Thread: [matplotlib-devel] Bug in pyplot.hist when using histtype="step"

[matplotlib-devel] Bug in pyplot.hist when using histtype="step"

From: Detlef M. (IKP) <det...@ki...> - 2013-04-16 10:14:17

Attachments: 1.png 2.png 0001-fix-y-axis-range-for-histogram-in-step-mode.patch

Hi all,

there seems to be a bug in in pyplot.hist when using histtype="step".

I am plotting the attached data to a histogram (see 1.png). In this case 
I set the limits on the y axis manually.

When I don't do this (let the hist function choose the limits), I get 
the picture in the file 2.png. Only the highest bin is drawn. To be 
exact, ymin=second highest value, ymax=highest value.

To solve this, I suggest the patch in the attachment.

I also removed the for loop here, because I don't see why 0 height bins 
should be filtered out (but this is just a suggestion).


cheers,

Detlef Maurel

Re: [matplotlib-devel] Bug in pyplot.hist when using histtype="step"

From: Pierre H. <pie...@cr...> - 2013-04-16 13:16:54

Hi,

Le 16/04/2013 12:14, Detlef Maurel (IKP) a écrit :
> there seems to be a bug in in pyplot.hist when using histtype="step".
>
> I am plotting the attached data to a histogram (see 1.png). In this
> case I set the limits on the y axis manually.
>
> When I don't do this (let the hist function choose the limits), I get
> the picture in the file 2.png. Only the highest bin is drawn. To be
> exact, ymin=second highest value, ymax=highest value. 
Can you also send a minimal code that reproduces the bug. I just tried a
hist(x, histtype='step') command, with x being a random vector and it
works fine.

best,
Pierre

Re: [matplotlib-devel] Bug in pyplot.hist when using histtype="step"

From: Detlef M. (IKP) <det...@ki...> - 2013-04-16 13:42:10

Attachments: data.txt

On 16.04.2013 14:58, Pierre Haessig wrote:
> Hi,
>
> Le 16/04/2013 12:14, Detlef Maurel (IKP) a écrit :
>> there seems to be a bug in in pyplot.hist when using histtype="step".
>>
>> I am plotting the attached data to a histogram (see 1.png). In this
>> case I set the limits on the y axis manually.
>>
>> When I don't do this (let the hist function choose the limits), I get
>> the picture in the file 2.png. Only the highest bin is drawn. To be
>> exact, ymin=second highest value, ymax=highest value.
> Can you also send a minimal code that reproduces the bug. I just tried a
> hist(x, histtype='step') command, with x being a random vector and it
> works fine.
>

this works for me, too. Strange... I get the problem with the attached 
dataset using the command

hist(loadtxt("data.txt"),bins=300,histtype="step")

When I plot only a subset of the data (first/second/third/... thousand 
numbers),
the histogram looks fine. It seems it gets wrong when more than roughly 
2000 numbers are used.

I will try to dig more into this and let you know if I find something.

cheers
Detlef

Re: [matplotlib-devel] Bug in pyplot.hist when using histtype="step"

From: Pierre H. <pie...@cr...> - 2013-04-17 12:31:57

Attachments: signature.asc

Hi,

Le 16/04/2013 15:41, Detlef Maurel (IKP) a écrit :
> this works for me, too. Strange... I get the problem with the attached
> dataset using the command
>
> hist(loadtxt("data.txt"),bins=300,histtype="step") 
Indeed, I could reproduce your problem using your specific dataset.

The automated selection of ylim is not working as expected, and I don't
know why !

best,
Pierre