All Downloads are FREE. Search and download functionalities are using the official Maven repository.

docutils.utils.math.latex2mathml.py Maven / Gradle / Ivy

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# :Id: $Id: latex2mathml.py 8878 2021-11-05 11:10:44Z milde $
# :Copyright: © 2005 Jens Jørgen Mortensen [1]_
#             © 2010, 2021 Günter Milde.
#
# :License: Released under the terms of the `2-Clause BSD license`_, in short:
#
#    Copying and distribution of this file, with or without modification,
#    are permitted in any medium without royalty provided the copyright
#    notice and this notice are preserved.
#    This file is offered as-is, without any warranty.
#
# .. _2-Clause BSD license: https://opensource.org/licenses/BSD-2-Clause
#
# .. [1] the original `rst2mathml.py` in `sandbox/jensj/latex_math`

"""Convert LaTex maths code into presentational MathML.

This module is provisional:
the API is not settled and may change with any minor Docutils version.
"""

# Usage:
#
# >>> from latex2mathml import *

import collections
import copy
import re
import sys
import unicodedata
if sys.version_info >= (3, 0):
    unicode = str  # noqa

from docutils.utils.math import tex2unichar, toplevel_code


# Character data
# --------------

# LaTeX math macro to Unicode mappings.
# Character categories.

# identifiers -> 

letters = tex2unichar.mathalpha
letters['hbar'] = u'\u210F' # compatibility mapping to ℏ (\hslash).
# (ħ LATIN SMALL LETTER H WITH STROKE is upright)

# special case: Capital Greek letters: (upright in TeX style)
greek_capitals = {
    'Phi':u'\u03a6', 'Xi':u'\u039e', 'Sigma':u'\u03a3',
    'Psi':u'\u03a8', 'Delta':u'\u0394', 'Theta':u'\u0398',
    'Upsilon':u'\u03d2', 'Pi':u'\u03a0', 'Omega':u'\u03a9',
    'Gamma':u'\u0393', 'Lambda':u'\u039b'}

# functions -> 
functions = {# functions with a space in the name
             'liminf': u'lim\u202finf',
             'limsup': u'lim\u202fsup',
             'injlim': u'inj\u202flim',
             'projlim': u'proj\u202flim',
             # embellished function names (see handle_cmd() below)
             'varlimsup': 'lim',
             'varliminf': 'lim',
             'varprojlim': 'lim',
             'varinjlim': 'lim',
             # custom function name
             'operatorname': None,
            }
functions.update((name, name) for name in
                 ('arccos', 'arcsin', 'arctan', 'arg',  'cos',
                  'cosh',   'cot',    'coth',   'csc',  'deg',
                  'det',    'dim',    'exp',    'gcd',  'hom',
                  'ker',    'lg',     'ln',     'log',  'Pr',
                  'sec',    'sin',    'sinh',   'tan',  'tanh'))
# Function with limits: 'lim', 'sup', 'inf', 'max', 'min':
# use  to allow "movablelimits" attribute (see below).


# math font selection ->  or 
math_alphabets = {# 'cmdname': 'mathvariant value'  # package
                  'boldsymbol': 'bold',
                  'mathbf':     'bold',
                  'mathit':     'italic',
                  'mathtt':     'monospace',
                  'mathrm':     'normal',
                  'mathsf':     'sans-serif',
                  'mathcal':    'script',
                  'mathbfit':   'bold-italic',            # isomath
                  'mathbb':     'double-struck',          # amssymb
                  'mathfrak':   'fraktur',                # amssymb
                  'mathsfit':   'sans-serif-italic',      # isomath
                  'mathsfbfit': 'sans-serif-bold-italic', # isomath
                  'mathscr':    'script',                 # mathrsfs
                  # unsupported: bold-fraktur
                  #              bold-script
                  #              bold-sans-serif
                 }

# operator, fence, or separator -> 


stretchables = {# extensible delimiters allowed in left/right cmds
                'backslash':   '\\',
                'uparrow':     u'\u2191', # ↑ UPWARDS ARROW
                'downarrow':   u'\u2193', # ↓ DOWNWARDS ARROW
                'updownarrow': u'\u2195', # ↕ UP DOWN ARROW
                'Uparrow':     u'\u21d1', # ⇑ UPWARDS DOUBLE ARROW
                'Downarrow':   u'\u21d3', # ⇓ DOWNWARDS DOUBLE ARROW
                'Updownarrow': u'\u21d5', # ⇕ UP DOWN DOUBLE ARROW
                'lmoustache':  u'\u23b0', # ⎰ UPPER LEFT OR LOWER RIGHT CURLY BRACKET SECTION
                'rmoustache':  u'\u23b1', # ⎱ UPPER RIGHT OR LOWER LEFT CURLY BRACKET SECTION
                'arrowvert':   u'\u23d0', # ⏐ VERTICAL LINE EXTENSION
                'bracevert':   u'\u23aa', # ⎪ CURLY BRACKET EXTENSION
                'lvert':      u'|',      # left  |
                'lVert':      u'\u2016', # left  ‖
                'rvert':      u'|',      # right |
                'rVert':      u'\u2016', # right ‖
                'Arrowvert':  u'\u2016', # ‖
               }
stretchables.update(tex2unichar.mathfence)
stretchables.update(tex2unichar.mathopen)  # Braces
stretchables.update(tex2unichar.mathclose) # Braces

