ChaoYue wrote:
> Hi Martin,
>
> I don't know tight_layout quite well. Probably you could also split the
> handlers of the barplot into
> and 2 or 3 or 4 parts depending on the number, and then show them in sperate
> axes?
>
> then you create n+1 subplots for the whole figure?
No, the reason why I use stacked bar chart is that I can squeeze the data
into a single figure. It is only difficult for the reader to compare 20
separate barcharts between each other whereas seeing 20 stacked bars in a
single chart on top of each other is easy.
I think your approach combined with something like a future improvement:
fig.savefig('foobar.png', bbox_inches='tight', keep_fig_width=True)
will do the best for me (unused space below the legend will be chopped away
while figure width will be untouched).
Lets see what Ben or others say about the tight_layout rescaling issue and the
"feature request". ;-)
Maritn
>
> probably this is quite stupid.
>
> cheers,
>
> Chao
>
> On Wed, May 22, 2013 at 1:03 PM, Martin Mokrejs [via matplotlib] <[hidden
> email] </user/SendEmail.jtp?type=node&node=41105&i=0>> wrote:
>
> Hi Chao,
>
> ChaoYue wrote:
> > Dear Martin,
> >
> > I worked out a similar example for your reference as I don't catch your
> example very well.
>
> I think you got the idea quite well.
>
> >
> > fig = plt.figure()
>
> > ax1 = fig.add_subplot(211)
>
> > ax2 = fig.add_subplot(212)
>
> > arrlist = [np.random.normal(size=100) for i in range(50)]
>
> > ret = ax1.hist(arrlist,histtype='barstacked')
>
> > reclist = [patchlist[0] for patchlist in ret[2]]
>
> > labellist = ['data'+str(i) for i in range(50)]
>
> > ax2.legend(reclist,labellist,loc='upper
> left',bbox_to_anchor=(0,0,1,1),borderaxespad=0.,ncol=5,mode='expand')
> > ax2.set_frame_on(False)
>
> > ax2.tick_params(bottom='off',left='off',right='off',top='off')
>
> > plt.setp(ax2.get_yticklabels(),visible=False)
>
> > plt.setp(ax2.get_xticklabels(),visible=False)
> >
> I added plt.show() and it demonstrates my problem: the legend is not
> complete in the
> figure. That is why I think I could instead use:
>
> import pylab as plt
> import numpy as np
>
> fig = plt.figure()
> DefaultSize = tuple(fig.get_size_inches())
> fig.set_size_inches(DefaultSize[0], 4*DefaultSize[1])
> ax1 = fig.add_subplot(411)
> ax2 = fig.add_subplot(412)
> arrlist = [np.random.normal(size=100) for i in range(50)]
> ret = ax1.hist(arrlist,histtype='barstacked')
> reclist = [patchlist[0] for patchlist in ret[2]]
> labellist = ['data'+str(i) for i in range(50)]
> ax2.legend(reclist,labellist,loc='upper
> left',bbox_to_anchor=(0,0,1,1),borderaxespad=0.,ncol=5,mode='expand')
> ax2.set_frame_on(False)
> ax2.tick_params(bottom='off',left='off',right='off',top='off')
> plt.setp(ax2.get_yticklabels(),visible=False)
> plt.setp(ax2.get_xticklabels(),visible=False)
> plt.show()
>
> But, this does not make the image 4* taller than I thought. But thank you
> for the example how to extract the legend of ax1 and place it under ax2.
>
>
>
> >
> > you're asking some object-oriented way, I personally don't think
> > using pylab and set_tight_layout are the good way to be
> > "object-oriented" as pylab is only a bounding wrapper by my
> > understanding (maybe I am wrong!). legend and hist are all
> > matplotlib.axes.Axes method.
> >
> > Also, I think it's unrealistic to ask the figure do a nice job for
> > you if there are 50 legend handlers and you want to show them in 2
> > columns with a very high width/height ratio of the figure....
>
> The problem is that the data are calculated dynamically and sometimes
> I need to display data for 20 data types while sometimes for 200 data
> types (and for each I need a legend).
