pyqtgraph not compatible with matplotlib

[Jianbao Tao]

Hi,

I am playing with pyqtgraph recently. I am impressed by its speed. :-) However, there is one thing that really bothers me: It doesn't appear  to be compatible with matplotlib, which is very strange, considering the fact that pyqtgraph and matplotlib have their own name space. Here is a snippet that demonstrates that problem.

#--------------------- code ----------------------------------

import matplotlib.pyplot as plt

import numpy as np

x = np.arange(100)

plt.plot(x)  # Working

plt.show()

import pyqtgraph

plt.plot(x)  # Not working

#----------------------- end of code -------------------------

Cheers,

Jianbao

[Jianbao Tao]

I forgot to mention my platform.

pyqtgraph version: pyqtgraph-dev-r201.tar.gz

OS: Mountain lion (OS X 10.8.2)

Python: 2.7.2 (apple version)

matplotlib: 1.2.0rc2

Let me know if you need more information to diagnose the problem.

Jianbao

[Luke Campagnola]

I bet the problem is that you can't use two different GUI libraries in the same application. In this case, pyplot probably creates a window with TkInter, which prevents Qt from loading properly. Instead, use matplotlib's Qt backend to ensure that the two libraries play nicely together.  

(see pyqtgraph/widgets/MatplotlibWidget.py )

Luke

[Jianbao Tao]

Luke,

With all due respect, your bet is wrong. I used the Qt4Agg for matplotlib backend, and the same problem is still there. Besides, in the example code, it is not matplotlib that makes pyqtgraph not working. It is pyqtgraph that makes matplotlib not working. So, your argument doesn't really make too much sense to me.

Jianbao

-- [ You are subscribed to pyqt...@googlegroups.com. To unsubscribe, send email to pyqtgraph+...@googlegroups.com ]

[Luke Campagnola]

On Tue, Oct 9, 2012 at 11:43 AM, Jianbao Tao <jianb...@gmail.com> wrote:

With all due respect, your bet is wrong. I used the Qt4Agg for matplotlib backend, and the same problem is still there. Besides, in the example code, it is not matplotlib that makes pyqtgraph not working. It is pyqtgraph that makes matplotlib not working. So, your argument doesn't really make too much sense to me.

Sorry, I did not read your question carefully enough. You are absolutely correct--pyqtgraph has a wrapper function that it installs around numpy.concatenate as a workaround for a segmentation fault bug in numpy.  My implementation of that wrapper is not quite correct, and causes matplotlib to stop working. 

This is fixed in the latest release.

Thanks!

Luke

 

[Jianbao Tao]

Thanks for following up on this, Luke. I will try the release later.

Jianbao

-- [ You are subscribed to pyqt...@googlegroups.com. To unsubscribe, send email to pyqtgraph+...@googlegroups.com ]

[ML___]

Hello,

Thanks to all those who've worked on PyQtGraph.  I switched to it for a real time plotting app as I was bogging down Matplotlib. 

It's easy to use, highly flexible, and fast!

I'm seeing a similar issue as was reported in 2012, where importing pyqtgraph stops matplotlib from showing plots:

--------------------my code------------------------------

import matplotlib.pyplot as plt

plt.ion()

plt.show()

plt.plot([1,2,3,4])    # 1) call succeeds, interactive plot window is shown

import pyqtgraph  # 2) after this call the existing plot window from (1) is unresponsive

plt.plot([4,3,2,1])    # 3) function succeeds as in (1), but plot window is not updated, and remains unresponsive

-----------------------------------------------------------------

I'm running Debian Testing (bookworm) recently updated

Python 3.10.4

pyqtgraph.__version__  = 0.12.4

matplotlib.__version__  = 3.5.1

Regards,

Michael

[Ognyan Moore]

The only support that pyqtgraph has with matplotlib is that you can export windows to a matplotlib window, there is a little more in the documentation here: https://pyqtgraph.readthedocs.io/en/latest/exporting.html#export-formats

Trying to use both matplotlib and pyqtgraph at the same time will likely lead to unintentionally bad behavior from some issues w/ QApplication I would imagine.  If you specify a non Qt back-end of matplotlib, it might work, but this is fairly untested, and I would generally not recommend.

--
You received this message because you are subscribed to the Google Groups "pyqtgraph" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pyqtgraph+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pyqtgraph/496afc6c-3c4c-44e0-bb28-2f1ef7c67860n%40googlegroups.com.

[Patrick]

Hi,

I have successfully used both pyqtgraph and matplotlib together in a data processing/analysis project. I used pyqtgraph for the most part, all the interactive plots etc, and then matplotlib for the final step of generating publication-quality plots. It's something I always wanted to open-source but never really sorted out the details for that, but I'll try to summarise here. Perhaps email me directly if you have problems and I may be able to sort out and supply some better code.

Imports look something like:

# Do usual Qt imports, pyqtgraph and matplotlib

from PyQt5 import QtWidgets, uic
import pyqtgraph as pg
import matplotlib as mpl
mpl.use('Qt5Agg')

# For matplotlib plots in the GUI

from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas

# For matplotlib plot exports to pdf

from matplotlib.backends.backend_pdf import FigureCanvasPdf, PdfPages

In relevant UI class, set up a matplotlib FigureCanvas widget to render onto:

self.resultplot_figureCanvas = FigureCanvas(mpl.figure.Figure(constrained_layout=True))

self.resultplots_groupBox.layout().addWidget(self.resultplot_figureCanvas)

In appropriate place, render matplotlib plot to the FigureCanvas instead of letting mpl.pyplot create one for you:

fig = self.resultplot_figureCanvas.figure
fig.clear()
ax = fig.subplots(1, 1)
ax.plot([1, 2, 3], [4, 5, 6])
#...
self.resultplot_figureCanvas.draw()

If you want to render to a pdf, then render to a FigureCanvasPdf something like:

canvas = FigureCanvasPdf(mpl.figure.Figure(constrained_layout=True, figsize=(figwidth, figheight)))
fig = canvas.figure
ax = fig.subplots(1, 1)
ax.plot([1, 2, 3], [4, 5, 6])
# ...
fig.savefig("filename.pdf", bbox_inches="tight", dpi=150, metadata=metadata_pdf)

The application is a little old now (Qt5 etc), but hopefully that might give you something to work with.

Patrick

[ML___]

Hi Patrick,

Thanks for all the info, will have a look shortly, and report back.

Michael

[ML___]

Hi,

So I have matplotlib and pyqtgraph coexisting.   Not sure exactly what changed, but had something to do with pyqtgraph auto selecting

pyqt5.  Got my imports in order and now my app works as desired with pyside2, and separate matplotlibs scripts work as well. 

Michael

[Ognyan Moore]

If you have multiple Qt versions in your environment, you can force pyqtgraph to use a specific binding by setting the PYQTGRAPH_QT_LIB environment variable:

https://github.com/pyqtgraph/pyqtgraph/blob/master/pyqtgraph/Qt/__init__.py#L23

Unfortunately, this functionality isn't documented, but we have no plans in removing it.  Having multiple Qt bindings is an issue for people who install matplotlib using the conda defaults channel (which installs PyQt5 5.9, which pyqtgraph does not support).  A couple of other workarounds, in that case, involve installing matplotlib via pip, or installing the conda-forge variant of matplotlib.

--
You received this message because you are subscribed to the Google Groups "pyqtgraph" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pyqtgraph+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/pyqtgraph/64a22a45-ebaf-4d01-a9fa-80e314e48103n%40googlegroups.com.