# >>> print(' '.join(sorted(set(stretchables.values()))))
# [ \ ] { | } ‖ ↑ ↓ ↕ ⇑ ⇓ ⇕ ⌈ ⌉ ⌊ ⌋ ⌜ ⌝ ⌞ ⌟ ⎪ ⎰ ⎱ ⏐ ⟅ ⟆ ⟦ ⟧ ⟨ ⟩ ⟮ ⟯ ⦇ ⦈

operators = {# negated symbols without pre-composed Unicode character
             'nleqq':      u'\u2266\u0338', # ≦̸
             'ngeqq':      u'\u2267\u0338', # ≧̸
             'nleqslant':  u'\u2a7d\u0338', # ⩽̸
             'ngeqslant':  u'\u2a7e\u0338', # ⩾̸
             'ngtrless':   u'\u2277\u0338', # txfonts
             'nlessgtr':   u'\u2276\u0338', # txfonts
             'nsubseteqq': u'\u2AC5\u0338', # ⫅̸
             'nsupseteqq': u'\u2AC6\u0338', # ⫆̸
             # compatibility definitions:
             'centerdot': u'\u2B1D', # BLACK VERY SMALL SQUARE | mathbin
             'varnothing': u'\u2300', # ⌀ DIAMETER SIGN | empty set
             'varpropto': u'\u221d', # ∝ PROPORTIONAL TO | sans serif
             'triangle': u'\u25B3', # WHITE UP-POINTING TRIANGLE | mathord
             'triangledown': u'\u25BD', # WHITE DOWN-POINTING TRIANGLE | mathord
             # alias commands:
             'dotsb': u'\u22ef', # ⋯ with binary operators/relations
             'dotsc': u'\u2026', # … with commas
             'dotsi': u'\u22ef', # ⋯ with integrals
             'dotsm': u'\u22ef', # ⋯ multiplication dots
             'dotso': u'\u2026', # … other dots
             # functions with movable limits (requires )
             'lim': 'lim',
             'sup': 'sup',
             'inf': 'inf',
             'max': 'max',
             'min': 'min',
            }
operators.update(tex2unichar.mathbin)   # Binary symbols
operators.update(tex2unichar.mathrel)   # Relation symbols, arrow symbols
operators.update(tex2unichar.mathord)   # Miscellaneous symbols
operators.update(tex2unichar.mathpunct) # Punctuation
operators.update(tex2unichar.mathop)    # Variable-sized symbols
operators.update(stretchables)


# special cases

thick_operators = {# style='font-weight: bold;'
                   'thicksim':   u'\u223C', # ∼
                   'thickapprox':u'\u2248', # ≈
                  }

small_operators = {# mathsize='75%'
                   'shortmid':       u'\u2223', # ∣
                   'shortparallel':  u'\u2225', # ∥
                   'nshortmid':      u'\u2224', # ∤
                   'nshortparallel': u'\u2226', # ∦
                   'smallfrown':     u'\u2322', # ⌢ FROWN
                   'smallsmile':     u'\u2323', # ⌣ SMILE
                   'smallint':       u'\u222b', # ∫ INTEGRAL
                  }

# Operators and functions with limits above/below in display formulas
# and in index position inline (movablelimits=True)
movablelimits = ('bigcap', 'bigcup', 'bigodot', 'bigoplus', 'bigotimes',
                  'bigsqcup', 'biguplus', 'bigvee', 'bigwedge',
                  'coprod', 'intop', 'ointop', 'prod', 'sum',
                  'lim', 'max', 'min', 'sup', 'inf')
# Depending on settings, integrals may also be in this category.
# (e.g. if "amsmath" is loaded with option "intlimits", see
#  http://mirror.ctan.org/macros/latex/required/amsmath/amsldoc.pdf)
# movablelimits.extend(('fint', 'iiiint', 'iiint', 'iint', 'int', 'oiint',
#                       'oint', 'ointctrclockwise', 'sqint',
#                       'varointclockwise',))

