Too small relative size of axes labels w.r.t. tick marks in 2D plots

[egourgoulhon]

Consider this plot:

sage: plot(sin(x^2), (x, -3, 3), axes_labels=['$x$','$y$'])

The size of the labels x and y is quite small and barely readable when the figure is reproduced in a article with moderate size (cf. the attached figure). In the current settings, the only way to enlarge it is via the optional argument fontsize:

sage: plot(sin(x^2), (x, -3, 3), axes_labels=['$x$','$y$'], fontsize=20)

But then the tick marks are too large! One should be able to set separately the size of the axes labels and that of the tick marks.

[egourgoulhon]

New commits:

​10b6924

Add parameter axes_labels_size to Graphics.

[egourgoulhon]

The attached commit introduces an optional parameter for 2D graphics, axes_labels_size, which sets the size of axes labels relative to that of tick marks. The default value of this parameter has been set to 1.7 (the previous behavior would have corresponded to axes_labels_size=1.).

[kcrisman]

Seems fine on the face of it but I truly have nearly zero Sage reviewing time these days. Wait until school lets out and I'll have a mountain of reviews for you, I'm sure :-)

[vdelecroix]

Hello,

If you set

+ self._axes_labels_size = 1.7

in the class Graphics, why would you need this try/except block

+ try:
+     return self._axes_labels_size
+ except AttributeError:
+     self._axes_labels_size = 1.7
+     return self._axes_labels_size

Vincent

[egourgoulhon]

Hi Vincent,

You are perfectly right! I am going to change this. Btw, the same occurs for the attribute _fontsize (probably I did a copy-and-paste of the latter...): there is

self._fontsize = 10

in Graphics.__init__ and

            try:
                return self._fontsize
            except AttributeError:
                self._fontsize = 10
                return self._fontsize

in Graphics.fontsize. Shall we correct it as well, although this does not pertain to the current ticket ?

Replying to vdelecroix:

Hello,

If you set

+ self._axes_labels_size = 1.7

in the class Graphics, why would you need this try/except block

+ try:
+     return self._axes_labels_size
+ except AttributeError:
+     self._axes_labels_size = 1.7
+     return self._axes_labels_size

Vincent

[vdelecroix]

Replying to egourgoulhon:

Hi Vincent,

You are perfectly right! I am going to change this. Btw, the same occurs for the attribute _fontsize (probably I did a copy-and-paste of the latter...): there is

self._fontsize = 10

in Graphics.__init__ and

            try:
                return self._fontsize
            except AttributeError:
                self._fontsize = 10
                return self._fontsize

in Graphics.fontsize. Shall we correct it as well, although this does not pertain to the current ticket ?

Sure. Do it at the same time.

Vincent

[egourgoulhon]

Replying to vdelecroix:

Shall we correct it as well, although this does not pertain to the current ticket ?

Sure. Do it at the same time.

A systematic search for AttributeError in graphics.py reveals that the same issue occurs for the attributes _show_axes, _axes_color, _axes_label_color, _axes_width and _tick_label_colors. So I guess we shall correct these as well?... The only possible justification is that the current implementation is more robust in case of a derived class of Graphics, A let us say, such that A.__init__() does not call Graphics.__init__() (but this is bad design IMHO).

Eric.

[vdelecroix]

Replying to egourgoulhon:

Replying to vdelecroix:

Shall we correct it as well, although this does not pertain to the current ticket ?

Sure. Do it at the same time.

A systematic search for AttributeError in graphics.py reveals that the same issue occurs for the attributes _show_axes, _axes_color, _axes_label_color, _axes_width and _tick_label_colors. So I guess we shall correct these as well?... The only possible justification is that the current implementation is more robust in case of a derived class of Graphics, A let us say, such that A.__init__() does not call Graphics.__init__() (but this is bad design IMHO).

Right. The simplest would be to let it as you initially did... sorry.

Vincent

[egourgoulhon]

Replying to vdelecroix:

Replying to egourgoulhon:

Replying to vdelecroix:

Shall we correct it as well, although this does not pertain to the current ticket ?

Sure. Do it at the same time.

A systematic search for AttributeError in graphics.py reveals that the same issue occurs for the attributes _show_axes, _axes_color, _axes_label_color, _axes_width and _tick_label_colors. So I guess we shall correct these as well?... The only possible justification is that the current implementation is more robust in case of a derived class of Graphics, A let us say, such that A.__init__() does not call Graphics.__init__() (but this is bad design IMHO).

Right. The simplest would be to let it as you initially did... sorry.

OK. Btw note that the case of _axes_width looks strange: in Graphics.__init__(), one has

        self._axes_width = 0.8

while in Graphics.axes_width() there is

            try:
                return self._axes_width
            except AttributeError:
                self._axes_width = True
                return self._axes_width

Shouldn't it be there self._axes_width = 0.8 for consistency ?

Eric.

Vincent

[vdelecroix]

Replying to egourgoulhon:

Shouldn't it be there self._axes_width = 0.8 for consistency ?

True. But I suggest to not do it here and open a new ticket and get rid of all these try/except.

Vincent

[egourgoulhon]

Hi,

Another point that could be discussed is the default value of the new parameter axes_labels_size: I've set it to 1.7 for it looks good for small labels like $x$ and $y$, but maybe it can be lowered a little bit, to 1.6 or 1.5 say, to accomodate for larger labels. Of course, this can be adjusted by the user (the whole point of this ticket!), but it's always nice if the default value fits most usages.

[vdelecroix]

Hello Eric,

You added two trailing whitespaces! This is very bad.

The problem with the right size is that it depends if the label is in math mode (like $x$) or not (like x). I would say that:

• 1.4 is a good default for non math mode
• 1.7 is a good default for math mode

I just tried

plot(x^2, (x,0,1), axes_labels=('x','$x$'), axes_labels_size=1.7)

There is a big difference in the aspects of both letters!

What do people use, label='x' or label='$x$'?

Vincent

[egourgoulhon]

Hi Vincent,

Replying to vdelecroix:

You added two trailing whitespaces! This is very bad.

Sorry about this. Yet I've configured my editor to automatically removing trailing whitespaces while saving a file. Apparently two of them went through...

The problem with the right size is that it depends if the label is in math mode (like $x$) or not (like x). I would say that:

• 1.4 is a good default for non math mode
• 1.7 is a good default for math mode

I just tried

plot(x^2, (x,0,1), axes_labels=('x','$x$'), axes_labels_size=1.7)

There is a big difference in the aspects of both letters!

You are right!

What do people use, label='x' or label='$x$'?

Probably it depends, but if you want to have nice labels (in particular labels that coincide with the symbols used in the main text of your article or labels with subscripts or Greek letters), LaTeX seems mandatory. So I would favor the math mode. Maybe lowered to 1.6 as a compromise?... Eric.

[vdelecroix]

Replying to egourgoulhon:

Replying to vdelecroix: So I would favor the math mode. Maybe lowered to 1.6 as a compromise?...

Looks reasonable to me.

Vincent

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​69354fc

Set the default value of axes_labels_size to 1.6

[egourgoulhon]

Replying to vdelecroix:

Replying to egourgoulhon:

Replying to vdelecroix: So I would favor the math mode. Maybe lowered to 1.6 as a compromise?...

Looks reasonable to me.

All right; in the new commit, I've set the default value to 1.6.

Eric.

[vdelecroix]

Good to go then. Thanks.

Vincent

[egourgoulhon]

Thanks for the review Vincent!

[kcrisman]

See #18421 for followup.