colorbar without overlap

[SteffenMeinecke]

Dear @kravitsjacob,
thank you very much for this small but smart python package.

I try to use it as explained in the pandas example. Unfortunately, with my data set I get an overlap of the figure with the colormap. How can that be solved? (In other contexts plt.tight_layout() may help but not here.)
A solution increasing the width as much as needed, e.g. paxfig.set_figwidth(18), is undesired since it needs to much space.

The example code:

import pandas as pd
import matplotlib.pyplot as plt
import paxplot

json_str = '{"$v_{m_{0}}$":{"0":1.05101414,"1":0.97541944,"2":1.02250952,"3":1.05724531,"4":0.91881114,"5":1.00289782,"6":1.02356087,"7":0.91879598,"8":0.91856527,"9":1.05471083,"10":1.03512956,"11":1.01495961,"12":0.99814463,"13":1.08194644,"14":1.00435071,"15":0.91008701,"16":1.02191995,"17":1.0104185,"18":1.05577619,"19":1.03793584},"$v_{m_{1}}$":{"0":1.08675354,"1":0.99289773,"2":1.0006575,"3":1.03279003,"4":1.04013416,"5":1.00186099,"6":0.94705512,"7":1.08785596,"8":0.9901675,"9":1.03153871,"10":1.00552627,"11":0.97142502,"12":1.0635523,"13":0.94439644,"14":1.04267524,"15":1.00328082,"16":0.96776328,"17":1.05307138,"18":1.06106298,"19":1.04153575},"$v_{m_{2}}$":{"0":1.06366783,"1":0.91104843,"2":1.06438557,"3":1.00671504,"4":1.05875128,"5":0.97205114,"6":1.05476759,"7":0.99542596,"8":0.95052276,"9":1.02192044,"10":0.9318766,"11":0.92399757,"12":1.01605039,"13":1.0046992,"14":1.02006368,"15":1.04490573,"16":0.95427917,"17":0.92809566,"18":1.01760408,"19":1.06606671},"$v_{m_{3}}$":{"0":1.07216311,"1":0.97654954,"2":1.03222464,"3":0.96508126,"4":0.92882952,"5":0.9553515,"6":1.02212038,"7":0.99203902,"8":1.0135808,"9":1.07826395,"10":1.0875509,"11":1.02785425,"12":1.03946075,"13":1.04979572,"14":1.04030671,"15":1.07928366,"16":0.94665728,"17":1.02104275,"18":1.04965773,"19":0.97794591},"$v_{m_{4}}$":{"0":0.96261164,"1":0.98230649,"2":0.99018279,"3":0.91334141,"4":1.0495412,"5":1.04518659,"6":1.07116521,"7":1.00761336,"8":0.95882059,"9":1.05051721,"10":0.95355408,"11":0.94120487,"12":0.91258196,"13":1.01900706,"14":0.94362856,"15":1.03923165,"16":0.99024211,"17":0.95139986,"18":1.00734866,"19":0.98286251},"results":{"0":-33488.121893849,"1":3980.888057873,"2":-11523.4361694497,"3":16582.4829596293,"4":-24886.4934994639,"5":-2551.3204885725,"6":-1097.4260832155,"7":4460.7742568187,"8":441.248979536,"9":-435.2766055767,"10":-19832.8660960374,"11":-19358.1526127029,"12":-37944.4309613923,"13":-12791.378116259,"14":-28921.1231681309,"15":3990.7823822596,"16":1037.0001667862,"17":-9150.0172480353,"18":-11273.0333321415,"19":22697.3994558519}}'

# Import data
df = pd.read_json(json_str)
cols = df.columns

# Create figure
paxfig = paxplot.pax_parallel(n_axes=len(cols))
paxfig.plot(df.to_numpy())

# Add labels
paxfig.set_labels(cols)

# Add colorbar
color_col = 5
paxfig.add_colorbar(
    ax_idx=color_col,
    cmap='viridis',
    colorbar_kwargs={'label': cols[color_col]}
)

plt.show()

[kravitsjacob]