# horizontal space -> 

spaces = {'qquad':         '2em',       # two \quad
          'quad':          '1em',       # 18 mu
          'thickspace':    '0.2778em',  # 5mu = 5/18em
          'medspace':      '0.2222em',  # 4mu = 2/9em
          'thinspace':     '0.1667em',  # 3mu = 1/6em
          'negthinspace':  '-0.1667em', # -3mu = -1/6em
          'negmedspace':   '-0.2222em', # -4mu = -2/9em
          'negthickspace': '-0.2778em', # -5mu = -5/18em
          ' ':             '0.25em',    # inter word space
          ';':             '0.2778em',  # 5mu thickspace
          ':':             '0.2222em',  # 4mu medspace
          ',':             '0.1667em',  # 3mu thinspace
          '!':             '-0.1667em', # negthinspace
         }

# accents -> 
accents = {# TeX:      (spacing, combining)
           'acute':    (u'´', u'\u0301'),
           'bar':      (u'ˉ', u'\u0304'),
           'breve':    (u'˘', u'\u0306'),
           'check':    (u'ˇ', u'\u030C'),
           'dot':      (u'˙', u'\u0307'),
           'ddot':     (u'¨', u'\u0308'),
           'dddot':    (u'⋯', u'\u20DB'),
           'grave':    (u'`', u'\u0300'),
           'hat':      (u'ˆ', u'\u0302'),
           'mathring': (u'˚', u'\u030A'),
           'tilde':    (u'˜', u'\u0303'), # tilde ~ or small tilde ˜?
           'vec':      (u'→', u'\u20d7'), # → too heavy, accents="false"
           # TODO: ddddot
       }

# limits etc. ->  or 
over = {# TeX:                  (char,     offset-correction/em)
        'overbrace':            (u'\u23DE', -0.2), # DejaVu Math -0.6
        'overleftarrow':        (u'\u2190', -0.2),
        'overleftrightarrow':   (u'\u2194', -0.2),
        'overline':             (u'_',      -0.2),   # \u2012' FIGURE DASH does not stretch
        'overrightarrow':       (u'\u2192', -0.2),
        'widehat':              (u'^',      -0.5),
        'widetilde':            (u'~',      -0.3),
       }
under = {'underbrace':          (u'\u23DF',  0.1), # DejaVu Math -0.7
         'underleftarrow':      (u'\u2190', -0.2),
         'underleftrightarrow': (u'\u2194', -0.2),
         'underline':           (u'_',      -0.8),
         'underrightarrow':     (u'\u2192', -0.2),
        }

# Character translations
# ----------------------
# characters with preferred alternative in mathematical use
# cf. https://www.w3.org/TR/MathML3/chapter7.html#chars.anomalous
anomalous_chars = {'-': u'\u2212', # HYPHEN-MINUS -> MINUS SIGN
                   ':': u'\u2236', # COLON -> RATIO
                   '~': u'\u00a0', # NO-BREAK SPACE
                  }

# blackboard bold (Greek characters not working with "mathvariant" (Firefox 78)
mathbb = {u'Γ': u'\u213E',    # ℾ
          u'Π': u'\u213F',    # ℿ
          u'Σ': u'\u2140',    # ⅀
          u'γ': u'\u213D',    # ℽ
          u'π': u'\u213C',    # ℼ
         }

# Matrix environments
matrices = {# name:    fences
            'matrix':  ('', ''),
            'smallmatrix':  ('', ''), # smaller, see begin_environment()!
            'pmatrix': ('(', ')'),
            'bmatrix': ('[', ']'),
            'Bmatrix': ('{', '}'),
            'vmatrix': ('|', '|'),
            'Vmatrix': (u'\u2016', u'\u2016'), # ‖
            'cases':   ('{', ''),
           }

layout_styles = {
    'displaystyle':      {'displaystyle': True,  'scriptlevel': 0},
    'textstyle':         {'displaystyle': False, 'scriptlevel': 0},
    'scriptstyle':       {'displaystyle': False, 'scriptlevel': 1},
    'scriptscriptstyle': {'displaystyle': False, 'scriptlevel': 2},
    }
# See also https://www.w3.org/TR/MathML3/chapter3.html#presm.scriptlevel

fractions = {# name:   style_attrs, frac_attrs
             'frac':   ({}, {}),
             'cfrac':  ({'displaystyle': True,  'scriptlevel': 0,
                         'CLASS': 'cfrac'}, {}), # in LaTeX with padding
             'dfrac':  (layout_styles['displaystyle'], {}),
             'tfrac':  (layout_styles['textstyle'], {}),
             'binom':  ({}, {'linethickness': 0}),
             'dbinom': (layout_styles['displaystyle'], {'linethickness': 0}),
             'tbinom': (layout_styles['textstyle'], {'linethickness': 0}),
            }

delimiter_sizes = ['', '1.2em', '1.623em', '2.047em', '2.470em']
bigdelimiters = {'left':  0,
                 'right': 0,
                 'bigl':  1,
                 'bigr':  1,
                 'Bigl':  2,
                 'Bigr':  2,
                 'biggl': 3,
                 'biggr': 3,
                 'Biggl': 4,
                 'Biggr': 4,
                }


