Watermark Versions Save

• Adds a new --gpu flag to print out GPU information (currently limited to NVIDIA devices) (#90, via contribution by 907Resident)

• Upper limit on importlib-metadata caused trouble installing on Python 3.7. Instead pin to minimum version with Python 3.8 functionality according to https://github.com/python/importlib_metadata#compatibility (#86, via contribution by James Myatt)

• Added the following arguments: --github_username - for prints author github username, --email - for prints author email, --website - for prints author or project website. (#82, via contribution by joschkazj)
• Added a --conda option to print the name of the current conda environment. (#79, via contribution by Alexander Krasnikov)
• It is now possible to inject globals when used outside IPython (#80, via contribution by Hugo Lapré). For example, version numbers of imported packages can now be obtained as follows:

import numpy
import watermark.watermark as watermark


watermark(iversions=True, globals_=globals())

• Refactoring such that a watermark() function now also works outside IPython and Jupyter. Now, the magic %watermark calls watermark(). Calling %watermark via IPython and Juypter still works as usual. However, in addition the watermark() function can be used within regular Python for unit testing purposes etc. (#46, via contribution by Tymoteusz Wołodźko)

• Support VERSION attributes, in addition to __version__ attributes.

• Now uses pkg_resources as the default method for getting version numbers.
• Fixes a whitespace bug when printing the timezone.

• When no Python library was imported and the --iversion is used, print an empty string instead of raising an error.

• Fixes string alignment issues when the -iv/--iversion flag is used.

• The -iv/--iversion flag now also shows package versions that were imported as from X import Y and import X.Y as Y. For example,

import scipy as sp
from sklearn import metrics
import numpy.linalg as linalg

%watermark --iversions

will return

scipy     1.1.0
sklearn   0.20.1
numpy     1.15.4

(Via contribution by James Myattt)

• Shows "not installed" for version of packages/modules that cannot be imported.
• Shows "unknown" for version of packages/modules when version attribute cannot be found.
• Add Python 3.6 and 3.7 to Travis CI builds.
• Add classifiers to setuptools configuration.