3b1b-manim/manimlib/mobject/svg/mtex_mobject.py
2022-01-29 21:06:54 +08:00

653 lines
24 KiB
Python

import itertools as it
import re
import sys
from types import MethodType
from manimlib.constants import BLACK
from manimlib.mobject.svg.svg_mobject import SVGMobject
from manimlib.mobject.types.vectorized_mobject import VMobject
from manimlib.mobject.types.vectorized_mobject import VGroup
from manimlib.utils.color import color_to_int_rgb
from manimlib.utils.iterables import adjacent_pairs
from manimlib.utils.iterables import remove_list_redundancies
from manimlib.utils.tex_file_writing import tex_to_svg_file
from manimlib.utils.tex_file_writing import get_tex_config
from manimlib.utils.tex_file_writing import display_during_execution
from manimlib.logger import log
SCALE_FACTOR_PER_FONT_POINT = 0.001
TEX_HASH_TO_MOB_MAP = {}
def _contains(span_0, span_1):
return span_0[0] <= span_1[0] and span_1[1] <= span_0[1]
def _get_neighbouring_pairs(iterable):
return list(adjacent_pairs(iterable))[:-1]
class _TexSVG(SVGMobject):
CONFIG = {
"color": BLACK,
"height": None,
"path_string_config": {
"should_subdivide_sharp_curves": True,
"should_remove_null_curves": True,
},
}
@staticmethod
def color_to_label(fill_color):
r, g, b = color_to_int_rgb(fill_color)
rg = r * 256 + g
return rg * 256 + b
def parse_labels(self):
for glyph in self:
glyph.glyph_label = _TexSVG.color_to_label(glyph.fill_color)
return self
class _TexSpan(object):
def __init__(self, script_type, label):
# `script_type`: 0 for normal, 1 for subscript, 2 for superscript.
# Only those spans with `script_type == 0` will be colored.
self.script_type = script_type
self.label = label
self.containing_labels = []
def __repr__(self):
return "_TexSpan(" + ", ".join([
attrib_name + "=" + str(getattr(self, attrib_name))
for attrib_name in ["script_type", "label", "containing_labels"]
]) + ")"
class _TexParser(object):
def __init__(self, tex_string, additional_substrings):
self.tex_string = tex_string
self.tex_spans_dict = {}
self.current_label = -1
self.brace_index_pairs = self.get_brace_index_pairs()
self.existing_color_command_spans = self.get_existing_color_command_spans()
self.has_existing_color_commands = any(self.existing_color_command_spans.values())
self.specified_substring_spans = []
self.add_tex_span((0, len(tex_string)))
self.break_up_by_double_braces()
self.break_up_by_scripts()
self.break_up_by_additional_substrings(additional_substrings)
self.specified_substrings = remove_list_redundancies([
tex_string[slice(*span_tuple)]
for span_tuple in self.specified_substring_spans
])
self.check_if_overlap()
self.analyse_containing_labels()
@staticmethod
def label_to_color_tuple(rgb):
# Get a unique color different from black.
rg, b = divmod(rgb, 256)
r, g = divmod(rg, 256)
return r, g, b
def add_tex_span(self, span_tuple, script_type=0, label=-1):
if span_tuple in self.tex_spans_dict:
return
if script_type == 0:
# Should be additionally labelled.
self.current_label += 1
label = self.current_label
tex_span = _TexSpan(script_type, label)
self.tex_spans_dict[span_tuple] = tex_span
def get_brace_index_pairs(self):
result = []
left_brace_indices = []
for match_obj in re.finditer(r"(\\*)(\{|\})", self.tex_string):
# Braces following even numbers of backslashes are counted.
if len(match_obj.group(1)) % 2 == 1:
continue
if match_obj.group(2) == "{":
left_brace_index = match_obj.span(2)[0]
left_brace_indices.append(left_brace_index)
else:
left_brace_index = left_brace_indices.pop()
right_brace_index = match_obj.span(2)[1]
result.append((left_brace_index, right_brace_index))
if left_brace_indices:
self.raise_tex_parsing_error("unmatched braces")
return result
def get_existing_color_command_spans(self):
tex_string = self.tex_string
color_related_commands_dict = _TexParser.get_color_related_commands_dict()
commands = color_related_commands_dict.keys()
result = {
command_name: []
for command_name in commands
}
brace_index_pairs = self.brace_index_pairs
pattern = "|".join([
re.escape(command_name)
for command_name in commands
])
for match_obj in re.finditer(pattern, tex_string):
span_tuple = match_obj.span()
command_begin_index = span_tuple[0]
command_name = match_obj.group()
n_braces = color_related_commands_dict[command_name]
for _ in range(n_braces):
span_tuple = min(filter(
lambda t: t[0] >= span_tuple[1],
brace_index_pairs
))
result[command_name].append(
(command_begin_index, span_tuple[1])
)
return result
def break_up_by_double_braces(self):
