Potential Bug in matplotlib ax.loglog() and downstream functions?

Question

Im trying to create some graphs which have lots of points towards smaller magnitudes, and covers numbers from 10^5 to about 10^8, so of course im trying to make the axis have a log scale to better see all the points and their distribution.

My problem is, when i use ax.loglog() i dont get consistent behaviour, i.e. other axes have their scale changed. (I have also tried downstream functions like ax.semilogx/semilogy, and ax.set_xscale/set_yscale which have the same behaviour)

Heres an example: Here i plot a bunch of points using sns.scatterplot, and i also plot a lineplot over the top (simply y=x). I plot 4 subplots of this data, and try to set each axis to log scale right after I plot it.

fig, ax = plt.subplots(1,4,sharey=True, figsize=(15,6), sharex=True)


y=variable_to_predict # called NSW Equivalent Emissions
xs=["Prediction (TOBD)","Prediction (TOBD,Sep/TOBD)","Prediction (TOBD,Sep/TOBD,ED/TOBD)","Prediction (TOBD,Sep/TOBD,ED/TOBD,Peer)"]

for i, x in enumerate(xs):
    maxval = 10**8
    sns.lineplot(x=np.linspace(ED_DATA[y].min(),maxval,10),y=np.linspace(ED_DATA[y].min(),maxval,10),
                 color="red", ax=ax[i])

    sns.scatterplot(data=ED_DATA,y=y,x=x,
                    hue = "State", alpha=0.7, linewidth=0, ax=ax[i])
    
    ax[i].loglog()

plt.show()

Whats interesting is, if i only set the last plot/axes to loglog scale, it actually changes all of them in a way that I want! see code below:

fig, ax = plt.subplots(1,4,sharey=True, figsize=(15,6), sharex=True)


y=variable_to_predict # called NSW Equivalent Emissions
xs=["Prediction (TOBD)","Prediction (TOBD,Sep/TOBD)","Prediction (TOBD,Sep/TOBD,ED/TOBD)","Prediction (TOBD,Sep/TOBD,ED/TOBD,Peer)"]

for i, x in enumerate(xs):
    maxval = 10**8
    sns.lineplot(x=np.linspace(ED_DATA[y].min(),maxval,10),y=np.linspace(ED_DATA[y].min(),maxval,10),
                 color="red", ax=ax[i])

    sns.scatterplot(data=ED_DATA,y=y,x=x,
                    hue = "State", alpha=0.7, linewidth=0, ax=ax[i])
    
ax[3].loglog()

plt.show()

If anyone has any insight into whats actually happening in the depths of the matplotlib library I would be very appreciative if you could explain it. I have absolutely no clue, and now, no understanding on how to use the loglog() and related functions the right way.

Thanks in advance!

Potential Bug in matplotlib ax.loglog() and downstream functions?

0 Answers