0.4.post1: test failures

Hi,

I am just trying to update the Debian package to 0.4.post1, but get a number of test failures due to limited accuracy:

platform linux -- Python 3.7.3, pytest-3.10.1, py-1.8.0, pluggy-0.8.0 -- /usr/bin/python3
cachedir: .pytest_cache
Matplotlib: 3.0.2
Freetype: 2.9.1

Running tests with astroMl version 0.4.post1.
Running tests in astroML.

Date: 2019-07-31T15:28:23

Platform: Linux-4.19.0-5-amd64-x86_64-with-debian-bullseye-sid

Executable: /usr/bin/python3

Full Python Version: 
3.7.3 (default, Apr  3 2019, 05:39:12) 
[GCC 8.3.0]

encodings: sys: utf-8, locale: UTF-8, filesystem: utf-8
byteorder: little
float info: dig: 15, mant_dig: 15

Numpy: 1.16.2
Scipy: 1.1.0
Matplotlib: 3.0.2
Cython: 0.29.2
Astropy: 3.1.2
scikit-learn: 0.20.2
astroML_addons: not available
astropy_helpers: 3.1.1
Using Astropy options: remote_data: none.

rootdir: /home/oles/Projects/2011/debian/python/astropy/astroml, inifile: setup.cfg
plugins: remotedata-0.3.1, openfiles-0.3.2, mpl-0.10, doctestplus-0.2.0, arraydiff-0.3
[…]
=================================== FAILURES ===================================
___________ [doctest] astroML.datasets.LIGO_bigdog.fetch_LIGO_bigdog ___________
084         The data is 10 seconds of measurements from three sites, along with
085         the time of each measurement.
086 
087     Examples
088     --------
089     >>> from astroML.datasets import fetch_LIGO_bigdog
090     >>> data = fetch_LIGO_bigdog()  # doctest: +IGNORE_OUTPUT
091     >>> print(data.dtype.names)
092     ('t', 'Hanford', 'Livingston', 'Virgo')
093     >>> print(data['t'][:3])
Expected:
    [  0.00000000e+00   6.10400000e-05   1.22070000e-04]
Got:
    [0.0000e+00 6.1040e-05 1.2207e-04]

/home/oles/Projects/2011/debian/python/astropy/astroml/astroML/datasets/LIGO_bigdog.py:93: DocTestFailure
_____ [doctest] astroML.datasets.sdss_S82standards.fetch_sdss_S82standards _____
099        Herr              H-band error
100        K                 K-band magnitude
101        Kerr              K-band error
102        theta             difference between SDSS and 2MASS position (arcsec)
103 
104     Examples
105     --------
106     >>> data = fetch_sdss_S82standards()  # doctest: +IGNORE_OUTPUT
107     >>> u_g = data['mmed_u'] - data['mmed_g']
108     >>> print(u_g[:4])
Expected:
    [-22.23500061   1.34900093   1.43799973   2.08200073]
Got:
    [-22.235       1.3490009   1.4379997   2.0820007]

/home/oles/Projects/2011/debian/python/astropy/astroml/astroML/datasets/sdss_S82standards.py:108: DocTestFailure
_________ [doctest] astroML.datasets.sdss_specgals.fetch_sdss_specgals _________
073 
074     Examples
075     --------
076     >>> from astroML.datasets import fetch_sdss_specgals
077     >>> data = fetch_sdss_specgals()  # doctest: +IGNORE_OUTPUT
078     >>> data.shape  # number of objects in dataset
079     (661598,)
080     >>> data.dtype.names[:5]  # first five column names
081     ('ra', 'dec', 'mjd', 'plate', 'fiberID')
082     >>> print(data['ra'][:3])  # first three RA values
Expected:
    [ 146.71419105  146.74414186  146.62857334]
Got:
    [146.71419105 146.74414186 146.62857334]

/home/oles/Projects/2011/debian/python/astropy/astroml/astroML/datasets/sdss_specgals.py:82: DocTestFailure
_____________ [doctest] astroML.datasets.sdss_sspp.fetch_sdss_sspp _____________
095 
096     Examples
097     --------
098     >>> from astroML.datasets import fetch_sdss_sspp
099     >>> data = fetch_sdss_sspp()  # doctest: +IGNORE_OUTPUT
100     >>> data.shape  # number of objects in dataset
101     (327260,)
102     >>> print(data.dtype.names[:5])  # names of the first five columns
103     ('ra', 'dec', 'Ar', 'upsf', 'uErr')
104     >>> print(data['ra'][:1])  # first RA value
Expected:
    [49.6275024]
Got:
    [49.627502]

/home/oles/Projects/2011/debian/python/astropy/astroml/astroML/datasets/sdss_sspp.py:104: DocTestFailure
_____ [doctest] astroML.density_estimation.empirical.EmpiricalDistribution _____
065     ----------
066     data : one-dimensional array
067         input data
068 
069     Examples
070     --------
071     >>> import numpy as np
072     >>> np.random.seed(0)
073     >>> x = np.random.normal(size=10000)  # normally-distributed variables
074     >>> x.mean(), x.std()
Expected:
    (-0.018433720158265783, 0.98755656817612003)
Got:
    (-0.018433720158265818, 0.9875565681761208)

/home/oles/Projects/2011/debian/python/astropy/astroml/astroML/density_estimation/empirical.py:74: DocTestFailure
___________________ [doctest] astroML.utils.utils.logsumexp ____________________
011 
012     Returns log(sum(exp(arr))) while minimizing the possibility of
013     over/underflow.
014 
015     Examples
016     --------
017 
018     >>> import numpy as np
019     >>> a = np.arange(10)
020     >>> np.log(np.sum(np.exp(a)))
Expected:
    9.4586297444267107
Got:
    9.45862974442671

[olebole]

I can confirm that it works with doctestplus 0.3.0. I’ll update the dependencies accordingly, thanks.

[bsipocz]

@olebole - is there any chance to upgrade pytest-doctestplus? Locally, using 0.3.0 I see no failures, while with 0.2.0 I can reproduce the failures from above.