Menu
▾
▴
matplotlib-devel
|
From: Tomo L. <laz...@gm...> - 2015-03-08 00:20:19
|
Hello matplotlib developers, I'm not sure if this is the right mailing list for this question, so please re-direct me if it is not. I am wondering whether it is possible to have a histogram in pyplot normalized to the total length of the list input, rather than just the bins showing on the plot (i.e. include those entries in the "overflow" and "underflow", off the right and left edges of the plot). As far as I can tell, the normed option of pyplot.hist currently makes it so that the area under the bins showing is 1. This can lead to a situation like the one pasted below, where when I look at the whole histogram the bins have certain values but when I try to zoom in to see one part of the plot better those values change. I can think of two ways to solve this as of now: 1) Use the weights option to scale each entry by 1/len(input) rather than using normed=True. 2) Somehow add the contents of the overflow to the last bin of the plot, which would keep the normalizations constant for earlier bins even if you extend the axes. Is there a better way of doing this? If the option does not currently exist, I am also happy to help implement it if the community would find it desirable. Thanks for your help! Tomo Lazovich P.S. Here is a toy example of what I mean: >> import numpy as np >> import matplotlib.pyplot as plt >> h1 = [0, 0, 0, 1, 1, 2, 3] >> my_bins = np.linspace(-0.5, 4.5, 6) >> plt.hist(h1, bins=my_bins, normed=True) >> plt.show() gives [image: Inline image 1] Now, if I change the range on the x axis that I would like plot: >> my_bins2 = np.linspace(-0.5, 1.5, 3) >> plt.hist(h1, bins=my_bins2, normed=True) >> plt.show() [image: Inline image 2] The y values have changed to 0.6 and 0.4 because the normalization does not include the values that are cut off to the right of the plot. |
|
From: Paul H. <pmh...@gm...> - 2015-03-08 00:28:49
|
IMO, this seems like a bug. I would expect bars to change height with zoom/limit levels. -p — Sent from Mailbox On Sat, Mar 7, 2015 at 4:20 PM, Tomo Lazovich <laz...@gm...> wrote: > Hello matplotlib developers, > I'm not sure if this is the right mailing list for this question, so please > re-direct me if it is not. > I am wondering whether it is possible to have a histogram in pyplot > normalized to the total length of the list input, rather than just the bins > showing on the plot (i.e. include those entries in the "overflow" and > "underflow", off the right and left edges of the plot). As far as I can > tell, the normed option of pyplot.hist currently makes it so that the area > under the bins showing is 1. This can lead to a situation like the one > pasted below, where when I look at the whole histogram the bins have > certain values but when I try to zoom in to see one part of the plot better > those values change. > I can think of two ways to solve this as of now: > 1) Use the weights option to scale each entry by 1/len(input) rather than > using normed=True. > 2) Somehow add the contents of the overflow to the last bin of the plot, > which would keep the normalizations constant for earlier bins even if you > extend the axes. > Is there a better way of doing this? If the option does not currently > exist, I am also happy to help implement it if the community would find it > desirable. > Thanks for your help! > Tomo Lazovich > P.S. Here is a toy example of what I mean: >>> import numpy as np >>> import matplotlib.pyplot as plt >>> h1 = [0, 0, 0, 1, 1, 2, 3] >>> my_bins = np.linspace(-0.5, 4.5, 6) >>> plt.hist(h1, bins=my_bins, normed=True) >>> plt.show() > gives > [image: Inline image 1] > Now, if I change the range on the x axis that I would like plot: >>> my_bins2 = np.linspace(-0.5, 1.5, 3) >>> plt.hist(h1, bins=my_bins2, normed=True) >>> plt.show() > [image: Inline image 2] > The y values have changed to 0.6 and 0.4 because the normalization does not > include the values that are cut off to the right of the plot. |
|
From: Thomas C. <tca...@gm...> - 2015-03-08 00:40:03
|
