From: Angus McMorland <amcmorl@gm...>  20120318 12:07:29

From: klo uo <klonuo@gm...>  20120318 08:40:05

From: Angus McMorland <amcmorl@gm...>  20120318 12:07:29

From: klo uo <klonuo@gm...>  20120318 12:43:10
Attachments:
Message as HTML

On Sun, Mar 18, 2012 at 1:07 PM, Angus McMorland <amcmorl@...> wrote: > > > The xlim command can be used to set the x limits. For example: > > xlim(.5, 2.5) > > will prevent the points lying on the axis boundaries for your case. > > Thanks Angus, that worked with ease for separate MPL window, but not inline in IPython. I guess there is no setting, that will allow setting MPL to automatically adjust default plot window  add padding if bars (which can also be lines or points as in example) are drawn on axes; trim window if if there is no data to plot instead trimming based on grid range; and similar intuitive expectation 
From: Angus McMorland <amcmorl@gm...>  20120318 12:51:17

On 18 March 2012 08:43, klo uo <klonuo@...> wrote: > On Sun, Mar 18, 2012 at 1:07 PM, Angus McMorland <amcmorl@...> wrote: >> >> >> >> The xlim command can be used to set the x limits. For example: >> >> xlim(.5, 2.5) >> >> will prevent the points lying on the axis boundaries for your case. >> > > Thanks Angus, > > that worked with ease for separate MPL window, but not inline in IPython. That's because the first command draws the plot, and when inlined, further changes within the cell aren't propagated to the plot. > I guess there is no setting, that will allow setting MPL to automatically > adjust default plot window  add padding if bars (which can also be lines or > points as in example) are drawn on axes; trim window if if there is no data > to plot instead trimming based on grid range; and similar intuitive > expectation For inline ipython, you want to switch to the objectoriented use of pylab. Something like this should work with xlim. a = [0.1, 0.2, 0.1] fig = plt.figure() ax = fig.add_subplot(111) ax.errorbar(arange(3), a, yerr=asum(a)/len(a), fmt='ro') ax.set_xlim(.5,2.5) ax.show() I'm not aware of automatic settings for padding, but with this set_xlim, it's easy enough to roll your own using the data limits. Angus  AJC McMorland Postdoctoral research fellow Neurobiology, University of Pittsburgh 
From: klo uo <klonuo@gm...>  20120318 13:14:31
Attachments:
Message as HTML

On Sun, Mar 18, 2012 at 1:50 PM, Angus McMorland <amcmorl@...> wrote: > > For inline ipython, you want to switch to the objectoriented use of > pylab. Something like this should work with xlim. > > a = [0.1, 0.2, 0.1] > fig = plt.figure() > ax = fig.add_subplot(111) > ax.errorbar(arange(3), a, yerr=asum(a)/len(a), fmt='ro') > ax.set_xlim(.5,2.5) > ax.show() > > I'm not aware of automatic settings for padding, but with this > set_xlim, it's easy enough to roll your own using the data limits. > > OK, thanks It's not very elegant (assuming pylab freedom) but I take it as only way to correct clipping example (or differently put  to use custom range for axis) 
From: Tony Yu <tsyu80@gm...>  20120318 13:36:26
Attachments:
Message as HTML

On Sun, Mar 18, 2012 at 9:14 AM, klo uo <klonuo@...> wrote: > On Sun, Mar 18, 2012 at 1:50 PM, Angus McMorland <amcmorl@...>wrote: > >> >> For inline ipython, you want to switch to the objectoriented use of >> pylab. Something like this should work with xlim. >> >> a = [0.1, 0.2, 0.1] >> fig = plt.figure() >> ax = fig.add_subplot(111) >> ax.errorbar(arange(3), a, yerr=asum(a)/len(a), fmt='ro') >> ax.set_xlim(.5,2.5) >> ax.show() >> >> I'm not aware of automatic settings for padding, but with this >> set_xlim, it's easy enough to roll your own using the data limits. >> >> > OK, thanks > > It's not very elegant (assuming pylab freedom) but I take it as only way > to correct clipping example (or differently put  to use custom range for > axis) > > You can roll out a utility function that can automatically adjust the limits with some specified padding (i.e. the function doesn't calculate the marker size, but you can just give it sufficient padding). Here's an example where you specify a padding passed on the axes size (0.05 is 5% of axes height/width): #~~~~ import numpy as np import matplotlib.pyplot as plt def pad_limits(pad_frac=0.05, ax=None): ax = ax if ax is not None else plt.gca() ax.set_xlim(_calc_limits(ax.xaxis, pad_frac)) ax.set_ylim(_calc_limits(ax.yaxis, pad_frac)) def _calc_limits(axis, frac): limits = axis.get_data_interval() mag = np.diff(limits)[0] pad = np.array([mag*frac, mag*frac]) return limits + pad a = np.array([0.1, 0.2, 0.1]) plt.errorbar(np.arange(3), a, yerr=asum(a)/len(a), fmt='ro') pad_limits() plt.show() #~~~~ I've put this is my personal mpl toolkit with the added ability of handling log scales: https://github.com/tonysyu/mpltools/blob/master/mpltools/layout.py#L80 Best, Tony 
From: Benjamin Root <ben.root@ou...>  20120318 15:08:29
Attachments:
Message as HTML

