As stated in #25, patchworklib doesn't currently support the current plotnine release.

The issue in place arises from the way plotnine's version is checked, since my understanding is that patchworklib supports plotnine releases starting from 0.8. The next major release (0.9) made some breaking changes which patchworklib handles by doing a simple string check to ensure whether the version is 0.8 or 0.9 so as to handle the ggplot drawing. This has, however, two unintended consequences:

(1) Any plotnine release newer than 0.9 breaks patchworklib. (2) Any plotnine release with substrings containing 0.9 or 0.8 (such as a possible 0.10.8, for instance) will match the incorrect version and will be handled in the wrong manner.

The fix is pretty trivial, and depends on the parse_version function of pkg_resources, which is in itself a dependency of setuptools that is already a dependency of patchworklib, so we aren't adding any extra dependencies.

Comparisons are now made through the properly parsed versions, and any newer versions of plotnine should be handled appropiately now.

By the way, thanks for the very useful library!

Hello!

I am currently working on some code whereby I iteratively generate a list of plots using plotnine and I want to also be able to combine them all into one figure. Depending on when I run this code there may be a variable number of plots, and so I can not use the basic syntax of patchworklib.

So my question is if there is a way with patchworklib to combine a list of plots together, instead of having to specify individual plot objects (ie "plot1 | plot2"). There doesn't seem to be a lot of options to do this using plotnine, but prefer it over the murky waters of matplotlib.

Thanks for any help! Kurt

Dear ponnhide,

I would like firstly to thank you very much for creating and maintaining the patchworklib.

In addition, I would like to comment about an issue I am getting when using it to arrange plotnine graphics. As an example, I will use the same one you have previously shared in: https://pypi.org/project/patchworklib/

-- Code -- import patchworklib as pw from plotnine import * from plotnine.data import * g1 = pw.load_ggplot(ggplot(mtcars) + geom_point(aes("mpg", "disp")) + theme(figure_size=(2, 3))) g2 = pw.load_ggplot(ggplot(mtcars) + geom_boxplot(aes("gear", "disp", group = "gear")) + theme(figure_size=(2, 3))) g12 = g1 | g2 g12.savefig()

-- Output --

AttributeError Traceback (most recent call last) in <cell line: 4>() [2]from plotnine import * [3]from plotnine.data import * ----> [4]g1 = pw.load_ggplot(ggplot(mtcars) + geom_point(aes("mpg", "disp")) + theme(figure_size=(2, 3))) [5]g2 = pw.load_ggplot(ggplot(mtcars) + geom_boxplot(aes("gear", "disp", group = "gear")) + theme(figure_size=(2, 3))) [6]g12 = g1 | g2

File C:\Users\rodrife\Anaconda3\lib\site-packages\patchworklib\patchworklib.py:401, in load_ggplot(ggplot, figsize) [399]ggplot._draw_breaks_and_labels() [400]ggplot._draw_watermarks() --> [401]ggplot._apply_theme() [403]if len(ggplot.axs) == 1: [404]ax = Brick(ax=ggplot.axs[0])

File c:\Users\rodrife\Anaconda3\lib\site-packages\plotnine\ggplot.py:612, in ggplot._apply_theme(self) [608]def _apply_theme(self): [609]""" [610]Apply theme attributes to Matplotlib objects [611]""" --> [612]self.theme.apply_axs(self.axs) [613]self.theme.apply_figure(self.figure)

AttributeError: 'theme_gray' object has no attribute 'apply_axs'

I see it might be trying to apply theme attributes to Matplotlib objects, although in this case there is a plotnine object. However, since this is your example, it may be that I could have something wrong in my environment (e.g. plotnine or patchworklib). In any case, I would really appreciate any helping comment from your side.

Thank you very much in advance! With my very best regards, Fermin

Hello, thank you for your amazing tool so that I can make better pictures for my graduation thesis. I'm now having trouble on turning off/modifying the size of the legend in the combined pictures. I'm not a professional programmer so I'm not sure if it's my problem or if this function is not included in this package. I'll paste my code below:

p1 = (
        ggplot(flight_01, aes(x='t_plot', y='FL1', shape="legend", color="legend"))
        + geom_point(alpha=0.5, size=0.5)
        + scale_x_datetime(labels=date_format('%H:%M'))
        + scale_color_manual(values=['tomato'], breaks=['trajectory'])
        + xlab("time [h]")
        + ylab("altitude x 10^2 [ft]")
        + ggtitle("Flight trajectory "+str(num[0])+"")
        + theme_gray()
        + theme(
                legend_title=element_blank(),
                figure_size=(12, 9),
                legend_position='none',
                legend_box='horizontal',
        )
)

p2 = (
        ggplot(flight_02, aes(x='t_plot', y='FL1', shape="legend", color="legend"))
        + geom_point(alpha=0.5, size=0.5)
        + scale_x_datetime(labels=date_format('%H:%M'))
        + scale_color_manual(values=['tomato'], breaks=['trajectory'])
        # + scale_x_datetime(breaks="30 minute", labels=date_format('%H:%M'))
        + xlab("time [h]")
        + ylab("altitude x 10^2 [ft]")
        + ggtitle("Flight trajectory "+str(num[1])+"")
        + theme_gray()
        + theme(
                legend_title=element_blank(),
                figure_size=(12, 9),
                legend_position='right',
                legend_box='horizontal',
                legend_text=element_text(size=5)
        )
)
print(p1)
print(p2)

p1 = pw.load_ggplot(p1, figsize=(4, 3))
p2 = pw.load_ggplot(p2, figsize=(4, 3))
g1234 = p1|p2
g1234.savefig("G:/DA/combi.png")

As you can see I turn the legend off for p1 and keep only p2. I also turn off the word "legend" for p2 and make the text size smaller so that the width of the combined pic can be thinner.