# Match paired double braces (`{{...}}`).
skip_pair = False
for prev_span_tuple, span_tuple in _get_neighbouring_pairs(
self.brace_index_pairs
):
if skip_pair:
skip_pair = False
continue
if all([
span_tuple[0] == prev_span_tuple[0] - 1,
span_tuple[1] == prev_span_tuple[1] + 1
]):
self.add_tex_span(span_tuple)
self.specified_substring_spans.append(span_tuple)
skip_pair = True
def break_up_by_scripts(self):
# Match subscripts & superscripts.
tex_string = self.tex_string
brace_indices_dict = dict(self.brace_index_pairs)
for match_obj in re.finditer(r"((?<!\\)(_|\^)\s*)|(\s+(_|\^)\s*)", tex_string):
script_type = 1 if "_" in match_obj.group() else 2
token_begin, token_end = match_obj.span()
if token_end in brace_indices_dict:
content_span = (token_end, brace_indices_dict[token_end])
else:
content_match_obj = re.match(r"\w|\\[a-zA-Z]+", tex_string[token_end:])
if not content_match_obj:
self.raise_tex_parsing_error("unclear subscript/superscript")
content_span = tuple([
index + token_end for index in content_match_obj.span()
])
self.add_tex_span(content_span)
label = self.tex_spans_dict[content_span].label
self.add_tex_span(
(token_begin, content_span[1]),
script_type=script_type,
label=label
)
def break_up_by_additional_substrings(self, additional_substrings):
tex_string = self.tex_string
all_span_tuples = []
for string in additional_substrings:
# Only match non-crossing strings.
for match_obj in re.finditer(re.escape(string), tex_string):
all_span_tuples.append(match_obj.span())
script_spans_dict = dict([
span_tuple[::-1]
for span_tuple, tex_span in self.tex_spans_dict.items()
if tex_span.script_type != 0
])
for span_begin, span_end in all_span_tuples:
if span_end in script_spans_dict.values():
# Deconstruct spans with subscripts & superscripts.
while span_end in script_spans_dict:
span_end = script_spans_dict[span_end]
if span_begin >= span_end:
continue
span_tuple = (span_begin, span_end)
self.add_tex_span(span_tuple)
self.specified_substring_spans.append(span_tuple)
def check_if_overlap(self):
span_tuples = sorted(
self.tex_spans_dict.keys(),
key=lambda t: (t[0], -t[1])
)
overlapping_span_pairs = []
for i, span_0 in enumerate(span_tuples):
for span_1 in span_tuples[i + 1 :]:
if span_0[1] <= span_1[0]:
continue
if span_0[1] < span_1[1]:
overlapping_span_pairs.append((span_0, span_1))
if overlapping_span_pairs:
tex_string = self.tex_string
log.error("Overlapping substring pairs occur in MTex:")
for span_tuple_pair in overlapping_span_pairs:
log.error(", ".join(
f"\"{tex_string[slice(*span_tuple)]}\""
for span_tuple in span_tuple_pair
))
raise ValueError
def analyse_containing_labels(self):
for span_0, tex_span_0 in self.tex_spans_dict.items():
if tex_span_0.script_type != 0:
continue
for span_1, tex_span_1 in self.tex_spans_dict.items():
if _contains(span_1, span_0):
tex_span_1.containing_labels.append(tex_span_0.label)
def get_labelled_tex_string(self):
tex_string = self.tex_string
if self.current_label == 0 and not self.has_existing_color_commands:
return tex_string
# Remove the span of extire tex string.
indices_with_labels = sorted([
(span_tuple[i], i, span_tuple[1 - i], tex_span.label)
for span_tuple, tex_span in self.tex_spans_dict.items()
if tex_span.script_type == 0
for i in range(2)
], key=lambda t: (t[0], -t[1], -t[2]))[1:]