Paul, Note that by zoom the op means they are changing the bins, not actual zooming(by just changing the x axis). I was going to say we deal with normalization by delegating to numpy, but we actually handle it internally (with a note that when we drop np 1.5 to make numpy do it). I think the best course of action here is to do that conversion and forward this feature request to numpy (if it does not already do this). Tom On Sat, Mar 7, 2015, 18:29 Paul Hobson <pmh...@gm...> wrote: > IMO, this seems like a bug. I would expect bars to change height with > zoom/limit levels. > -p > > — > Sent from Mailbox <https://www.dropbox.com/mailbox> > > > On Sat, Mar 7, 2015 at 4:20 PM, Tomo Lazovich <laz...@gm...> wrote: > >> Hello matplotlib developers, >> >> I'm not sure if this is the right mailing list for this question, so >> please re-direct me if it is not. >> >> I am wondering whether it is possible to have a histogram in pyplot >> normalized to the total length of the list input, rather than just the bins >> showing on the plot (i.e. include those entries in the "overflow" and >> "underflow", off the right and left edges of the plot). As far as I can >> tell, the normed option of pyplot.hist currently makes it so that the area >> under the bins showing is 1. This can lead to a situation like the one >> pasted below, where when I look at the whole histogram the bins have >> certain values but when I try to zoom in to see one part of the plot better >> those values change. >> >> I can think of two ways to solve this as of now: >> >> 1) Use the weights option to scale each entry by 1/len(input) rather than >> using normed=True. >> 2) Somehow add the contents of the overflow to the last bin of the plot, >> which would keep the normalizations constant for earlier bins even if you >> extend the axes. >> >> Is there a better way of doing this? If the option does not currently >> exist, I am also happy to help implement it if the community would find it >> desirable. >> >> Thanks for your help! >> >> Tomo Lazovich >> >> P.S. Here is a toy example of what I mean: >> >> >> import numpy as np >> >> import matplotlib.pyplot as plt >> >> h1 = [0, 0, 0, 1, 1, 2, 3] >> >> my_bins = np.linspace(-0.5, 4.5, 6) >> >> plt.hist(h1, bins=my_bins, normed=True) >> >> plt.show() >> >> gives >> >> <image.png> >> >> >> Now, if I change the range on the x axis that I would like plot: >> >> >> my_bins2 = np.linspace(-0.5, 1.5, 3) >> >> plt.hist(h1, bins=my_bins2, normed=True) >> >> plt.show() >> >> <image.png> >> >> >> The y values have changed to 0.6 and 0.4 because the normalization does >> not include the values that are cut off to the right of the plot. >> > > ------------------------------------------------------------ > ------------------ > Dive into the World of Parallel Programming The Go Parallel Website, > sponsored > by Intel and developed in partnership with Slashdot Media, is your hub for > all > things parallel software development, from weekly thought leadership blogs > to > news, videos, case studies, tutorials and more. Take a look and join the > conversation now. http://goparallel.sourceforge.net/ > _______________________________________________ > Matplotlib-devel mailing list > Mat...@li... > https://lists.sourceforge.net/lists/listinfo/matplotlib-devel > |
|
From: Tomo L. <laz...@gm...> - 2015-03-08 01:00:36
|
Thanks for the suggestion...I will see how numpy handles this. Sorry for not being clearer earlier. Tom is right that by "zooming" I meant changing the bins so that they covered a smaller range. Is there a better way of "zooming" in on an axis so that I don't have this issue? Thanks! Tomo On Sat, Mar 7, 2015 at 7:39 PM, Thomas Caswell <tca...@gm...> wrote: > Paul, > > Note that by zoom the op means they are changing the bins, not actual > zooming(by just changing the x axis). > > I was going to say we deal with normalization by delegating to numpy, but > we actually handle it internally (with a note that when we drop np 1.5 to > make numpy do it). > I think the best course of action here is to do that conversion and > forward this feature request to numpy (if it does not already do this). > > Tom > > On Sat, Mar 7, 2015, 18:29 Paul Hobson <pmh...@gm...> wrote: > >> IMO, this seems like a bug. I would expect bars to change height with >> zoom/limit levels. >> -p >> >> — >> Sent from Mailbox <https://www.dropbox.com/mailbox> >> >> >> On Sat, Mar 7, 2015 at 4:20 PM, Tomo Lazovich <laz...@gm...