matplotlib / grid-strategy Goto Github PK

We need to add an explicit dependency on matplotlib for now (we can drop it later if it becomes optional because we've added another plotting backend).

It should be added under [options] in setup.cfg, it is under the install_requires section.

Currently the tests are emitting some warnings, that should be causing the tests to fail.

I think the setuptools, attrs or variants configurations work well for this.

When specifying the justification of a GraphStrategy, the spacing is no spread out far enough and it overlaps with other graphs.

Per my comments in #50, I think we should probably create a version of SquareStrategy that allows you to specify a target aspect ratio other than 1:1.

The nomenclature is all messed up and I'm wondering if we should rename this project and a bunch of the classes in it before too many people start using it, but for now we can just call it RectangleStrategy, and make SquareStrategy a subclass of it.

RectangularStrategy, if get_grid takes in prime N, creates a one-row by N cols plots grid. This poses an issue with larger prime numbers (N > 3)

In our setup.cfg, we should have python_requires = >=3.6

I believe you can use the quickstart utility from sphinx.

We may also want to add a tox env for building the docs.

There's at least one property test that I think would be useful: we can assert that:

def test_arrangement(strategy, n):
    sum(strategy.get_grid_arrangement(n)) == n

Getting a grid arrangement that has the number of plots we want in it.

We just need to add hypothesis to the test dependencies and design an appropriate (hypothesis) strategy.

Hi,

I'm very new to matplotlib and I'm creating figures in a loop and I'd like to arrange them nicely.

Can someone add example code to the README that shows how to apply the strategies in this package?

Use setup.cfg for everything if possible.

Best to base our tox configuration on the exemplary repos listed in the wiki.

I believe we should set isolated_build=True.

Tox documentation

Right now this only works with matplotlib, but it would be good to decouple grid-strategy from that a bit and make it possible to add additional backends.

Specifically, I'd like to support bqplot as well.

We should format all the code with black, this means:

1. Add black's configuration to the pyproject.toml
2. Add a tox environment that runs black on the code
3. Add a CI job that ensures that all the code has been formatted with black.

Can we add an insert function to let users insert their premade figures into specific grids?

This project came up on the discourse. I'm just curious if it still works well with the constrained_layout rewrite?

We do not currently have any xfail-ing tests, but we should still configure pytest to treat all XPASSes as failures.

We should have a few simple examples in the documentation, generated inline with the matplotlib plotting extensions for sphinx.

For example, we can use this to generate one of the main figures:

.. plot::

   import matplotlib.pyplot as plt
   from grid_strategy.strategies import SquareStrategy

   specs = SquareStrategy('center').get_grid(7)

   for i, spec in enumerate(specs):
      ax = plt.subplot(spec)

      ax.text(0.5, 0.5, f"Subplot: {i}", color='white',
             fontweight='bold', va="center", ha="center")
      ax.tick_params(axis='both', bottom=False, top=False, left=False,
                    right=False, labelbottom=False, labelleft=False)

      ax.patch.set_facecolor('#2E2E2E')

   plt.show()

Overall this looks fun.

Super happy you are using constrained layout, but suggest you make it optional somehow. There are some times when it'll fail and users will want a way to turn it off.

I'm also not sure about initializing the figure inside of your helper, instead of letting the user pass it (optionally?).

Once we have tox and CI set up, we would also like to start checking the code coverage automatically. See the repos to steal from, dateutil and setuptools both have code coverage configured through tox.

We should have tests/test_strategies.py, using pytest with parametrized tests.

I think we want to generate it from some Apache 2.0 boilerplate website.

Right now "fully justified" changes the shape of the plots, we could change that to two different justification modes. In the original MEP, I proposed two "fully justified" modes, fill-grow and fill-space, where fill-grow would have the plots grow to fill the space and fill-space would distribute the extra space between the plots.

Probably Azure Pipelines and/or Travis CI. See the repos to steal from section - tox, virtualenv and dateutil all have strong Azure Pipelines pipelines, but note!

We only support Python >= 3.6, so we can get away with a much less elaborate set of pipelines and jobs. 3.6, 3.7 and 3.8-dev (allowed failure), on Linux, Mac and Windows I think should be sufficient.

See the badges on our repos to steal from, we should add:

• Docs badge
• Azure pipelines badge
• PyPI badge (don't worry it won't work until we make a release)
• Codecov (though that can be added as part of the "add code cov" PR)

A lot of this can probably be cribbed from MEP 30

Should be a list of all the authors who have contributed, see the comparable list for dateutil; it is unnecessary to include the dateutil stuff about dual licensing and whatnot, but otherwise the format can be similar.

Currently I'm just globbing *.tar.gz and *.whl, maybe an assertion that exactly 2 things were found would be prudent.

We should make GridStrategy public.

We should also add a docstring and an __all__ = ['GridStrategy'] attribute as well.

We can crib most of this from dateutil.