# Prevent from "\\color[RGB]" being replaced.
# Hopefully tex string doesn't contain such a substring...
color_command_placeholder = "{{\\iffalse \\fi}}"
result = tex_string[: indices_with_labels[0][0]]
for index_with_label, next_index_with_label in _get_neighbouring_pairs(
indices_with_labels
):
index, flag, _, label = index_with_label
next_index, *_ = next_index_with_label
# Adding one more pair of braces will help maintain the glyghs of tex file...
if flag == 0:
color_tuple = _TexParser.label_to_color_tuple(label)
result += "".join([
"{{",
color_command_placeholder,
"{",
",".join(map(str, color_tuple)),
"}"
])
else:
result += "}}"
result += tex_string[index : next_index]
color_related_commands_dict = _TexParser.get_color_related_commands_dict()
for command_name, command_spans in self.existing_color_command_spans.items():
if not command_spans:
continue
n_braces = color_related_commands_dict[command_name]
command_to_replace = command_name + n_braces * "{black}"
commands = {
tex_string[slice(*span_tuple)]
for span_tuple in command_spans
}
for command in commands:
result = result.replace(command, command_to_replace)
return result.replace(color_command_placeholder, "\\color[RGB]")
def raise_tex_parsing_error(self, message):
raise ValueError(f"Failed to parse tex ({message}): \"{self.tex_string}\"")
@staticmethod
def get_color_related_commands_dict():
# Only list a few commands that are commonly used.
return {
"\\color": 1,
"\\textcolor": 1,
"\\pagecolor": 1,
"\\colorbox": 1,
"\\fcolorbox": 2,
}
class MTex(VMobject):
CONFIG = {
"fill_opacity": 1.0,
"stroke_width": 0,
"font_size": 48,
"alignment": "\\centering",
"tex_environment": "align*",
"isolate": [],
"tex_to_color_map": {},
"use_plain_tex_file": False,
}
def __init__(self, tex_string, **kwargs):
super().__init__(**kwargs)
tex_string = tex_string.strip()
# Prevent from passing an empty string.
if not tex_string:
tex_string = "\\quad"
self.tex_string = tex_string
self.generate_mobject()
self.set_color_by_tex_to_color_map(self.tex_to_color_map)
self.scale(SCALE_FACTOR_PER_FONT_POINT * self.font_size)
def get_additional_substrings_to_break_up(self):
result = remove_list_redundancies([
*self.tex_to_color_map.keys(), *self.isolate
])
if "" in result:
result.remove("")
return result
def get_parser(self):
return _TexParser(self.tex_string, self.get_additional_substrings_to_break_up())
def generate_mobject(self):
tex_string = self.tex_string
tex_parser = self.get_parser()
self.tex_spans_dict = tex_parser.tex_spans_dict
self.specified_substrings = tex_parser.specified_substrings
fill_color = self.get_fill_color()
# Cannot simultaneously be false, so at least one file is generated.
require_labelled_tex_file = tex_parser.current_label != 0
require_plain_tex_file = any([
self.use_plain_tex_file,
tex_parser.has_existing_color_commands,
tex_parser.current_label == 0
])
if require_labelled_tex_file:
labelled_full_tex = self.get_tex_file_body(tex_parser.get_labelled_tex_string())
labelled_hash_val = hash(labelled_full_tex)
if labelled_hash_val in TEX_HASH_TO_MOB_MAP:
self.add(*TEX_HASH_TO_MOB_MAP[labelled_hash_val].copy())
else:
with display_during_execution(f"Writing \"{tex_string}\""):
filename = tex_to_svg_file(labelled_full_tex)
labelled_svg_glyphs = _TexSVG(filename).parse_labels()
self.add(*labelled_svg_glyphs)
self.build_submobjects()
TEX_HASH_TO_MOB_MAP[labelled_hash_val] = self.copy()
if not require_plain_tex_file:
self.set_fill(color=fill_color)
return self
# require_plain_tex_file == True
self.set_submobjects([])