>
> I did not show that but I do calculate how many columns I could use
> legend display and pass that via pylab.legend(..., ncol= ). Of course
> at the same time I could calculate whether I will need 2 or 3 or 4
> subplots on the page (the first will be the barchart itself), the
> remaining space will be used by the long legend of subplot(211).
> I would hope that matplotlib does not mind that I actually issue any
> fig.add_subplot() foe the third or even fourth subplot at all. That
> would be just a trick to get more space for the legend. If I can live
> with just with subplot(211) and subplot(212)
>
> The fig.savefig('foobar.png', bbox_inches='tight') which Ben mentioned
> yesterday is nice but I want it to crop the image only vertically.
> An optional argument like:
> fig.savefig('foobar.png', bbox_inches='tight', keep_fig_width=True)
> would maybe do the job for me.
>
>
> What I still don't understand what is resizing the image in tight_layout.
> It doesn't seem to me that just the unused border space is chopped away.
> Fonts look different, ratio between x and y axes lengths seems different.
> Certainly not what I want.
>
>
> > hope it could be of a bit help,
>
> Sure, I am still learning to use matplotlib.
>
> Martin
>
> >
> > cheers,
> >
> > Chao
> >
>
> >
> > On Mon, May 20, 2013 at 6:43 PM, Martin Mokrejs [via matplotlib]
> <[hidden email] </user/SendEmail.jtp?type=node&node=41102&i=0>> wrote:
> >
> > Hi Ben,
> >
> > Benjamin Root wrote:
> >
> > >
> > >
> > >
> > > On Mon, May 20, 2013 at 12:02 PM, Martin Mokrejs <[hidden email]
> <http://user/SendEmail.jtp?type=node&node=41090&i=0> <mailto:[hidden email]
> <http://user/SendEmail.jtp?type=node&node=41090&i=1>>> wrote:
>
> > >
> > > Hi,
> > > I am having trouble to get space allocated for a long
> legend text,
> > > lets say spanning 2/3 - 3/4 of the whole output. I would like
> to have
> > > stacked barchart as 1st subplot and the place of remaining 3
> subplots
> > > to be actually allocated by the legend. Alternatively, could
> I get the
> > > legend saved into a separate figure?
> > >
> > > Or could the space for legend text be allocated automatically
> minimizing
> > > output figure size? For example, the width would be 1120px
> while height
> > > be multiples of 840px (840 for each subplot)?
> > >
> > > Attached is a quick example. It shows also that I tried
> tight_layout()
> > > but wasn't successful with this either. I would be glad for
> some help,
> > > ideally converting the whole thing into an object-oriented
> approach.
> > > I am generating several figures in a row and would like to
> clear()/del()
> > > any previously used data ASAP.
> > >
> > >
> > > Thank you,
> > > Martin
> > > Am using mpl-1.2.2
> > >
> > >
> > > Try "fig.savefig('foobar.png', bbox_inches='tight')" when saving
> the
> > > image. It will make the figure size such that all the visible
> > > elements of the figure will fit into the saved output.
> tight_layout()
> > > is meant to make sure the elements don't overlap each other, but
> does
> > > nothing about making sure nothing gets clipped.
> > Ah, would be nice to make this clear in the docs. So far was doing
> >
> >
> > import pylab
> > F = pylab.gcf()
> > F.set_tight_layout(True)
> >
> > which as you say does not help the way I thought.
> >
> >
> > Unfortunately, while
> >
> > fig.savefig('foobar.png', bbox_inches='tight')
> >
> > helped to get everything into the .png file (attached), the
> barchart itself
> > should span according to the code I posted just 1/2 of the figure.
> But somehow
> > it is enlarged and rescaled so that it occupies *more than* 1/2 of
> the figure.
> > What in pylab is resizing my image? Note: the final image is
> 625x1075.
> >
> > Martin
------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may
_______________________________________________
Matplotlib-users mailing list
Matplotlib-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-users