Hey Steffan,

I hope you are well. Thank you for the great question! Here is a hacky workaround using the currently unsupported subplots_adjust function (gasp):

import pandas as pd
import matplotlib.pyplot as plt
import paxplot

json_str = '{"$v_{m_{0}}$":{"0":1.05101414,"1":0.97541944,"2":1.02250952,"3":1.05724531,"4":0.91881114,"5":1.00289782,"6":1.02356087,"7":0.91879598,"8":0.91856527,"9":1.05471083,"10":1.03512956,"11":1.01495961,"12":0.99814463,"13":1.08194644,"14":1.00435071,"15":0.91008701,"16":1.02191995,"17":1.0104185,"18":1.05577619,"19":1.03793584},"$v_{m_{1}}$":{"0":1.08675354,"1":0.99289773,"2":1.0006575,"3":1.03279003,"4":1.04013416,"5":1.00186099,"6":0.94705512,"7":1.08785596,"8":0.9901675,"9":1.03153871,"10":1.00552627,"11":0.97142502,"12":1.0635523,"13":0.94439644,"14":1.04267524,"15":1.00328082,"16":0.96776328,"17":1.05307138,"18":1.06106298,"19":1.04153575},"$v_{m_{2}}$":{"0":1.06366783,"1":0.91104843,"2":1.06438557,"3":1.00671504,"4":1.05875128,"5":0.97205114,"6":1.05476759,"7":0.99542596,"8":0.95052276,"9":1.02192044,"10":0.9318766,"11":0.92399757,"12":1.01605039,"13":1.0046992,"14":1.02006368,"15":1.04490573,"16":0.95427917,"17":0.92809566,"18":1.01760408,"19":1.06606671},"$v_{m_{3}}$":{"0":1.07216311,"1":0.97654954,"2":1.03222464,"3":0.96508126,"4":0.92882952,"5":0.9553515,"6":1.02212038,"7":0.99203902,"8":1.0135808,"9":1.07826395,"10":1.0875509,"11":1.02785425,"12":1.03946075,"13":1.04979572,"14":1.04030671,"15":1.07928366,"16":0.94665728,"17":1.02104275,"18":1.04965773,"19":0.97794591},"$v_{m_{4}}$":{"0":0.96261164,"1":0.98230649,"2":0.99018279,"3":0.91334141,"4":1.0495412,"5":1.04518659,"6":1.07116521,"7":1.00761336,"8":0.95882059,"9":1.05051721,"10":0.95355408,"11":0.94120487,"12":0.91258196,"13":1.01900706,"14":0.94362856,"15":1.03923165,"16":0.99024211,"17":0.95139986,"18":1.00734866,"19":0.98286251},"results":{"0":-33488.121893849,"1":3980.888057873,"2":-11523.4361694497,"3":16582.4829596293,"4":-24886.4934994639,"5":-2551.3204885725,"6":-1097.4260832155,"7":4460.7742568187,"8":441.248979536,"9":-435.2766055767,"10":-19832.8660960374,"11":-19358.1526127029,"12":-37944.4309613923,"13":-12791.378116259,"14":-28921.1231681309,"15":3990.7823822596,"16":1037.0001667862,"17":-9150.0172480353,"18":-11273.0333321415,"19":22697.3994558519}}'

# Import data
df = pd.read_json(json_str)
cols = df.columns

# Create figure
paxfig = paxplot.pax_parallel(n_axes=len(cols))
paxfig.plot(df.to_numpy())

# Add labels
paxfig.set_labels(cols)

# Add colorbar
plt.subplots_adjust(right=0.8)  # Adding whitespace
color_col = 5
paxfig.add_colorbar(
    ax_idx=color_col,
    cmap='viridis',
    colorbar_kwargs={
        'label': cols[color_col],
        'fraction': 0.65,  # moving colorbar
        'anchor': (4.5, 1.0),  # moving colorbar
    }
)
paxfig.set_size_inches(8, 4)

plt.show()

Long-term, it would be great to actually address the overlapping colorbar/legend issue. Under the hood, paxplot creates an additional subplot but in the case of long tick labels on the last axis (in your example) this approach yields the overlapping you describe. Additionally, support for tight_layout would also be great, as you mentioned.

If you or anyone reading this thread wants to take on either of these issues, I would be happy to support your efforts!

[SteffenMeinecke]

Thanks @kravitsjacob for your answers and ideas! I would let the issue open to allow others to take up the ideas and contributing.