full_tex = self.get_tex_file_body(tex_string, fill_color=fill_color)
hash_val = hash(full_tex)
if hash_val in TEX_HASH_TO_MOB_MAP:
self.add(*TEX_HASH_TO_MOB_MAP[hash_val].copy())
else:
with display_during_execution(f"Writing \"{tex_string}\""):
filename = tex_to_svg_file(full_tex)
svg_glyphs = _TexSVG(filename)
if require_labelled_tex_file:
labelled_svg_mob = TEX_HASH_TO_MOB_MAP[labelled_hash_val]
for glyph, labelled_glyph in zip(svg_glyphs, it.chain(*labelled_svg_mob)):
glyph.glyph_label = labelled_glyph.glyph_label
else:
for glyph in svg_glyphs:
glyph.glyph_label = 0
self.add(*svg_glyphs)
self.build_submobjects()
TEX_HASH_TO_MOB_MAP[hash_val] = self.copy()
return self
def get_tex_file_body(self, new_tex, fill_color=None):
if self.tex_environment:
new_tex = "\n".join([
f"\\begin{{{self.tex_environment}}}",
new_tex,
f"\\end{{{self.tex_environment}}}"
])
if self.alignment:
new_tex = "\n".join([self.alignment, new_tex])
if fill_color:
int_rgb = color_to_int_rgb(fill_color)
color_command = "".join([
"\\color[RGB]",
"{",
",".join(map(str, int_rgb)),
"}"
])
new_tex = "\n".join(
[color_command, new_tex]
)
tex_config = get_tex_config()
return tex_config["tex_body"].replace(
tex_config["text_to_replace"],
new_tex
)
def build_submobjects(self):
if not self.submobjects:
return
self.init_colors()
for glyph in self.submobjects:
glyph.set_fill(glyph.fill_color)
self.group_submobjects()
self.sort_scripts_in_tex_order()
self.assign_submob_tex_strings()
def group_submobjects(self):
# Simply pack together adjacent mobjects with the same label.
new_submobjects = []
def append_new_submobject(glyphs):
if glyphs:
submobject = VGroup(*glyphs)
submobject.submob_label = glyphs[0].glyph_label
new_submobjects.append(submobject)
new_glyphs = []
current_glyph_label = 0
for glyph in self.submobjects:
if glyph.glyph_label == current_glyph_label:
new_glyphs.append(glyph)
else:
append_new_submobject(new_glyphs)
new_glyphs = [glyph]
current_glyph_label = glyph.glyph_label
append_new_submobject(new_glyphs)
self.set_submobjects(new_submobjects)
def sort_scripts_in_tex_order(self):
# LaTeX always puts superscripts before subscripts.
# This function sorts the submobjects of scripts in the order of tex given.
tex_spans_dict = self.tex_spans_dict
index_and_span_list = sorted([
(index, span_tuple)
for span_tuple, tex_span in tex_spans_dict.items()
if tex_span.script_type != 0
for index in span_tuple
])
switch_range_pairs = []
for index_and_span_0, index_and_span_1 in _get_neighbouring_pairs(
index_and_span_list
):
index_0, span_tuple_0 = index_and_span_0
index_1, span_tuple_1 = index_and_span_1
if index_0 != index_1:
continue
if not all([
tex_spans_dict[span_tuple_0].script_type == 1,
tex_spans_dict[span_tuple_1].script_type == 2
]):
continue
submob_range_0 = self.range_of_part(
self.get_part_by_span_tuples([span_tuple_0])
)
submob_range_1 = self.range_of_part(
self.get_part_by_span_tuples([span_tuple_1])
)
switch_range_pairs.append((submob_range_0, submob_range_1))
switch_range_pairs.sort(key=lambda t: (t[0].stop, -t[0].start))
indices = list(range(len(self.submobjects)))
for submob_range_0, submob_range_1 in switch_range_pairs:
indices = [
*indices[: submob_range_1.start],
*indices[submob_range_0.start : submob_range_0.stop],
*indices[submob_range_1.stop : submob_range_0.start],
*indices[submob_range_1.start : submob_range_1.stop],
*indices[submob_range_0.stop :]
]
submobs = self.submobjects
self.set_submobjects([submobs[i] for i in indices])
def assign_submob_tex_strings(self):
