[Bug]: Zoom und pan not working after writing pdf pages. #27865

Erwin-Wert · 2024-03-05T13:23:06Z

Bug summary

Ladies and Gentlemen,
Zoom und pan is not working after writing pdf pages.
Within matplotlib version 3.7.5 it is working.
In e.g. 3.8.3 it does not react any more.

Code for reproduction

import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
fig, ax = plt.subplots()
ax.plot([1, 2, 3], [4, 5, 6])
filename = "test4.pdf"
with PdfPages(filename) as pdf:
    fig.show()
    pdf.savefig() # <== if you comment this line, then zoom is working fine. However it is not getting stored.

Actual outcome

Expected outcome

Additional information

In earlier versions it has been working fine.
It is also not working within jupyterlab.
It was testet in Spyder IDE 5.5.1 and IDLE(Python 3.12 64-bit)
I take the earlier matplotlib version instead.

Operating system

Windows 7,

Matplotlib Version

3.8.3

Matplotlib Backend

QtAgg

Python version

Python 3.12.2 64-bit

Jupyter version

4.1.2

Installation

pip

Impaler343 · 2024-03-06T02:30:48Z

I believe a more appropriate code for reproduction is:

import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages

filename = "test4.pdf"

with PdfPages(filename) as pdf:
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [4, 5, 6])
    pdf.savefig() # <== if you comment this line, then zoom is working fine. However it is not getting stored.

plt.show()

Impaler343 · 2024-03-06T03:01:03Z

To solve this issue, you just need to write plt.show() before saving the figure and pass the figure to the savefig() function. Like:

import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
filename = "test4.pdf"

with PdfPages(filename) as pdf:
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [4, 5, 6])
    plt.show()
    pdf.savefig(fig) # <== if you comment this line, then zoom is working fine. However it is not getting stored.

But you might have to undo changes that you did while inspecting the figure to save the original figure

Impaler343 · 2024-03-06T04:53:02Z

On another note,

The import statement is slightly incorrect from matplotlib.backends.backend_pgf import PdfPages and needs to be backend_pdf instead of backend_pgf. Is it recommended to open a PR in small changes like this one?

QuLogic · 2024-03-06T10:52:21Z

This bisects to a6230cc, or #25546, by @anntzer.

QuLogic · 2024-03-06T11:07:38Z

However, that appears to be exposing an existing different bug. For some reason, this context manager is not restoring the old canvas:

matplotlib/lib/matplotlib/backends/backend_pdf.py

Lines 2758 to 2759 in 9618fc6

    
           with cbook._setattr_cm(figure, canvas=FigureCanvasPdf(figure)): 
        
               figure.savefig(self, format="pdf", **kwargs)

Calling `FigureCanvasPdf(figure)` will call `figure.set_canvas(self)`, meaning the `cbook._setattr_cm` context manager that wraps this change will see the _new_ canvas, and the old canvas isn't restored. Instead, just pass the `backend` parameter, as `savefig` already knows how to correctly save and restore the canvas if it needs to change backends. Fixes matplotlib#27865

QuLogic · 2024-03-06T11:57:05Z

The import statement is slightly incorrect from matplotlib.backends.backend_pgf import PdfPages and needs to be backend_pdf instead of backend_pgf. Is it recommended to open a PR in small changes like this one?

That is not a typo; there are two PdfPages classes, one in the PDF backend and one in the PGF backend. That example uses the PDF backend, and the line is a cross-reference to the PGF backend.

Erwin-Wert · 2024-03-07T08:32:08Z

To solve this issue, you just need to write plt.show() before saving the figure and pass the figure to the savefig() function. Like:
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
filename = "test4.pdf"

with PdfPages(filename) as pdf:
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [4, 5, 6])
    plt.show()
    pdf.savefig(fig) # <== if you comment this line, then zoom is working fine. However it is not getting stored.
But you might have to undo changes that you did while inspecting the figure to save the original figure

The code snippet was just used to reproduce the effect. The real application has multiple figures opened in a tkinter window. Then a function is called to export these figures to the pdf. After that I need the figures to be responsive to zoom and pan. Adding plt.show() opens an extra window with the figure. So this is not an option.

