Plot square or cube by default for geopandas Point/MultiPoint types

[yohaimagen]

In response to issue #1373 adding if statement to the plot function to create desirable default behavior for plotting geopandas dataframe with points data(geometry) to be plotted as points instead of a line.

Fixes #1373

[yohaimagen]

I am not very happy with my solution.
first of all, as we discuss in #1373 it involves the plot function which has many many more utilities than just plotting geopandas DataFrams so it seems a bit too specific for this function.
second, I need to make sure that data is geopandas DataFrame before I am checking if it contains points only. In order to do so without importing geopandas I am checking that data contains the attribute _geometry_column_name "_geometry_column_name" in data.__dict__ by doing so I am assuming no other data option(string, np.ndarray etc...) will have this attribute. Another option will be to import geopands and to directly check that it is a geopandas DataFeame instant.

So if you have an idea for better implementation ill be happy to know.

[weiji14]

You've got the right idea here to check for an attribute that only geopandas.GeoDataFrame objects have 😄 I've offered a few suggestions that can help to improve things a bit.

Once we settle on the if-statement condition, it would be good to:

1. copy the same chunk of code to plot3d.py.
2. Add a unit test to test_geopandas.py to ensure this works. Are you interested in learning about unit test writing? If so, we can give you some tips on that.

P.S. I've converted this Pull Request (PR) to draft mode to save on Continuous Integration resources (test will only run on Linux instead of all 3 OS types). We will convert the PR back to non-draft mode once things are more polished.

[yohaimagen]

I am not sure about the p style. at least in my example, you almost can't see the 1-pixel points

from
fig.plot( projection='X10', frame='f', region=[0, 10, 0, 10, 0, 10], data=gdf, )
we get

and from
fig.plot( projection='X10', frame='f', region=[0, 10, 0, 10, 0, 10], data=gdf, style='s0.2c' )
we get

the same goes for plot3d with and without style='u0.2c'

[weiji14]

I am not sure about the p style. at least in my example, you almost can't see the 1-pixel points

Hmm, you're right. Let's go back to square 0.2cm then for plot, and perhaps we should use u0.2c for plot3d as you suggested. Sorry for the back and forth!

[yohaimagen]

it's O.K.

now I'll be happy to learn about the tests. thanks!

[weiji14]

now I'll be happy to learn about the tests. thanks!

Alright! Testing involves checking that code produces the expected behaviour. Specifically, you'll need to write some new test functions to put in pygmt/tests/test_geopandas.py, and ensure that the function will generate an image that matches a baseline. The guidelines are at https://www.pygmt.org/dev/contributing.html#testing-plots.

The test itself can be adapted from what you did in #1405 (comment). You might also find the examples in test_plot.py useful. Without giving you too much of a full answer, it should look something like this:

@pytest.mark.mpl_image_compare
def test_geopandas_plot_default_square():
    """
    """
    point = shapely.geometry.Point(...)
    gdf = gpd.GeoDataFrame(...)
    fig = Figure()
    fig.plot(data=gdf, ...)
    return fig

Start with the above for a 2D plot with a Point type geometry, and once you've got the hang of it, try making a 3D plot with a MultiPoint type geometry. Good luck!

[yohaimagen]

@weiji14 can you please add me as a collaborator at DAGshub.

[michaelgrund]

The dvc image diff test fails. Is that caused by the draft status or went something wrong with the upload @weiji14 ?

[maxrjones]

The dvc image diff test fails. Is that caused by the draft status or went something wrong with the upload @weiji14 ?

dvc image failure is not a problem, see this post for an explanation #1405 (comment).

[seisman]

Please hold off merging this PR and see my comment in #1373 (comment) first.

[seisman]

I think we also need to add two tests to make sure users can plot the symbols using other styles when passing -Sc or others.

[yohaimagen]

I'll add them

[seisman]

Looks good to me.

[yohaimagen]

great thanks for the review!

[seisman]

Ping @weiji14 and @michaelgrund to give this PR another round of final review, because this PR has changed a little after you approved it.

[weiji14]

I think it looks fine, just need to incorporate the suggestions at #1405 (comment) and should be good to merge 😄

[yohaimagen]

Sorry did not notice them, committed the changes.

[weiji14]

Cool, I'll merge this PR in now, thanks again @yohaimagen!