One related aspect to think about: Do we anticipate that we'll need a distinction between x and y axis as well? Because if so, we should directly introduce these.

I made the suggested changes.

One related aspect to think about: Do we anticipate that we'll need a distinction between x and y axis as well? Because if so, we should directly introduce these.

Can you clarify what you meant by "a distinction between x and y axis"? Do you mean something like

rcParams["grid.major.xaxis.linestyle"]=":"rcParams["grid.major.yaxis.linestyle"]="-"

Sorry I forgot to update the ".pyi" file

between x and y axis"? Do you mean something like

rcParams["grid.major.xaxis.linestyle"]=":"rcParams["grid.major.yaxis.linestyle"]="-"

Basically yes. Exact spelling t.b.d.

Basically yes. Exact spelling t.b.d.

I think this is possible however it might look "ugly". The change will require theTick class to know which child (XTick,YTick) was instantiated since the lines are created in the parent class.

One possible way is by accessing theself.__name__ attribute which should either be"xtick" or"ytick" since the base class is never instantiated. However, this might cause problems if this attribute ever changes which might be hard to debug. It is also possible to add a new attribute to theTick class which will default toNone and will be set by the child class to be used explicitly for this purpose so that we don't use the__name__ attribute.

Another way is to add an extra required argument to the constructor ofTick, probably an enum or a bool. For instance, an argumentaxis that will be either 1 or 2 ifXTick orYTick has called theTick.__init__ method, respectively.

Any thoughts on those?

We already rely onTick.__name__ see

It's ok to use that.

Spelling: We have "xaxis", "xticks", "xscale" as axis-dependent names (but they are all also used in the python interface). We do not have any xgrid or similar names. Options:

• "xgrid.major.linestyle" - in analogy to "xticks"
• "grid.xaxis.major.linestyle" / "grid.x.major.linestyle" - still put it under the "grid" top level group
• "xtick.major.grid.linestyle" - put everything under the corresponding tick
• "xaxis.grid.major.linestyle" - put everything under the corresponding axis

None of them look like the obvious right choice to me.

Out of the four, I prefer two:

• "xtick.major.grid.linestyle" - many options already exist under "xtick.major.*" so it is logical.
• "grid.xaxis.major.linestyle" - configures similar to both axis. Also, it will be in a similar place in the matplotlibrc making easier to find if you are looking for it.

I will try to implement the second one for now. I can change the key later.

I implemented the x and y axis distinction however I came across something that needs further clarification. Right now "grid.xaxis.major.linestyle" is a valid key but "grid.xaxis.linestyle" is not. Should it be a valid option as well? If so it is unclear which of "grid.major.linestyle"and "grid.xaxis.linestyle" should take precedence.

No, we don't need "grid.xaxis.linestyle". If starting from scratch, I'd just have the granular settings "grid.xaxis.major.linestyle" etc. After all, these are advanced style configuration options, and if somebody cares to adapt them, it's ok to write out the details. We keep the high-level ones "grid.linestyle" for backward compatibility (and it is a nice convenience) but we don't need the middle-ground.

I added the What's New note. I am not if the syntax is correct for the rc parameters. I have never used rst before so let me know if it needs correcting. Also, I tried to resolve the stubtest unsuccessfully.

Also, I tried to resolve the stubtest unsuccessfully.

I suppose you've tricked yourself. You've made the_validate_linestyle parameter names consistent, which renders the allowlist entry unnecessary (and mypy complains on that). Then, you've repeated the same for_validate_linestyle_or_None.

Try removing both allowlist entries.

Do you want me to create the test in the file "test_rcparams.py" or create a separate file?

Please put it intest_axis.py. This is where the semantic code lives and where it logically belongs.test_rcparams.py is only for the rcparams infrastructure not testind individual values.

I was working on the test and I realised that theget_gridlines() only returns the major gridlines. If it's okay I will introduce two new functionsget_major_gridlines() andget_minor_gridlines() to be explicit. That way I will able to easily extract the minor gridlines as well. Also, the original function will now call theget_major_gridlines() to be backwards compatible

Then useax.xaxis.get_major_ticks()[0].gridline. There are various ways to drill down, and not all are complete or consistent. Sorting/consolidating that is for another time. Let's not create more functions as part of this PR.

I fix the spelling so all tests should pass now. If there is anything else please let me know.

@timhoffm Any idea when this might be merged?

This needs a second review by a core developer, and review time is scarce. However, this is flagged for 3.11 and I'm pretty sure it will get the attention in time for the next release.