Refactor StringMobject and relevant classes

[YishiMichael]

Motivation

Refactor StringMobject (renamed from LabelledString), MarkupText, MTex classes.

Proposed changes

• M manimlib/mobject/svg/labelled_string.py: Unify logic from MarkupText, MTex; remove use_plain_tex attributes; support a much more flexible Selector type; allow passing in substrings whose braces or tags aren't paired; rename some methods.
• M manimlib/mobject/svg/mtex_mobject.py: Allow passing in a 2-tuple to tex_environment, regarded as a prefix and a suffix respectively; remove use_plain_tex attributes.
• M manimlib/mobject/svg/text_mobject.py: Parse entities to prevent from cutting halfway; remove use_plain_tex attributes.
• M manimlib/animation/creation.py
• M manimlib/animation/transform_matching_parts.py
• M manimlib/scene/interactive_scene.py: Add from __future__ import annotations.

The method renaming in StringMobject class includes:

get_parts_by_string => select_parts
get_part_by_string => select_part
set_color_by_string => set_parts_color
set_color_by_string_to_color_map => set_parts_color_by_dict

The names of aliases remain unchanged (e.g. MTex.get_parts_by_tex)

The type definition of Selector is given below:

Selector = Union[
    str,  # Match all identical strings
    re.Pattern,  # Match the pattern
    tuple[Union[int, None], Union[int, None]],  # A normal span
    Iterable[Union[  # A collection of the above
        str,
        re.Pattern,
        tuple[Union[int, None], Union[int, None]]
    ]]
]

The StringMobject.select_parts method takes in a selector as an input. The isolate parameter, keys of tex_to_color_map, keys of t2c, etc. are also of type selector. StringMobject.select_parts no longer supports case_sensitive, substring and regex parameters --- only substrings inside spans specified by selectors will be selected.

# Equivalent:
text_ = Text("abc", isolate="ab")
#text_ = Text("abc", isolate=["ab"])
#text_ = Text("abc", isolate=(0, 2))
#text_ = Text("abc", isolate=(None, -1))

# Equivalent:
text = Text("abcBC", isolate=("a", "bc", "BC"))
#text = Text("abcBC", isolate=("a", re.compile("bc", re.I)))
#text = Text("abcBC", isolate=((0, 1), (1, 3), (3, 5)))
#text = Text("abcBC", isolate=((None, 1), (1, 3), (-2, None)))

# Equivalent:
text.select_part("bcBC").set_fill(BLUE)
#text.select_part(("bc", "BC")).set_fill(BLUE)
#text.select_part(re.compile("bc", re.I)).set_fill(BLUE)
#text.select_part((1, None)).set_fill(BLUE)

# Equivalent:
poly = MTex(
    "p(x) = a_0 x^0 + a_1 x^1 + \\cdots + a_{n-1} x^{n-1} + a_n x^n",
    tex_to_color_map={re.compile(r"a_(.+?) x\^\1"): ORANGE}
)
#poly = MTex(
#    "p(x) = a_0 x^0 + a_1 x^1 + \\cdots + a_{n-1} x^{n-1} + a_n x^n",
#    tex_to_color_map={
#        s: ORANGE
#        for s in ["a_0 x^0", "a_1 x^1", "a_{n-1} x^{n-1}", "a_n x^n"]
#    }
#)

self.add(VGroup(text_, text, poly).arrange(DOWN))

This PR handles imbalanced strings and entity strings, so the following will work:

tex = MTex("\\frac{-b\\pm\\sqrt{b^2-4ac}}{2a}", tex_to_color_map={"{b^2-4ac": TEAL})
text = MarkupText("<tt>Hello Markup</tt>!", t2c={re.compile(r"\W+"): RED})
self.add(VGroup(tex, text).arrange(DOWN))

This PR fixes #1792 by removing line_height from the global attribute dictionary if Pango doesn't support.

[YishiMichael]

I notice that after refactoring Mobject.copy method, the TransformMatchingStrings animation becomes really slow, and sometimes a RecursionError is raised. I wonder how this happens, and how to fix it.

Code

kw = {
    "isolate": [
        "=",
        "\\int_{0}^{\\infty} \\mathrm{e}^{- t}",
        "\\mathrm{d} t"
    ],
    "tex_to_color_map": {
        "{z}": YELLOW,
        "\\mathrm{e}": MAROON_A,
        "c_{0}": BLUE,
        "c_{1}": BLUE,
        "c_{2}": BLUE,
        "c_{n}": BLUE,
        "{p}": BLUE_B,
        "P \\left( t \\right)": BLUE
    }
}
explicit_formula = MTex(
    "= \\int_{0}^{\\infty} \\mathrm{e}^{- t}"
    " \\left( c_{0} + c_{1} t + c_{2} t^{2} + \\cdots + c_{n} t^{n} \\right) \\mathrm{d} t",
    **kw
)
expanded_formula = MTex(
    "= \\int_{0}^{\\infty} \\mathrm{e}^{- t} c_{0} \\mathrm{d} t"
    " + \\int_{0}^{\\infty} \\mathrm{e}^{- t} c_{1} t \\mathrm{d} t"
    " + \\int_{0}^{\\infty} \\mathrm{e}^{- t} c_{2} t^{2} \\mathrm{d} t"
    " + \\cdots"
    " + \\int_{0}^{\\infty} \\mathrm{e}^{- t} c_{n} t^{n} \\mathrm{d} t",
    **kw
).shift(DOWN)
self.add(explicit_formula)
self.wait()
self.play(TransformMatchingStrings(explicit_formula, expanded_formula))
self.wait()

[3b1b]

I'll fix up Mobject.copy to avoid the recursive explosion there—apologies for a hasty push there.

[YishiMichael]

I noticed some weird behavior of pango as ligatures kick in. I guess it's something deeply rooted in pango that currently I cannot fix. This shall require a deeper look into pango in the future...

text = MarkupText("fast", t2c={"f": TEAL}, font="Source Han Sans Light")  # works
text = MarkupText("fire", t2c={"f": TEAL}, font="Source Han Sans Light")  # fails
text = MarkupText("<span font_features='liga=0'>fire</span>", t2c={"f": TEAL}, font="Source Han Sans Light")  # works

[YishiMichael]

The original svg is used instead of the labelled one, so whitespaces around operators won't be eliminated if isolated. That means, MTex("a\\,=\\,b", isolate="=") can now simply be written as MTex("a=b", isolate="=").

[3b1b]

Thank you, this is definitely an improved structure and syntax.