But in the combined pic g1234, the word "legend" is still there and the size of the legend remains the same as the title, which I didn't change at all. And the pics made by "print" are working as I wanted.

Could you give me some advice on this? Thx a lot!

Seaborn plots generated based on figure-level function cannot be handled as matplotlib subplots. The issue is mentioned in some places (https://github.com/mwaskom/seaborn/issues/2666, https://github.com/mwaskom/seaborn/issues/2059). Now, patchworklib can import the following plots as Bricks class objects and align them with other plots.

• [x] pairplot
• [x] relplot
• [x] jointplot
• [x] catplot
• [x] displot

However, it still cannot handle clustermap plots based on clustermap function.

I've noticed that my original ggplot legend position is not maintained after patching together. I used...

• theme(legend_position=(.5, -0.02), legend_direction='horizontal')

...to position it underneath each individual ggplot before combining with patchworklib. However, after patching together, the legend goes back to the right side for every sub-plot.

Is this related to the same issue as in #15? Thanks.

Hi, thanks for this cool library!

I noticed that several set_*label calls in load_ggplot doesn't pass on any other style settings such as font size so the axis title always ends up being size 12.

Let me know if you need more information on this.

patchworklib has saved me a lot of effort consumed in the layout.

But recently I found that in some special cases it automatically modifies the size of the subgraph.

ls_ax = []
for i in range(10):
    ls_ax.append(pw.Brick(f"ax{i}", (3, 3)))
ls_ax = [ls_ax[:3], ls_ax[3:6], ls_ax[6:9], ls_ax[9]]
_axs = [reduce(lambda x,y: x|y, z) for z in ls_ax[:-1]]
_axs = reduce(lambda x,y: x/y, _axs)
_axs = _axs['ax6'] / ls_ax[-1]

_axs.savefig()

Based on the work of @AlFontal's commit #26 in response to issue #25, replace the final plotnine version check as it still failed when using plotnine's facet_grid and facet_wrap.

The failure can be seen in the 'subplots for plotnine' Colab linked from the ReadMe.

Colab Result:

The failure can also be reproduced with the example below:

import pandas as pd
import patchworklib as pw
import plotnine as p9

df = pd.DataFrame({'a': range(10), 'b': ['x', 'y']*5})

work_around = False
if work_around:
	df1 = df.query('b == "x"')
	g1 = (p9.ggplot(df1, p9.aes(x=df1.index, y='a')) 
		+ p9.geom_line()
		)
	df2 = df.query('b == "y"')
	g2 = (p9.ggplot(df2, p9.aes(x=df2.index, y='a')) 
		+ p9.geom_line()
		)
	
	success1 = pw.load_ggplot(g1)
	success2 = pw.load_ggplot(g2)

	g = success1|success2
	g.savefig('success')
else:
	g = (p9.ggplot(df, p9.aes(x=df.index, y='a', colour='factor(b)')) 
		+ p9.geom_line()
		+ p9.facet_wrap('~b')
		)
	failure = pw.load_ggplot(g)
	# Produces error: Type Error: argument of type 'Version' is not iterable
	print(g)

From my testing it now works with facet_grid and facet_wrap.

Please excuse any incompetence as I am a novice.

Hi, ponnhide. When I use your package and plotnine in spyder IDE, I find that the generated plots are all in the console window and not in the plots pane. But when using only plotnine, the image is displayed in the plots pane.

When I use patchworklib with Plotnine 0.9.0,it reminds me “AttributeError: 'theme_gray' object has no attribute 'apply_axs'”.But I revert plotnine to 0.8.0,patchworklib work normally.

This PR is created to resolve issue #21.

GitHub Actions Workflow for publishing packages to PyPI for release on GitHub.

:warning: To run this workflow, you will need to set up a secret key (PYPI_USERNAME, PYPI_PASSWORD) to access PyPI in advance.

Secrect Key Setting URL: https://github.com/ponnhide/patchworklib/settings/secrets/actions

Hi, it would be very cool to have a conda-forge package of this:

• https://conda-forge.org/docs/user/introduction.html
• https://conda-forge.org/docs/maintainer/adding_pkgs.html#the-staging-process

I created a draft PR here: https://github.com/conda-forge/staged-recipes/pull/21603

Hey,

amazing package making life so much easier :)

I was wondering if it is possible to use patchworklib with PdfPages from matplotlib?

I am generating plots using plotnine, then assemble two side by side using patchworklib and then I would like to export a list of the patchwork brick objects to a multi-page PDF. So far I did not figure out how to properly chain the commands, so I was wondering if it is possible in the first place?

Thanks, Artyom

Would make me more comfortable depending on this package in a production environment if there was a test suite.

I'm not quite sure how to test everything, but some tests that run the patchwork examples would probably go a good distance in catching different errors that arise. The cowpatch tests might be of some inspiration.

You can configure github actions to run the tests on every commit / PR as well as on a scheduled basis to detect for example if a dependency update breaks something. Happy to help set up the actions.

Thanks for making frequent releases of this package, as seen on PyPI. Noting that there are only two git tags currently. Therefore it's hard to know which commits have made it into which releases.

Have you using GitHub Actions to deploy to PyPI? This way whenever you create a release on GitHub at https://github.com/ponnhide/patchworklib/releases/new, it will automatically be deployed to PyPI. You'll get the git tag automatically. Also you can put release notes on the GitHub Release (rather than in the readme).

An example workflow to do this at https://github.com/manubot/manubot/blob/a2ba5780fbbdef9dfc063180b98e6dea500aec48/.github/workflows/release.yml.

Happy to help with this if you're interested! Just found this package and excited for something that can combine python plots!