Make Text and Tex robust against non-existing text_dir and tex_dir

[behackl]

List of Changes

• When creating Tex and Text, check whether tex_dir and text_dir exist, respectively -- and if they don't, create them.
• Remove TextWithBackground from text_mobject.py. This was a non-working feature that basically copied code from Text. Code like that makes manim extremely hard to maintain in the long run. (Actually, also currently.)

Motivation

Making Tex and Text robust against not created directories is required to use them in a setting where there is no Scene to set everything up (e.g., when writing tests for them outside of a scene, see #476).

Testing Status

Added doctests for testing initialization of Tex and Text.

Acknowledgement

• I have read the Contributing Guidelines

[behackl]

Okay, I've added some very simple __repr__ methods to both Text and SingleStringMathTex (from which all other TeX classes inherit -- weird, but maybe a story for another day) and I've added a bunch of more doctests so that creating text and tex is always tested.

Review at will!

[huguesdevimeux]

I know you're going to hate me for that but ...

Do you think you could write a test for that ?

[behackl]

I know you're going to hate me for that but ...

Do you think you could write a test for that ?

It is, in fact, already tested with the doctests I have added. :-)

[huguesdevimeux]

I know you're going to hate me for that but ...
Do you think you could write a test for that ?

It is, in fact, already tested with the doctests I have added. :-)

Could you clarify this ? I don't understand

[behackl]

I know you're going to hate me for that but ...
Do you think you could write a test for that ?

It is, in fact, already tested with the doctests I have added. :-)

Could you clarify this ? I don't understand

@huguesdevimeux The creation of Text etc is tested in https://github.com/ManimCommunity/manim/pull/492/files#diff-cf9247b89e1b4debba20823314a969f8R86 -- without an existing text_dir etc, these tests would fail.

If you would like to have a more explicit test, let me know what precisely you would like to have tested. I'll probably have to adapt to the PR that enabled Pango anyways.

[behackl]

There was a merge conflict in the TextWithBackground class (the inheritance changed from Text to CairoText) -- but I still argue that it should be removed. It is not functional, and if someone wants to add it at a later point, then they are welcome to do so.

I don't find the current co-existance of PangoText and CairoText with Text still defaulting to CairoText a particularly good idea, and I think we should have switched the default and logged information that in case of problems one should fall back to the old implementation, CairoText -- but this is probably a discussion for a different issue / PR.

Anyhow, the reason I bring this up is the representation string: currently, CairoText('asdf') and Text('asdf') both have Text('asdf') as their representation string (and I think this is okay, as Text is just a wrapper with the logged warning). In 6e50d94 I've added a __repr__ to PangoText that lets PangoText('asdf') also have Text('asdf') as its representation string, to keep this sort of "symmetry".

[huguesdevimeux]

If you would like to have a more explicit test, let me know what precisely you would like to have tested. I'll probably have to adapt to the PR that enabled Pango anyways.

I hear that, but I do think that at some point we will need a test for the whole directory generation.
So, it's fine here, a test will be implemented in the future

[behackl]

I would appreciate if this could get a second approval so I can stop resolving merge conflicts over and over. 😅

[huguesdevimeux]

SGTM

[nilaybhatia]

LGTM!

[behackl]

Thank you!