> wrote: >> >>> Hello matplotlib developers, >>> >>> I'm not sure if this is the right mailing list for this question, so >>> please re-direct me if it is not. >>> >>> I am wondering whether it is possible to have a histogram in pyplot >>> normalized to the total length of the list input, rather than just the bins >>> showing on the plot (i.e. include those entries in the "overflow" and >>> "underflow", off the right and left edges of the plot). As far as I can >>> tell, the normed option of pyplot.hist currently makes it so that the area >>> under the bins showing is 1. This can lead to a situation like the one >>> pasted below, where when I look at the whole histogram the bins have >>> certain values but when I try to zoom in to see one part of the plot better >>> those values change. >>> >>> I can think of two ways to solve this as of now: >>> >>> 1) Use the weights option to scale each entry by 1/len(input) rather >>> than using normed=True. >>> 2) Somehow add the contents of the overflow to the last bin of the plot, >>> which would keep the normalizations constant for earlier bins even if you >>> extend the axes. >>> >>> Is there a better way of doing this? If the option does not currently >>> exist, I am also happy to help implement it if the community would find it >>> desirable. >>> >>> Thanks for your help! >>> >>> Tomo Lazovich >>> >>> P.S. Here is a toy example of what I mean: >>> >>> >> import numpy as np >>> >> import matplotlib.pyplot as plt >>> >> h1 = [0, 0, 0, 1, 1, 2, 3] >>> >> my_bins = np.linspace(-0.5, 4.5, 6) >>> >> plt.hist(h1, bins=my_bins, normed=True) >>> >> plt.show() >>> >>> gives >>> >>> <image.png> >>> >>> >>> Now, if I change the range on the x axis that I would like plot: >>> >>> >> my_bins2 = np.linspace(-0.5, 1.5, 3) >>> >> plt.hist(h1, bins=my_bins2, normed=True) >>> >> plt.show() >>> >>> <image.png> >>> >>> >>> The y values have changed to 0.6 and 0.4 because the normalization does >>> not include the values that are cut off to the right of the plot. >>> >> >> ------------------------------------------------------------ >> ------------------ >> Dive into the World of Parallel Programming The Go Parallel Website, >> sponsored >> by Intel and developed in partnership with Slashdot Media, is your hub >> for all >> things parallel software development, from weekly thought leadership >> blogs to >> news, videos, case studies, tutorials and more. Take a look and join the >> conversation now. http://goparallel.sourceforge.net/ >> _______________________________________________ >> Matplotlib-devel mailing list >> Mat...@li... >> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel >> > |
|
From: Tomo L. <laz...@gm...> - 2015-03-08 01:47:34
|
Sorry for the spam, but I just wanted to say that I now understand that I should be using plt.xlim to zoom in on the x-axis rather than changing the bins. When I zoom in with that, the bin height is indeed constant as expected. On Sat, Mar 7, 2015 at 8:00 PM, Tomo Lazovich <laz...@gm...> wrote: > Thanks for the suggestion...I will see how numpy handles this. > > Sorry for not being clearer earlier. Tom is right that by "zooming" I > meant changing the bins so that they covered a smaller range. Is there a > better way of "zooming" in on an axis so that I don't have this issue? > > Thanks! > > Tomo > > On Sat, Mar 7, 2015 at 7:39 PM, Thomas Caswell <tca...@gm...> wrote: > >> Paul, >> >> Note that by zoom the op means they are changing the bins, not actual >> zooming(by just changing the x axis). >> >> I was going to say we deal with normalization by delegating to numpy, but >> we actually handle it internally (with a note that when we drop np 1.5 to >> make numpy do it). >> I think the best course of action here is to do that conversion and >> forward this feature request to numpy (if it does not already do this). >> >> Tom >> >> On Sat, Mar 7, 2015, 18:29 Paul Hobson <pmh...@gm...> wrote: >> >>> IMO, this seems like a bug. I would expect bars to change height with >>> zoom/limit levels. >>> -p >>> >>> — >>> Sent from Mailbox <https://www.dropbox.com/mailbox> >>> >>> >>> On Sat, Mar 7, 2015 at 4:20 PM, Tomo Lazovich <laz...@gm...