Compatibility with Python 3.7.0b5

[rth]

When running the tests suite on Python 3.7.0b5, with

Dockerfile

FROM python:3.7-rc

RUN apt-get update && apt-get install -y build-essential git make

RUN apt-get install -y libopenblas-dev

RUN python --version

RUN pip3 --version

RUN pip3 install Cython nose pytest numpy
RUN pip3 install scipy

RUN git clone --depth 1 https://github.com/scikit-learn/scikit-learn.git && \
    cd scikit-learn && git checkout master


RUN cd scikit-learn && pip3 install -e .

RUN cd scikit-learn && pytest sklearn

most tests appears to pass, but there are two minor doctest failures,

________________ [doctest] sklearn.exceptions.FitFailedWarning _________________
106     Examples
107     --------
108     >>> from sklearn.model_selection import GridSearchCV
109     >>> from sklearn.svm import LinearSVC
110     >>> from sklearn.exceptions import FitFailedWarning
111     >>> import warnings
112     >>> warnings.simplefilter('always', FitFailedWarning)
113     >>> gs = GridSearchCV(LinearSVC(), {'C': [-1, -2]}, error_score=0, cv=2)
114     >>> X, y = [[1, 2], [3, 4], [5, 6], [7, 8]], [0, 0, 1, 1]
115     >>> with warnings.catch_warnings(record=True) as w:
Expected:
    FitFailedWarning('Estimator fit failed. The score on this train-test
    partition for these parameters will be set to 0.000000.
    Details: \nValueError: Penalty term must be positive; got (C=-2)\n',)
Got:
    FitFailedWarning('Estimator fit failed. The score on this train-test partition for these parameters will be set to 0.000000. Details: \nValueError: Penalty term must be positive; got (C=-2)\n')

/scikit-learn/sklearn/exceptions.py:115: DocTestFailure
_________________ [doctest] sklearn.exceptions.NotFittedError __________________
019 Exception class to raise if estimator is used before fitting.
020 
021     This class inherits from both ValueError and AttributeError to help with
022     exception handling and backward compatibility.
023 
024     Examples
025     --------
026     >>> from sklearn.svm import LinearSVC
027     >>> from sklearn.exceptions import NotFittedError
028     >>> try:
Expected:
    NotFittedError('This LinearSVC instance is not fitted yet',)
Got:
    NotFittedError('This LinearSVC instance is not fitted yet')

Also there is a DeprecationWarning when running import sklearn reported in #11121

[amueller]

That is a weird change in the standard exception __repr__. That's what this is, right?

[rth]

Yes, seems due to https://bugs.python.org/issue30399 that was resolved - not sure why we still get this in 3.7.0b5...

[rth]

No actually that issue / PR is where the behavior changed: in 3.7 it seems there is no trailing comma in Exceptions __repr__ anymore.

[wenhaoz-fengcai]

Can I take this issue as my first contribution?

My idea of fixing this would be just removing the trailing comma in NotFittedError('This LinearSVC instance is not fitted yet',) and Details: \nValueError: Penalty term must be positive; got (C=-2)\n',).

[rth]

Yes, thank you @jiaowoshabi !

[wenhaoz-fengcai]

@rth Hi, since Exceptions __rep__ in 3.6.5 still automatically adds the trailing comma at the end, test run failed on Python3.6.5 after simply removing the trailing comma.

How should I handle this backward compatibility if test run needs to succeed on 3.6.5?

[jnothman]

Since we don't have many doctests which raise exceptions, I think using doctest's ELLIPSIS is fine.

…

On Thu, 14 Jun 2018 at 11:55, Wenhao Zhang ***@***.***> wrote: @rth <https://github.com/rth> Hi, since Exceptions __rep__ in 3.6.5 still automatically adds the trailing comma at the end, test run failed on Python3.6.5 after simply removing the trailing comma. How should I handle this backward compatibility if test run needs to succeed on 3.6.5? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#11224 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz6xyNLs_ObrFXb17slIpbf_AA_0aLks5t8cKCgaJpZM4UhS6u> .