Unusual layout (GridSpec?)

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Unusual layout (GridSpec?)

Bruno Pagani
Hi there,

I’m trying to combine three graphs into one, but with some specific
constraints on the layout. The three graphs are having the same x and y
variables, but not over the same span. Each graph size should reflects
that (e.g. if a change of 1 on the x axis is 1 cm on one graph, so
should it be for the other ones). Also, I’d like the y label and ticks
labels to be shared amongst the three graphs. Same goes for the colorbar
(those are scatter plots) and the legend.

Attached are the three figures (.pdf), and what the combined version
should look like (target.png, but here it was obtained by removing parts
by hand and stretching everything to fit).

Below is some minimal code to produce three similarly looking graphs, if
someone has an idea (GridSpec?) for this, I should be able to adapt it
to my actual code:

import matplotlib.pyplot as plt
import numpy as np

plt.scatter(np.repeat(np.linspace(0,3.5,15)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],15,axis=0),c=np.random.random((15,11)),label="Some
label")

plt.scatter(np.repeat(np.linspace(0,6,25)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],25,axis=0),c=np.random.random((25,11)),marker='*',label="Some
other label")

plt.scatter(np.repeat(np.linspace(0,11,23)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],23,axis=0),c=np.random.random((23,11)),marker='D',label="Yet
another label")

Regards,
Bruno


_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users

explosion2.pdf (29K) Download Attachment
explosion3.pdf (30K) Download Attachment
explosion4.pdf (25K) Download Attachment
target.png (343K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Unusual layout (GridSpec?)

Paul Hobson-2
I don't think I understand your problem fully, but a GridSpec sounds like it'll fit the bill. We have a tutorial on it at our website. You'll probably get the most mileage out of the width_ratios and height_ratios parameters:

On Thu, Sep 19, 2019 at 12:02 PM Bruno Pagani <[hidden email]> wrote:
Hi there,

I’m trying to combine three graphs into one, but with some specific
constraints on the layout. The three graphs are having the same x and y
variables, but not over the same span. Each graph size should reflects
that (e.g. if a change of 1 on the x axis is 1 cm on one graph, so
should it be for the other ones). Also, I’d like the y label and ticks
labels to be shared amongst the three graphs. Same goes for the colorbar
(those are scatter plots) and the legend.

Attached are the three figures (.pdf), and what the combined version
should look like (target.png, but here it was obtained by removing parts
by hand and stretching everything to fit).

Below is some minimal code to produce three similarly looking graphs, if
someone has an idea (GridSpec?) for this, I should be able to adapt it
to my actual code:

import matplotlib.pyplot as plt
import numpy as np

plt.scatter(np.repeat(np.linspace(0,3.5,15)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],15,axis=0),c=np.random.random((15,11)),label="Some
label")

plt.scatter(np.repeat(np.linspace(0,6,25)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],25,axis=0),c=np.random.random((25,11)),marker='*',label="Some
other label")

plt.scatter(np.repeat(np.linspace(0,11,23)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],23,axis=0),c=np.random.random((23,11)),marker='D',label="Yet
another label")

Regards,
Bruno

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users
Reply | Threaded
Open this post in threaded view
|

Re: Unusual layout (GridSpec?)

Eric Firing
I don't think this is what gridspec is for, or can handle.  Instead, I
think you will have to use a very manual approach to figure out the
positions of the Axes objects that will make them and their tick
locations line up as desired. An example is in the attached script.

Caution: the example will work correctly only if your dpi setting is
such that the specified figsize can be fully displayed on your screen;
if it is too big, mpl will shrink it to fit, and then all the
calculations will be out of whack.

Eric

On 2019/09/19 10:39 AM, Paul Hobson wrote:

> I don't think I understand your problem fully, but a GridSpec sounds
> like it'll fit the bill. We have a tutorial on it at our website. You'll
> probably get the most mileage out of the width_ratios and height_ratios
> parameters:
> https://matplotlib.org/tutorials/intermediate/gridspec.html
>
> On Thu, Sep 19, 2019 at 12:02 PM Bruno Pagani
> <[hidden email] <mailto:[hidden email]>> wrote:
>
>     Hi there,
>
>     I’m trying to combine three graphs into one, but with some specific
>     constraints on the layout. The three graphs are having the same x and y
>     variables, but not over the same span. Each graph size should reflects
>     that (e.g. if a change of 1 on the x axis is 1 cm on one graph, so
>     should it be for the other ones). Also, I’d like the y label and ticks
>     labels to be shared amongst the three graphs. Same goes for the colorbar
>     (those are scatter plots) and the legend.
>
>     Attached are the three figures (.pdf), and what the combined version
>     should look like (target.png, but here it was obtained by removing parts
>     by hand and stretching everything to fit).
>
>     Below is some minimal code to produce three similarly looking graphs, if
>     someone has an idea (GridSpec?) for this, I should be able to adapt it
>     to my actual code:
>
>     import matplotlib.pyplot as plt
>     import numpy as np
>
>     plt.scatter(np.repeat(np.linspace(0,3.5,15)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],15,axis=0),c=np.random.random((15,11)),label="Some
>     label")
>
>     plt.scatter(np.repeat(np.linspace(0,6,25)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],25,axis=0),c=np.random.random((25,11)),marker='*',label="Some
>     other label")
>
>     plt.scatter(np.repeat(np.linspace(0,11,23)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],23,axis=0),c=np.random.random((23,11)),marker='D',label="Yet
>     another label")
>
>     Regards,
>     Bruno
>
>     _______________________________________________
>     Matplotlib-users mailing list
>     [hidden email] <mailto:[hidden email]>
>     https://mail.python.org/mailman/listinfo/matplotlib-users
>
>
> _______________________________________________
> Matplotlib-users mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-users
>

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users

custom_layout.py (1K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Unusual layout (GridSpec?)

Eric Firing
In reply to this post by Paul Hobson-2
Evidently mailing lists like this are crippled by not being able to
handle attachments.  Here is the script inline:
=================================

import numpy as np
import matplotlib.pyplot as plt

def axes_inches(fig, rect, **kw):
     """
     Wrapper for Figure.add_axes in which *rect* is given in inches.
     The translation to normalized coordinates is done immediately
     based on the present figsize.

     *rect* is left, bottom, width, height in inches
     *kw* are passed to Figure.add_axes

     """

     fw = fig.get_figwidth()
     fh = fig.get_figheight()
     l, b, w, h = rect
     relrect = [l / fw, b / fh, w / fw, h / fh]
     ax = fig.add_axes(relrect, **kw)
     return ax

# We will assume that the x and y data limits all start with 0.

# The actual data limits:
xranges = np.array([[0, 15], [0, 25], [0, 23]])
yranges = np.array([[0, 0.5], [0, 0.3], [0, 0.2]])

# Add margins:
xmranges = xranges + np.array([-0.5, 0.5])
ymranges = yranges + np.array([-0.05, 0.05])

# Scales in data units per inch:
xscale = 10  # so the max, 25, is 2.5 inches
yscale = 0.1 # so the max, 0.5, is 5 inches

# Separation in inches:
xsep = 0.4

# Force the y tick locations:
yticks = np.arange(0, 0.501, 0.1)

fig = plt.figure(figsize=(8.5, 7))
axs = []
left, bottom = 0.5, 0.5  # Starting point in inches.
for xm, ym in zip(xmranges, ymranges):
     w = (xm[1] - xm[0]) / xscale
     h = (ym[1] - ym[0]) / yscale
     ax = axes_inches(fig, (left, bottom, w, h))
     ax.set_xlim(xm)
     ax.set_ylim(ym)
     ax.set_yticks(yticks[yticks < ym[-1]])
     axs.append(ax)
     left += (xsep + w)

plt.show()

============================
Eric

On 2019/09/19 10:39 AM, Paul Hobson wrote:

