There is a bit of discussion in #10911 (comment) and following comments.

As to why it isn't implemented, the answer is simply noone has done it yet. The tick params used to be a big mess; by now it has become much cleaner code-wise. So maybe now is the time to actually add the missing properties.

Note: The horizontal alignment of Text is is something different compared to what's probably expected from horizontal alignment of tick labels. Text aligns itself relative to its reference point. The ticks should probably align to a common reference line. See the discussion in #13440.

I'm not in favour of overloading the notion horizontalalignment with different things. What I would expect and find useful is exactly what horizontalalignment does for text. Though of course one can introduce additional options like horizontalalignment="leftjustify" for the thing proposed in #13440.

I‘m not suggesting if or how to handle this. But horizontalalignment="left" is probably not that useful here. It basically would result in this example: #13440 (comment)

I don't think we need to question the usefulness. But let me reiterate.

As an example, consider using long ticklabels as in A simple categorical heatmap. You want them to be right-aligned and rotated

which is achieved using setters via

plt.setp(ax.get_xticklabels(), rotation=45, ha="right", rotation_mode="anchor")

One could hence argue that setters are the way to set the properties of ticklabels in matplotlib. But, since the introduction of tick_params, there are some cases where those properties are overwritten by a more general axis-wide setting. An example of this is precisely #10911 (comment). It was hence argued that tick_params should be the recommended interface to set those properties; but if that is to be the case, clearly horizontalalignment, verticalalignment and rotation_mode are missing.

So more generally, one could summarize this into the question: What is the recommended interface to set axis label properties in matplotlib?
(And please read through #10911 completely, before trying to answer that question. Also, a similar issue is #10882)

Disclaimer: Did not take the time to read #10911 completely, but still want to give short feedback. Please let me know in case I'm completely out of context.

From my naive point of view, your use case looks valid; and I would expect to be able to use tick_params there. Shall we maybe restrict horizontalalignment to xaxis and verticalalignment to yaxis` to not raise any false expectations concerning the not supported justify-alignment.

Following the principle of least astonishment I'd say that if specifiying ax.tick_params(axis="y", horizontalalignment="left") one would expect the horizontalalignment to apply to the y axis. This is still useful in conjunction with pad, e.g. ax.tick_params(axis="y", horizontalalignment="left", pad=-5) would give:

or ax.tick_params(axis="y", horizontalalignment="left", pad=20)

Worth noting: The "justify" case can currently be accomplished via mpl_toolkits.axisartist as shown in the demo_ticklabel_alignment example.

@ImportanceOfBeingErnest you are right with your example.

Should we maybe allow the alignment parameters only for single axis? For ax.tick_params(axis="y", horizontalalignment="left") the alignment has quite different effects and it would be a rare usecase that this is desired behavior. OTOH it's not too bad leaving this up to the user.

+1

+1

xref #20644

With 3.9.2, I haven't be able to control anchor position for xticklabels especially in a secondary_xaxis.
The parameter 'ha' is missing in

second_xaxis.tick_params(labelrotation=30, labelsize=8)

Am I missing something ?

@PBrockmann there is no out-of-the-box solution yet because the meaning of alignment is somewhat subtle. See the discussions above for details.

Therefore, alignment is currently not available in tick_params. As a workaround, you can set the alignment on the individual ticklabel instances plt.setp(second_xaxis.get_xticklabels(), horizontalalignment='left'). This works for static plots, i.e. as long as you don't modify the ticks afterwards.

Ok thanks ! It works.
See https://github.com/PBrockmann/PyAnalyseries

I was expecting a control on a specific axe rather than with plt.