# Not sure whether this is the best practice...
# This temporarily supports `TransformMatchingTex`.
tex_string = self.tex_string
tex_spans_dict = self.tex_spans_dict
# Use tex strings including "_", "^".
label_dict = {}
for span_tuple, tex_span in tex_spans_dict.items():
if tex_span.script_type != 0:
label_dict[tex_span.label] = span_tuple
else:
if tex_span.label not in label_dict:
label_dict[tex_span.label] = span_tuple
curr_labels = [submob.submob_label for submob in self.submobjects]
prev_labels = [curr_labels[-1], *curr_labels[:-1]]
next_labels = [*curr_labels[1:], curr_labels[0]]
tex_string_spans = []
for curr_label, prev_label, next_label in zip(
curr_labels, prev_labels, next_labels
):
curr_span_tuple = label_dict[curr_label]
prev_span_tuple = label_dict[prev_label]
next_span_tuple = label_dict[next_label]
containing_labels = tex_spans_dict[curr_span_tuple].containing_labels
tex_string_spans.append([
prev_span_tuple[1] if prev_label in containing_labels else curr_span_tuple[0],
next_span_tuple[0] if next_label in containing_labels else curr_span_tuple[1]
])
tex_string_spans[0][0] = label_dict[curr_labels[0]][0]
tex_string_spans[-1][1] = label_dict[curr_labels[-1]][1]
for submob, tex_string_span in zip(self.submobjects, tex_string_spans):
submob.tex_string = tex_string[slice(*tex_string_span)]
# Support `get_tex()` method here.
submob.get_tex = MethodType(lambda inst: inst.tex_string, submob)
def get_part_by_span_tuples(self, span_tuples):
tex_spans_dict = self.tex_spans_dict
labels = set(it.chain(*[
tex_spans_dict[span_tuple].containing_labels
for span_tuple in span_tuples
]))
return VGroup(*filter(
lambda submob: submob.submob_label in labels,
self.submobjects
))
def find_span_components_of_custom_span(self, custom_span_tuple):
tex_string = self.tex_string
span_choices = sorted(filter(
lambda t: _contains(custom_span_tuple, t),
self.tex_spans_dict.keys()
))
# Filter out spans that reach the farthest.
span_choices_dict = dict(span_choices)
span_begin, span_end = custom_span_tuple
result = []
while span_begin != span_end:
if span_begin not in span_choices_dict:
if tex_string[span_begin].strip():
return None
# Whitespaces may occur between spans.
span_begin += 1
continue
next_begin = span_choices_dict[span_begin]
result.append((span_begin, next_begin))
span_begin = next_begin
return result
def get_part_by_custom_span_tuple(self, custom_span_tuple):
span_tuples = self.find_span_components_of_custom_span(custom_span_tuple)
if span_tuples is None:
tex = self.tex_string[slice(*custom_span_tuple)]
raise ValueError(f"Failed to get span of tex: \"{tex}\"")
return self.get_part_by_span_tuples(span_tuples)
def get_parts_by_tex(self, tex):
return VGroup(*[
self.get_part_by_custom_span_tuple(match_obj.span())
for match_obj in re.finditer(re.escape(tex), self.tex_string)
])
def get_part_by_tex(self, tex, index=0):
all_parts = self.get_parts_by_tex(tex)
return all_parts[index]
def set_color_by_tex(self, tex, color):
self.get_parts_by_tex(tex).set_color(color)
return self
def set_color_by_tex_to_color_map(self, tex_to_color_map):
for tex, color in list(tex_to_color_map.items()):
try:
self.set_color_by_tex(tex, color)
except:
pass
return self
def indices_of_part(self, part):
indices = [
i for i, submob in enumerate(self.submobjects)
if submob in part
]
if not indices:
raise ValueError("Failed to find part in tex")
return indices
def indices_of_part_by_tex(self, tex, index=0):
part = self.get_part_by_tex(tex, index=index)
return self.indices_of_part(part)
def range_of_part(self, part):
indices = self.indices_of_part(part)
return range(indices[0], indices[-1] + 1)
def range_of_part_by_tex(self, tex, index=0):
part = self.get_part_by_tex(tex, index=index)
return self.range_of_part(part)
def index_of_part(self, part):
return self.indices_of_part(part)[0]
def index_of_part_by_tex(self, tex, index=0):
part = self.get_part_by_tex(tex, index=index)
return self.index_of_part(part)
def get_tex(self):
return self.tex_string
def get_all_isolated_substrings(self):
tex_string = self.tex_string
return remove_list_redundancies([
tex_string[slice(*span_tuple)]
for span_tuple in self.tex_spans_dict.keys()
])
def get_specified_substrings(self):
return self.specified_substrings
class MTexText(MTex):
CONFIG = {
"tex_environment": None,
}