On Sunday, March 18, 2012, Tony Yu <tsyu80@...> wrote: > > > On Sun, Mar 18, 2012 at 9:14 AM, klo uo <klonuo@...> wrote: >> >> On Sun, Mar 18, 2012 at 1:50 PM, Angus McMorland <amcmorl@...> wrote: >>> >>> For inline ipython, you want to switch to the objectoriented use of >>> pylab. Something like this should work with xlim. >>> >>> a = [0.1, 0.2, 0.1] >>> fig = plt.figure() >>> ax = fig.add_subplot(111) >>> ax.errorbar(arange(3), a, yerr=asum(a)/len(a), fmt='ro') >>> ax.set_xlim(.5,2.5) >>> ax.show() >>> >>> I'm not aware of automatic settings for padding, but with this >>> set_xlim, it's easy enough to roll your own using the data limits. >>> >> >> OK, thanks >> >> It's not very elegant (assuming pylab freedom) but I take it as only way to correct clipping example (or differently put  to use custom range for axis) >> > > You can roll out a utility function that can automatically adjust the limits with some specified padding (i.e. the function doesn't calculate the marker size, but you can just give it sufficient padding). > Here's an example where you specify a padding passed on the axes size (0.05 is 5% of axes height/width): > #~~~~ > import numpy as np > import matplotlib.pyplot as plt > def pad_limits(pad_frac=0.05, ax=None): > ax = ax if ax is not None else plt.gca() > ax.set_xlim(_calc_limits(ax.xaxis, pad_frac)) > ax.set_ylim(_calc_limits(ax.yaxis, pad_frac)) > def _calc_limits(axis, frac): > limits = axis.get_data_interval() > mag = np.diff(limits)[0] > pad = np.array([mag*frac, mag*frac]) > return limits + pad > a = np.array([0.1, 0.2, 0.1]) > plt.errorbar(np.arange(3), a, yerr=asum(a)/len(a), fmt='ro') > pad_limits() > plt.show() > #~~~~ > I've put this is my personal mpl toolkit with the added ability of handling log scales: > https://github.com/tonysyu/mpltools/blob/master/mpltools/layout.py#L80 > Best, > Tony > Uhm, don't we already have padx and pady kwargs for various limits functions? I know scatter and plot respects them. Ben Root 
From: Tony Yu <tsyu80@gm...>  20120318 17:04:21
Attachments:
Message as HTML

On Sun, Mar 18, 2012 at 11:08 AM, Benjamin Root <ben.root@...> wrote: > > > On Sunday, March 18, 2012, Tony Yu <tsyu80@...> wrote: > > > > > > On Sun, Mar 18, 2012 at 9:14 AM, klo uo <klonuo@...> wrote: > >> > >> On Sun, Mar 18, 2012 at 1:50 PM, Angus McMorland <amcmorl@...> > wrote: > >>> > >>> For inline ipython, you want to switch to the objectoriented use of > >>> pylab. Something like this should work with xlim. > >>> > >>> a = [0.1, 0.2, 0.1] > >>> fig = plt.figure() > >>> ax = fig.add_subplot(111) > >>> ax.errorbar(arange(3), a, yerr=asum(a)/len(a), fmt='ro') > >>> ax.set_xlim(.5,2.5) > >>> ax.show() > >>> > >>> I'm not aware of automatic settings for padding, but with this > >>> set_xlim, it's easy enough to roll your own using the data limits. > >>> > >> > >> OK, thanks > >> > >> It's not very elegant (assuming pylab freedom) but I take it as only > way to correct clipping example (or differently put  to use custom range > for axis) > >> > > > > You can roll out a utility function that can automatically adjust the > limits with some specified padding (i.e. the function doesn't calculate the > marker size, but you can just give it sufficient padding). > > Here's an example where you specify a padding passed on the axes size > (0.05 is 5% of axes height/width): > > #~~~~ > > import numpy as np > > import matplotlib.pyplot as plt > > def pad_limits(pad_frac=0.05, ax=None): > > ax = ax if ax is not None else plt.gca() > > ax.set_xlim(_calc_limits(ax.xaxis, pad_frac)) > > ax.set_ylim(_calc_limits(ax.yaxis, pad_frac)) > > def _calc_limits(axis, frac): > > limits = axis.get_data_interval() > > mag = np.diff(limits)[0] > > pad = np.array([mag*frac, mag*frac]) > > return limits + pad > > a = np.array([0.1, 0.2, 0.1]) > > plt.errorbar(np.arange(3), a, yerr=asum(a)/len(a), fmt='ro') > > pad_limits() > > plt.show() > > #~~~~ > > I've put this is my personal mpl toolkit with the added ability of > handling log scales: > > https://github.com/tonysyu/mpltools/blob/master/mpltools/layout.py#L80 > > Best, > > Tony > > > > > Uhm, don't we already have padx and pady kwargs for various limits > functions? I know scatter and plot respects them. > > Ben Root Oh, I didn't know anything about them. ... and where exactly? I can't seem to find them (I looked in `ax.autoscale`, `ax.autoscale_view`, and `plt.xlim`). Tony (Sorry for the duplicate, Ben. I forgot to reply all) 
From: klo uo <klonuo@gm...>  20120318 21:00:40
Attachments:
Message as HTML

After parsing matplotlibrc, I browsed module where errorbars are defined (axes.py) and tried changing various variables without success. In bar() function (line 4628) there is "adjust_xlim = False" line which calls line 4768 if set True. So I set it True, to find it's buggy if x starts from 0 (most common start value). I didn't tried to copy this code block in errorbars because of that I browsed then axis.py and then  transforms.py  total mess. Retreat. Didn't even figured out why IPython inline mode pads left side x range in above example. Seems like IPython/core/pylabtools.py just calls "canvas.print_figure(picdata)" and passes image in qt terminal, but I can't reproduce same range if not in inline mode. Idea was to learn how IPython inline mode sets one part of this range correctly, then use it to make what I wanted So, I guess wrapping some function that would calculate smart view range, like Tony replied, is the way to go Thanks Tony 