> I don't think I understand your problem fully, but a GridSpec sounds
> like it'll fit the bill. We have a tutorial on it at our website. You'll
> probably get the most mileage out of the width_ratios and height_ratios
> parameters:
> https://matplotlib.org/tutorials/intermediate/gridspec.html
>
> On Thu, Sep 19, 2019 at 12:02 PM Bruno Pagani
> <[hidden email] <mailto:[hidden email]>> wrote:
>
>     Hi there,
>
>     I’m trying to combine three graphs into one, but with some specific
>     constraints on the layout. The three graphs are having the same x and y
>     variables, but not over the same span. Each graph size should reflects
>     that (e.g. if a change of 1 on the x axis is 1 cm on one graph, so
>     should it be for the other ones). Also, I’d like the y label and ticks
>     labels to be shared amongst the three graphs. Same goes for the colorbar
>     (those are scatter plots) and the legend.
>
>     Attached are the three figures (.pdf), and what the combined version
>     should look like (target.png, but here it was obtained by removing parts
>     by hand and stretching everything to fit).
>
>     Below is some minimal code to produce three similarly looking graphs, if
>     someone has an idea (GridSpec?) for this, I should be able to adapt it
>     to my actual code:
>
>     import matplotlib.pyplot as plt
>     import numpy as np
>
>     plt.scatter(np.repeat(np.linspace(0,3.5,15)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],15,axis=0),c=np.random.random((15,11)),label="Some
>     label")
>
>     plt.scatter(np.repeat(np.linspace(0,6,25)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],25,axis=0),c=np.random.random((25,11)),marker='*',label="Some
>     other label")
>
>     plt.scatter(np.repeat(np.linspace(0,11,23)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],23,axis=0),c=np.random.random((23,11)),marker='D',label="Yet
>     another label")
>
>     Regards,
>     Bruno
>
>     _______________________________________________
>     Matplotlib-users mailing list
>     [hidden email] <mailto:[hidden email]>
>     https://mail.python.org/mailman/listinfo/matplotlib-users
>
>
> _______________________________________________
> Matplotlib-users mailing list
> [hidden email]
> https://mail.python.org/mailman/listinfo/matplotlib-users
>

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users
Reply | Threaded
Open this post in threaded view
|

Re: Unusual layout (GridSpec?)

Bruno Pagani
In reply to this post by Eric Firing
Hi there,

Thank you both for your advices. I was hoping for a GridSpec solution
because I think such a one would have been able to leverage
constrained_layout to automatically fit the space correctly. There might
exists one, but I did not had a lot of time for researching.

Instead, I went with Eric’s PoC that I was able to tweak nicely in order
to suit my needs (btw Eric, the attached script was perfectly received
on my end). The end result is attached to this email. So many thanks for
that piece of code Eric, it proved to be very helpful. ;)

Regards,
Bruno Pagani

Le 20/09/2019 à 00:01, Eric Firing a écrit :

> I don't think this is what gridspec is for, or can handle.  Instead, I
> think you will have to use a very manual approach to figure out the
> positions of the Axes objects that will make them and their tick
> locations line up as desired. An example is in the attached script.
>
> Caution: the example will work correctly only if your dpi setting is
> such that the specified figsize can be fully displayed on your screen;
> if it is too big, mpl will shrink it to fit, and then all the
> calculations will be out of whack.
>
> Eric
>
> On 2019/09/19 10:39 AM, Paul Hobson wrote:
>> I don't think I understand your problem fully, but a GridSpec sounds
>> like it'll fit the bill. We have a tutorial on it at our website.
>> You'll probably get the most mileage out of the width_ratios and
>> height_ratios parameters:
>> https://matplotlib.org/tutorials/intermediate/gridspec.html
>>
>> On Thu, Sep 19, 2019 at 12:02 PM Bruno Pagani
>> <[hidden email] <mailto:[hidden email]>>
>> wrote:
>>
>>     Hi there,
>>
>>     I’m trying to combine three graphs into one, but with some specific
>>     constraints on the layout. The three graphs are having the same x
>> and y
>>     variables, but not over the same span. Each graph size should
>> reflects
>>     that (e.g. if a change of 1 on the x axis is 1 cm on one graph, so
>>     should it be for the other ones). Also, I’d like the y label and
>> ticks
>>     labels to be shared amongst the three graphs. Same goes for the
>> colorbar
>>     (those are scatter plots) and the legend.
>>
>>     Attached are the three figures (.pdf), and what the combined version
>>     should look like (target.png, but here it was obtained by
>> removing parts
>>     by hand and stretching everything to fit).
>>
>>     Below is some minimal code to produce three similarly looking
>> graphs, if
>>     someone has an idea (GridSpec?) for this, I should be able to
>> adapt it
>>     to my actual code:
>>
>>     import matplotlib.pyplot as plt
>>     import numpy as np
>>
>>    
>> plt.scatter(np.repeat(np.linspace(0,3.5,15)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],15,axis=0),c=np.random.random((15,11)),label="Some
>>     label")
>>
>>    
>> plt.scatter(np.repeat(np.linspace(0,6,25)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],25,axis=0),c=np.random.random((25,11)),marker='*',label="Some
>>     other label")
>>
>>    
>> plt.scatter(np.repeat(np.linspace(0,11,23)[:,None],11,axis=1),np.repeat(np.linspace(0,0.5,11)[None,:],23,axis=0),c=np.random.random((23,11)),marker='D',label="Yet
>>     another label")
>>
>>     Regards,
>>     Bruno

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users

explosions.pdf (46K) Download Attachment