> >>> wrote: >>> >>>> Hello matplotlib developers, >>>> >>>> I'm not sure if this is the right mailing list for this question, so >>>> please re-direct me if it is not. >>>> >>>> I am wondering whether it is possible to have a histogram in pyplot >>>> normalized to the total length of the list input, rather than just the bins >>>> showing on the plot (i.e. include those entries in the "overflow" and >>>> "underflow", off the right and left edges of the plot). As far as I can >>>> tell, the normed option of pyplot.hist currently makes it so that the area >>>> under the bins showing is 1. This can lead to a situation like the one >>>> pasted below, where when I look at the whole histogram the bins have >>>> certain values but when I try to zoom in to see one part of the plot better >>>> those values change. >>>> >>>> I can think of two ways to solve this as of now: >>>> >>>> 1) Use the weights option to scale each entry by 1/len(input) rather >>>> than using normed=True. >>>> 2) Somehow add the contents of the overflow to the last bin of the >>>> plot, which would keep the normalizations constant for earlier bins even if >>>> you extend the axes. >>>> >>>> Is there a better way of doing this? If the option does not currently >>>> exist, I am also happy to help implement it if the community would find it >>>> desirable. >>>> >>>> Thanks for your help! >>>> >>>> Tomo Lazovich >>>> >>>> P.S. Here is a toy example of what I mean: >>>> >>>> >> import numpy as np >>>> >> import matplotlib.pyplot as plt >>>> >> h1 = [0, 0, 0, 1, 1, 2, 3] >>>> >> my_bins = np.linspace(-0.5, 4.5, 6) >>>> >> plt.hist(h1, bins=my_bins, normed=True) >>>> >> plt.show() >>>> >>>> gives >>>> >>>> <image.png> >>>> >>>> >>>> Now, if I change the range on the x axis that I would like plot: >>>> >>>> >> my_bins2 = np.linspace(-0.5, 1.5, 3) >>>> >> plt.hist(h1, bins=my_bins2, normed=True) >>>> >> plt.show() >>>> >>>> <image.png> >>>> >>>> >>>> The y values have changed to 0.6 and 0.4 because the normalization does >>>> not include the values that are cut off to the right of the plot. >>>> >>> >>> ------------------------------------------------------------ >>> ------------------ >>> Dive into the World of Parallel Programming The Go Parallel Website, >>> sponsored >>> by Intel and developed in partnership with Slashdot Media, is your hub >>> for all >>> things parallel software development, from weekly thought leadership >>> blogs to >>> news, videos, case studies, tutorials and more. Take a look and join the >>> conversation now. http://goparallel.sourceforge.net/ >>> _______________________________________________ >>> Matplotlib-devel mailing list >>> Mat...@li... >>> https://lists.sourceforge.net/lists/listinfo/matplotlib-devel >>> >> > |
|
From: OceanWolf <jui...@ya...> - 2015-03-10 11:18:54
|
Tom, ``When we drop numpy 1.5''? I thought we already had... I mean we only test numpy 1.6 on Travis... For the rebinning exercise, I don't have time to look, but I would expect a similar trick to imshow, quiver, etcetera when I want to compare to a baseline (e.g. for animation). Namely I calculate the normalisation parameters first, and then apply those parameters on subsequent plots. To ease the user, we could add a method to return the binning parameters from a single binning exercise, and then give an option to pass those params in to subsequent plots. |
|
From: Thomas C. <tca...@gm...> - 2015-03-10 11:47:58
|
Right on no longer supporting 1.5, but this code never got updated. This is a bit of a bigger job than I first anticipated as numpy has deprecated the norm kwarg, so we probably should too. On Tue, Mar 10, 2015, 07:19 OceanWolf <jui...@ya...> wrote: > Tom, ``When we drop numpy 1.5''? I thought we already had... I mean we > only test numpy 1.6 on Travis... > > For the rebinning exercise, I don't have time to look, but I would > expect a similar trick to imshow, quiver, etcetera when I want to > compare to a baseline (e.g. for animation). Namely I calculate the > normalisation parameters first, and then apply those parameters on > subsequent plots. > > To ease the user, we could add a method to return the binning parameters > from a single binning exercise, and then give an option to pass those > params in to subsequent plots. > > ------------------------------------------------------------ > ------------------ > Dive into the World of Parallel Programming The Go Parallel Website, > sponsored > by Intel and developed in partnership with Slashdot Media, is your hub for > all > things parallel software development, from weekly thought leadership blogs > to > news, videos, case studies, tutorials and more. Take a look and join the > conversation now. http://goparallel.sourceforge.net/ > _______________________________________________ > Matplotlib-devel mailing list > Mat...@li... > https://lists.sourceforge.net/lists/listinfo/matplotlib-devel > |
