Benefit of dynamic default backend; list issue titles.

anntzer · anntzer · commit a1dff3fe54b3 · 2017-10-23T01:47:41.000-07:00
diff --git a/doc/devel/MEP/MEP32.rst b/doc/devel/MEP/MEP32.rst
@@ -27,9 +27,7 @@ Detailed description
 The problem
 ~~~~~~~~~~~
 
-Current ("old-style") configuration files use a custom syntax, of the form
-
-.. code:: conf
+Current ("old-style") configuration files use a custom syntax, of the form ::
 
    key: value  # possible comment
    key: value
@@ -115,9 +113,7 @@ which file-parser to use.
 Maintain a matplotlibrc-like syntax
 ```````````````````````````````````
 
-The config files would maintain the format
-
-.. code:: conf
+The config files would maintain the format ::
 
    key: value  # possible comment
    key: value
@@ -205,6 +201,34 @@ convention is changed (e.g. ``.mplstyle-v2`` with the matplotlibrc-like syntax,
 or ``.py``\(!) with the full Python syntax), then a simple priority policy
 (such as "always prefer the newer syntax if available") can be implemented.
 
+Runtime detection for the default backend
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+As an additional benefit, either approach would allow setting the backend to
+the first of a list of supported backends *at import time*.  Specifically,
+``setup.py`` currently checks *at build time* what is the "best" "available"
+backend, where "best" is defined by a fixed ordering (OSX > Qt5Agg > ...) and
+"available" means that the platform is supported (macosx is only supported on
+OSX) and the dependencies are installed (PyQt5 for Qt5Agg, etc.).  The result
+of this check is written to the default ``matplotlibrc``, and used as the
+default backend.  Not only this code is fairly complex (in particular because
+one can generally not import multiple GUI toolkits in a single process -- in
+truth this could be fixed by just checking whether the module is available,
+*without* importing it), but more importantly, this default backend can
+easily be incorrect if the build environment is not the same as the install
+environment (quite common with wheels, conda packages and Linux packages),
+or become incorrect if the end-user installs or uninstalls some GUI toolkits
+(`#3679`_, `#8613`_).  The desired default backend may even depend on whether
+the environement is headless (`#6739`_).
+
+With an interpreted ``matplotlibrc``, this problem disappears: we can simply
+provide a ``first_available_backend`` function that finds *at runtime* the
+first available backend from a list, and write something like ::
+
+   rcParams["backend"] = first_available_backend(["macosx", "qt5agg", ...])
+
+to the default ``matplotlibrc``.
+
 Implementation
 --------------
 
@@ -241,16 +265,39 @@ Alternatives
   discussion that took place over PEP518_ (not that I particularly like the
   final choice of yet another obscure configuration language by that PEP).
 
+Relevant issues
+---------------
+
+- `#3670`_: "Document that rcparam values are unquoted"
+- `#3679`_: "Runtime detection for default backend"
+- `#4240`_: "Style RC parameter"
+- `#4371`_: "LaTeX and PGF preambles do not allow commas"
+- `#6157`_: "Add style for xkcd"
+- `#6274`_: "A feeble attempt to plug a theoretical security hole"
+- `#6320`_: "all matplotlibrc files should be loaded in order"
+- `#6739`_: "Make matplotlib fail more gracefully in headless environments"
+- `#7089`_: "matplotlibrc reader cannot handle rcparams with a hash"
+- `#7219`_: "Cycler passed as string for rc"
+- `#8235`_: "Investigate why some examples build properly in our gallery
+  despite using numpy and not importing it explicitely"
+- `#8430`_: "[feature request] move default colors into styles"
+- `#8613`_: "matplotlibrc: is there a way to specify "backend" as either
+  "Qt5Agg | Qt4Agg" (whatever is installed)?"
+- `#9184`_: "matplotlibrc parsing is inconsistent with matplotlibrc.template"
+
 .. _#3670: https://github.com/matplotlib/matplotlib/issues/3670
+.. _#3679: https://github.com/matplotlib/matplotlib/issues/3679
 .. _#4240: https://github.com/matplotlib/matplotlib/issues/4240
 .. _#4371: https://github.com/matplotlib/matplotlib/issues/4371
 .. _#6157: https://github.com/matplotlib/matplotlib/issues/6157
 .. _#6274: https://github.com/matplotlib/matplotlib/issues/6274
 .. _#6320: https://github.com/matplotlib/matplotlib/issues/6320
+.. _#6739: https://github.com/matplotlib/matplotlib/issues/6739
 .. _#7089: https://github.com/matplotlib/matplotlib/issues/7089
 .. _#7219: https://github.com/matplotlib/matplotlib/issues/7219
-.. _#8235: https://github.com/matplotlib/matplotlib/issues/8430
+.. _#8235: https://github.com/matplotlib/matplotlib/issues/8235
 .. _#8430: https://github.com/matplotlib/matplotlib/issues/8430
+.. _#8613: https://github.com/matplotlib/matplotlib/issues/8613
 .. _#9184: https://github.com/matplotlib/matplotlib/issues/9184
 .. _PEP518: https://www.python.org/dev/peps/pep-0518/#other-file-formats
 .. _explicit-imports: https://www.reddit.com/r/Python/comments/ex54j/seeking_clarification_on_pylonsturbogearspyramid/c1bo1v5/
