing a df with readonly array of Periods fails

[max-sixty]

Code Sample, a copy-pastable example if possible

In [1]: import pandas as pd

In [2]: import numpy as np

In [6]: pa = pd.PeriodIndex([pd.Period('2019-01-01')]).to_numpy()


In [8]: pa.setflags(write=False)

In [9]: pd.DataFrame(dict(date=pa, x=[1]))
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-9-7a4f1ee8be64> in <module>()
----> 1 pd.DataFrame(dict(date=pa, x=[1]))

/usr/local/lib/python2.7/site-packages/pandas/core/frame.pyc in __init__(self, data, index, columns, dtype, copy)
    390                                  dtype=dtype, copy=copy)
    391         elif isinstance(data, dict):
--> 392             mgr = init_dict(data, index, columns, dtype=dtype)
    393         elif isinstance(data, ma.MaskedArray):
    394             import numpy.ma.mrecords as mrecords

/usr/local/lib/python2.7/site-packages/pandas/core/internals/construction.pyc in init_dict(data, index, columns, dtype)
    210         arrays = [data[k] for k in keys]
    211
--> 212     return arrays_to_mgr(arrays, data_names, index, columns, dtype=dtype)
    213
    214

/usr/local/lib/python2.7/site-packages/pandas/core/internals/construction.pyc in arrays_to_mgr(arrays, arr_names, index, columns, dtype)
     54
     55     # don't force copy because getting jammed in an ndarray anyway
---> 56     arrays = _homogenize(arrays, index, dtype)
     57
     58     # from BlockManager perspective

/usr/local/lib/python2.7/site-packages/pandas/core/internals/construction.pyc in _homogenize(data, index, dtype)
    275                 val = lib.fast_multiget(val, oindex.values, default=np.nan)
    276             val = sanitize_array(val, index, dtype=dtype, copy=False,
--> 277                                  raise_cast_failure=False)
    278
    279         homogenized.append(val)

/usr/local/lib/python2.7/site-packages/pandas/core/internals/construction.pyc in sanitize_array(data, index, dtype, copy, raise_cast_failure)
    675         if inferred == 'period':
    676             try:
--> 677                 subarr = period_array(subarr)
    678             except IncompatibleFrequency:
    679                 pass

/usr/local/lib/python2.7/site-packages/pandas/core/arrays/period.pyc in period_array(data, freq, copy)
    786     data = ensure_object(data)
    787
--> 788     return PeriodArray._from_sequence(data, dtype=dtype)
    789
    790

/usr/local/lib/python2.7/site-packages/pandas/core/arrays/period.pyc in _from_sequence(cls, scalars, dtype, copy)
    197             periods = periods.copy()
    198
--> 199         freq = freq or libperiod.extract_freq(periods)
    200         ordinals = libperiod.extract_ordinals(periods, freq)
    201         return cls(ordinals, freq=freq)

pandas/_libs/tslibs/period.pyx in pandas._libs.tslibs.period.extract_freq()

/usr/local/lib/python2.7/site-packages/pandas/_libs/tslibs/period.so in View.MemoryView.memoryview_cwrapper()

/usr/local/lib/python2.7/site-packages/pandas/_libs/tslibs/period.so in View.MemoryView.memoryview.__cinit__()

ValueError: buffer source array is read-only

Problem description

I'm currently seeing this issue because xarray is constructing a read-only array of Periods in a .to_dataframe() call.

If pandas isn't expected to support these arrays, we could attempt to prevent this in xarray

Expected Output

Output of pd.show_versions()

In [10]: pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 2.7.15.final.0
python-bits: 64
OS: Darwin
OS-release: 18.2.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: None.None

pandas: 0.24.1
pytest: 4.2.0
pip: 19.0.3
setuptools: 40.6.3
Cython: 0.29.5
numpy: 1.16.1
scipy: 1.2.1
pyarrow: 0.12.0
xarray: 0.11.3
IPython: 5.8.0
sphinx: None
patsy: 0.5.1
dateutil: 2.8.0
pytz: 2018.9
blosc: None
bottleneck: 1.2.1
tables: None
numexpr: 2.6.9
feather: None
matplotlib: 2.2.3
openpyxl: None
xlrd: 1.2.0
xlwt: 1.3.0
xlsxwriter: None
lxml.etree: 4.3.1
bs4: 4.7.1
html5lib: None
sqlalchemy: 1.2.17
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: 0.9.0
pandas_datareader: 0.7.0
gcsfs: None

[jreback]

sounds like the same only cython issue that we had with take
cc @jbrockmendel

[max-sixty]

For reference, we're creating the readonly array with np.broadcast_to here: https://github.com/pydata/xarray/blob/master/xarray/core/variable.py#L1199

[jbrockmendel]

Easy fix: tslibs.period.extract_freq needs to be changed from def extract_freq(object[:] values): to def extract_freq(ndarray[object] values):. If cython ever supports const object[:] that will be the long-term solution.

[jorisvandenbossche]

If this is an easy fix, let's do it for 0.24.2, as this is a regression.