Impaler343 · 2024-03-07T09:21:42Z

To solve this issue, you just need to write plt.show() before saving the figure and pass the figure to the savefig() function. Like:
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
filename = "test4.pdf"

with PdfPages(filename) as pdf:
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [4, 5, 6])
    plt.show()
    pdf.savefig(fig) # <== if you comment this line, then zoom is working fine. However it is not getting stored.
But you might have to undo changes that you did while inspecting the figure to save the original figure
The code snippet was just used to reproduce the effect. The real application has multiple figures opened in a tkinter window. Then a function is called to export these figures to the pdf. After that I need the figures to be responsive to zoom and pan. Adding plt.show() opens an extra window with the figure. So this is not an option.

What I collect from this is that the showing functionality must work after saving the figure. The initial code had a fig.show() function before saving the figure due to which I interpreted it in that way

Calling `FigureCanvasPdf(figure)` will call `figure.set_canvas(self)`, meaning the `cbook._setattr_cm` context manager that wraps this change will see the _new_ canvas, and the old canvas isn't restored. Instead, just pass the `backend` parameter, as `savefig` already knows how to correctly save and restore the canvas if it needs to change backends. Fixes matplotlib#27865

pin pytest Deprecate plot_date Correctly set temporary pdf/pgf backends Calling `FigureCanvasPdf(figure)` will call `figure.set_canvas(self)`, meaning the `cbook._setattr_cm` context manager that wraps this change will see the _new_ canvas, and the old canvas isn't restored. Instead, just pass the `backend` parameter, as `savefig` already knows how to correctly save and restore the canvas if it needs to change backends. Fixes matplotlib#27865 Bump the actions group with 2 updates Bumps the actions group with 2 updates: [pypa/gh-action-pypi-publish](https://github.com/pypa/gh-action-pypi-publish) and [scientific-python/upload-nightly-action](https://github.com/scientific-python/upload-nightly-action). Updates `pypa/gh-action-pypi-publish` from 1.8.11 to 1.8.12 - [Release notes](https://github.com/pypa/gh-action-pypi-publish/releases) - [Commits](pypa/gh-action-pypi-publish@2f6f737...e53eb8b) Updates `scientific-python/upload-nightly-action` from 0.3.0 to 0.5.0 - [Release notes](https://github.com/scientific-python/upload-nightly-action/releases) - [Commits](scientific-python/upload-nightly-action@6e9304f...b67d7fc) --- updated-dependencies: - dependency-name: pypa/gh-action-pypi-publish dependency-type: direct:production update-type: version-update:semver-patch dependency-group: actions - dependency-name: scientific-python/upload-nightly-action dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions ... Signed-off-by: dependabot[bot] <support@github.com> Bump pydata-sphinx-theme to 0.15 Fix search button placement in navbar Fix doc sidebar Fix version switcher url for circleCI Don't use custom CSS for announcement banner Disable parallel build Don't show doc source link Add dtype/copy args to internal testing class Use pybind11 string formatter for exception messages This can eventually be replaced by `std::format` when we require C++20. doc: add description of **kwargs usage to collections (matplotlib#27872) * doc: add description of **kwargs usage to collections * Update lib/matplotlib/collections.py Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> --------- Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> TST: adding tests of current clear behavior on ticks closes matplotlib#23839 DOC: update comments and docstrings Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> Update lib/matplotlib/tests/test_axes.py BLD,Cygwin: Included Python.h first in src/_c_internal_utils.cpp Python.h needs to be included before any system includes. This is easiest if it is the first include. Interestingly, pybind11/pybind11.h does not include Python, possibly depending on downstream projects including Python.h beforehand. CI: Include Python.h first in _tkagg.cpp Needs to be included before system headers to set visibility macros for pybind11 BLD: Include Python.h first in _backend_agg.cpp BLD: Include Python.h before pybind11/pybind11.h pybind11 assumes Python.h is included first. CI: Specify python version for default pytest. CI,Cygwin: Run pytest with python -m pytest Cygwin pytest recently updated, and only installs pytest-3.9 CI,Cygwin: Revert use of alternatives to set pytest script. Cygwin pytest only installs pytest-3.9. I should possibly suggest using alternatives for that and pip. CI,Cygwin: Avoid running pytest in root directory. CI,Cygwin: Revert running pytest in different directory. CI,Cygwin: