Drops Py3.7, improves typing and adds a mypy config, improves the docstrings and isorts the imports.
* configure isort
* sort with isort
* add github action
* enable flake8-docstrings
* fix docstrings
* add mypy env
* no implicit optional
* add type for raw
* add some type hints
Python 2.7 EOL is arriving on 2020-01-01: https://pythonclock.org/
This will allow us to use Python 3 only libraries, like python-barcode,
which can maintain a reduced, simpler codebase, due to only one version
to support.
Closes#371.
Signed-off-by: Romain Porte <microjoe@microjoe.org>
* drop python 2 from supported languages
Python 2 compatibility will not be actively revoked, but will
not be worked on.
* update travis config
* remove flake8-test with Py2.7
* add explicit location of capabilities.json for travis
* pass on env variables in tox builds
* drop support for python 3.3 and remove python3.3 from supported versions list
The feature use_coverage of hypothesis caused the failing tests, because
the printer_file_test is sensitive to the coverage analysis of
hypothesis.
Fixed by disabling use_coverage for the crashing tests
hypothesis introduced a regression in 3.29.0 (or at least changed
behaviour). Until I have found the problem I will pin it to the last
working version.
Also two tests should be prevented from failing when they are slow.
Actually the hardware barcode implementation is very specific and not
generic enough for just adding a `soft_render=True` argument to it. This
is a first work that can be improved with other commits, maybe for
merging this method in the `barcode` method after some cleanup.
The width, height and text_distance were set using empiric
print-and-retry tests so that the generated barcode looks nice to the
eye (and to the eye of an Android scanner tool.
!WARNING! Printing a barcode that is too large in width will result in
the printer to go crazy trying to print an image that is too large for
it. This may be fixed by raising an exception in the `image` method.
jaconv is available for more Python-versions and seems to be more
professional. Apart from that I added jaconv to the test-requirements
but not the requirements. (If the katakana-stuff really works we can
later add it as a real dependency)
The version string is in the module as __version__ available.
In the doc the version will be automatically parsed. The version comes
from the installed module if on read the docs or directly from
setuptools_scm if you are working locally.
The CLI will issue the version string if you call it with the option
'version'. The CLI does not accept commands like '--version', since this
would not be conform with the rest of the interface (and argparse).
The configuration for loading the version-string is adapted from
pimutils/vdirsyncer. It autogenerates a version string setuptools_scm at
install-time and then adds it to the __version__ member in __init__.py
I adapted the GitHub-template with a fitting comment and bumped the
changelog.
closes#141