Repository navigation
imshow size limitation #2105
Description
imshow() and matshow() fail to display matrices larger than 2^15 in horizontal dimension,
e.g.:
M = np.ones((100,100000))
plt.imshow(M, cmap='jet', aspect='auto', alpha = 1, origin = 'lower', interpolation = 'none')
plt.show()
matplotlib 1.2.1,
ubuntu gnome 12.10
Activity
tacaswell commentedon Jun 1, 2013
I have also had issues with this.
On the other hand, one should ask why you are displaying something several orders of magnitude wider than your screen.
dmcdougall commentedon Jun 23, 2013
@NeedAName @tacaswell Is there an error message?
I've run into size limitations with Agg, but I get an error message telling me what the problem is. I can understand writing a big image file to disk rather than displaying on the screen.
tacaswell commentedon Jun 23, 2013
No error that I see (doing this interactively). Parts or all of the image just don't show up.
If you zoom in on a small region, it behaves as expected.
mdboom commentedon Jun 24, 2013
Which backend is failing for the OP @NeedAName or @tacaswell ?
tacaswell commentedon Jun 24, 2013
qt4agg (pysides)
pelson commentedon Aug 12, 2013
Do you really want to show an image which is 100000 pixels wide? I'm guessing you don't have a monitor which is capable of displaying each pixel.
Is it that you really want to make use of some downsampling algorithm first? I've been working on a bit of an example demonstrating the use of lazy arrays and sending them over to a matplotlib image for fast, lazy, subsampling. This would allow for zoomed in images of arbitrary size (I've been testing with 52428 ** 2 pixels) to be drawn without destroying a conventional machine... is that the kind of thing you are after?
tacaswell commentedon Aug 12, 2013
In my data each column is a time step and I have interesting features at the 100, 1,000, and 10,000 time scales so I am using this to a) very inefficiently down-sample my data for display by letting the rasterization code do it and b) easy data-exploration.
The lazy array example sounds exactly like what I want to use this for.
mdboom commentedon Aug 12, 2013
@pelson: That lazy array example sounds really nice to have. I wouldn't want this to hold up inclusion of something like that, but we should look to that to perhaps influence MEP18 about "plottable objects" and how or if we would create an abstract interface to a more dynamic interface to the data. (I'm basically just noting this for myself or someone else when they get around to writing MEP18).
petehuang commentedon Jan 7, 2017
Reconfirming on 1.5.3 and TkAgg
dstansby commentedon Dec 29, 2022
The original code snipped seems to work fine for me on both
macosxandQt5Aggbackends, so closing. If this is still an issue with a recent release, please re-open or post a comment!