|
From: OceanWolf <jui...@ya...> - 2015-03-10 15:51:03
|
Slightly off-topic, does github allow for one to tag lines/code-blocks with notes to the future like ``numpy1.5 removal'', similar to tags for issues and milestones? This would save us from trawling through the code looking for comments. If github does not have this feature, do you think github would add that as a feature if asked? |
|
From: Paul H. <pmh...@gm...> - 2015-03-10 16:12:30
|
You can comment on specific lines of code in the pull request interface, but that's not what I think you're describing. A better practice, IMO is to raise a DeprecationWarning when the soon-to-be-removed code is executed. Then you can just grep for those and get cracking. -p On Tue, Mar 10, 2015 at 8:51 AM OceanWolf <jui...@ya...> wrote: > Slightly off-topic, does github allow for one to tag lines/code-blocks > with notes to the future like ``numpy1.5 removal'', similar to tags for > issues and milestones? This would save us from trawling through the > code looking for comments. If github does not have this feature, do you > think github would add that as a feature if asked? > > ------------------------------------------------------------ > ------------------ > Dive into the World of Parallel Programming The Go Parallel Website, > sponsored > by Intel and developed in partnership with Slashdot Media, is your hub for > all > things parallel software development, from weekly thought leadership blogs > to > news, videos, case studies, tutorials and more. Take a look and join the > conversation now. http://goparallel.sourceforge.net/ > _______________________________________________ > Matplotlib-devel mailing list > Mat...@li... > https://lists.sourceforge.net/lists/listinfo/matplotlib-devel > |
|
From: OceanWolf <jui...@ya...> - 2015-03-10 16:34:09
|
Yes, you have almost got what I want. I know of both line comments in the pull interface, and raising DeprecationWarnings. The problem with the warnings that they only want to get used when you have an alternate in place and want to remove old code soon. The problem here lies in not knowing how long the life-cycle will last for an indeterminate amount of time, such as numpy-1.5 here, or python-2.7. Lots of such items can stack up, and easily get forgotten about, as we see here. On 10/03/15 17:12, Paul Hobson wrote: > You can comment on specific lines of code in the pull request > interface, but that's not what I think you're describing. A better > practice, IMO is to raise a DeprecationWarning when the > soon-to-be-removed code is executed. Then you can just grep for those > and get cracking. > -p > > On Tue, Mar 10, 2015 at 8:51 AM OceanWolf > <jui...@ya... <mailto:jui...@ya...>> > wrote: > > Slightly off-topic, does github allow for one to tag lines/code-blocks > with notes to the future like ``numpy1.5 removal'', similar to > tags for > issues and milestones? This would save us from trawling through the > code looking for comments. If github does not have this feature, > do you > think github would add that as a feature if asked? > > ------------------------------------------------------------------------------ > Dive into the World of Parallel Programming The Go Parallel > Website, sponsored > by Intel and developed in partnership with Slashdot Media, is your > hub for all > things parallel software development, from weekly thought > leadership blogs to > news, videos, case studies, tutorials and more. Take a look and > join the > conversation now. http://goparallel.sourceforge.net/ > _______________________________________________ > Matplotlib-devel mailing list > Mat...@li... > <mailto:Mat...@li...> > https://lists.sourceforge.net/lists/listinfo/matplotlib-devel > |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X