Call pytest-3.9 explicitly. BLD: Include Python.h first in src/_path.h BLD,BUG: Remove Python.h includes immediately preceeding pybind11/pybind11.h includes pybind11 seems decent about including Python.h before any system headers, once you chase three layers of includes in to see that. FIX: handle nans in RGBA input with ScalarMappables DOC: Update some animation related topics - use `set_data_3d` - cross-reference `Line2d` `set_data`, `set_xdata`, `set_ydata` - Rewrite parts of the FuncAnimation description Inspired through matplotlib#27830. Revert "Merge branch 'matplotlib:main' into doc-change" This reverts commit feeabe6, reversing changes made to bcbc2ef. Reapply "Merge branch 'matplotlib:main' into doc-change" This reverts commit ed91cee. Convert path extension to pybind11 Add a pybind11 type caster for agg::rect_d Add a pybind11 type caster for agg::trans_affine Add a pybind11 type caster for mpl::PathIterator Add a pybind11 type caster for e_snap_mode Add a pybind11 type caster for SketchParams Optimize convert_polygon_vector a bit FIX: don't copy twice on RGB input Fix devdocs version switcher Allow passing a transformation to secondary_[xy]axis Add transform argument to secondary axes Update _secax_docstring Move new params to end of functions Add input check to secondary axes Add tests Add examples Move transform type checks and improve docs Add type stubs Update _secax_docstring Move new params to end of functions Add input check to secondary axes Move transform type checks and improve docs Fix rebase error Fix stub for SecondaryAxis.__init__ Clarify example Add default param to secax constructor Fix stub for secax constructor Simplify imports Co-authored-by: Elliott Sales de Andrade <quantum.analyst@gmail.com> Remove redundancy in docs Co-authored-by: Elliott Sales de Andrade <quantum.analyst@gmail.com> Fix typo DOC: fix stray release note entry BLD: Add a fallback URL for FreeType FreeType is available from both Savannah and SourceForge, and sometimes AppVeyor seems to have trouble downloading from Savannah, so perhaps this fallback will help. We used to try both these URLs in the pre-Meson build system. DOC: State approximate documentation build time Since the doc build takes very long, an order-of-magnitude estimate is helpful to manage expectations. I'm not wedded to the actual numbers, if somebody has better suggestions. A quick test on an old i7 6700 (from 2015) yielded - `O=-j1`: 23min - `O=-j4`: 15min Doc build on CI is around 18min. Add BackendRegistry singleton class Use backend_registry for name of singleton instance Use class variables for immutable collections Review comments Update lib/matplotlib/backend_bases.py Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> Linting/mypy fixes Remove INTERACTIVE_NON_WEB backend filter Small changes from review Use _api.caching_module_getattr for deprecated module-level attributes Add docstrings Add api changes and what new docs Fix docs Inline the deprecation function calls Import BackendFilter and backend_registry into matplotlib.backends.__init__.py Mypy fixes Remove unused _safe_pyplot_import Remove unneeded type annotations Make sure custom alpha param does not change 'none' colors in a list of colors (matplotlib#27845) * fix issue 27839, make sure alpha param does not affect none colors Apply review feedbacks Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> * Update lib/matplotlib/colors.py Co-authored-by: Elliott Sales de Andrade <quantum.analyst@gmail.com> --------- Co-authored-by: Tim Hoffmann <2836374+timhoffm@users.noreply.github.com> Co-authored-by: Elliott Sales de Andrade <quantum.analyst@gmail.com> DOC: fix :mpltype:`color` role `inline.interpreted()` already returns a list of nodes. We mustn't wrap it in another list. Use :mpltype:`color` for color types

QuLogic added the status: confirmed bug label Mar 6, 2024

QuLogic mentioned this issue Mar 6, 2024

Correctly set temporary pdf/pgf backends #27869

Merged

5 tasks

timhoffm closed this as completed in #27869 Mar 6, 2024

QuLogic added this to the v3.9.0 milestone Mar 6, 2024

QuLogic mentioned this issue Apr 1, 2024

[Bug]: Cannot show figure after using PdfPages savefig #28004

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

[Bug]: Zoom und pan not working after writing pdf pages. #27865

[Bug]: Zoom und pan not working after writing pdf pages. #27865

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

[Bug]: Zoom und pan not working after writing pdf pages. #27865

[Bug]: Zoom und pan not working after writing pdf pages. #27865

Comments

Uh oh!

Bug summary

Code for reproduction

Actual outcome

Expected outcome

Additional information

Operating system

Matplotlib Version

Matplotlib Backend

Python version

Jupyter version

Installation

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!