# MathML element classes
# ----------------------

class math(object):
    """Base class for MathML elements and root of MathML trees."""

    nchildren = None
    """Expected number of children or None"""
    # cf. https://www.w3.org/TR/MathML3/chapter3.html#id.3.1.3.2
    parent = None
    """Parent node in MathML DOM tree."""
    _level = 0 # indentation level (static class variable)
    xml_entities = { # for invalid and invisible characters
                    ord('<'): u'<',
                    ord('>'): u'>',
                    ord('&'): u'&',
                    0x2061:   u'⁡',
                   }
    _boolstrings = {True: 'true', False: 'false'}
    """String representation of boolean MathML attribute values."""

    html_tagname = 'span'
    """Tag name for HTML representation."""

    def __init__(self, *children, **attributes):
        """Set up node with `children` and `attributes`.

        Attributes are downcased: Use CLASS to set "class" value.
        >>> math(mn(3), CLASS='test')
        math(mn(3), class='test')
        >>> math(CLASS='test').toprettyxml()
        '\n'

        """
        self.children = []
        self.extend(children)

        self.attributes = collections.OrderedDict()
        # sort attributes for predictable functional tests
        # as self.attributes.update(attributes) does not keep order in Python < 3.6
        for key in sorted(attributes.keys()):
            # Use .lower() to allow argument `CLASS` for attribute `class`
            # (Python keyword). MathML uses only lowercase attributes.
            self.attributes[key.lower()] = attributes[key]

    def __repr__(self):
        content = [repr(item) for item in getattr(self, 'children', [])]
        if hasattr(self, 'data'):
            content.append(repr(self.data))
        if isinstance(self, MathSchema) and self.switch:
            content.append('switch=True')
        content += ["%s=%r"%(k, v) for k, v in self.attributes.items()
                    if v is not None]

        return self.__class__.__name__ + '(%s)' % ', '.join(content)

    def __len__(self):
        return len(self.children)

    # emulate dictionary-like access to attributes
    # see `docutils.nodes.Element` for dict/list interface
    def __getitem__(self, key):
        return self.attributes[key]
    def __setitem__(self, key, item):
        self.attributes[key] = item
    def get(self, *args, **kwargs):
        return self.attributes.get(*args, **kwargs)

    def full(self):
        """Return boolean indicating whether children may be appended."""
        return (self.nchildren is not None
                and len(self) >= self.nchildren)

    def append(self, child):
        """Append child and return self or first non-full parent.

        If self is full, go up the tree and return first non-full node or
        `None`.
        """
        if self.full():
            raise SyntaxError('Node %s already full!' % self)
        self.children.append(child)
        child.parent = self
        if self.full():
            return self.close()
        return self

    def extend(self, children):
        for child in children:
            self.append(child)
        return self

    def close(self):
        """Close element and return first non-full parent or None."""
        parent = self.parent
        while parent is not None and parent.full():
            parent = parent.parent
        return parent

    def toprettyxml(self):
        """Return XML representation of self as string."""
        return ''.join(self._xml())

    def _xml(self, level=0):
        return ([self.xml_starttag()]
                + self._xml_body(level)
                + ['' % self.__class__.__name__])

    def xml_starttag(self):
        attrs = ['%s="%s"' % (k, str(v).replace('True', 'true').replace('False', 'false'))
                 for k, v in self.attributes.items()
                 if v is not None]
        return '<%s>' % ' '.join([self.__class__.__name__] + attrs)

    def _xml_body(self, level=0):
        xml = []
        for child in self.children:
            xml.extend(['\n', '  ' * (level+1)])
            xml.extend(child._xml(level+1))
        xml.extend(['\n', '  ' * level])
        return xml

# >>> n2 = math(mn(2))
# >>> n2
# math(mn(2))
# >>> n2.toprettyxml()
# '\n  2\n'
# >>> len(n2)
# 1
# >>> eq3 = math(id='eq3', display='block')
# >>> eq3
# math(display='block', id='eq3')
# >>> eq3.toprettyxml()
# '\n'
# >>> len(eq3)
# 0
# >>> math(CLASS='bold').xml_starttag()
# ''

class mtable(math): pass

# >>> mt = mtable(displaystyle=True)
# >>> mt
# mtable(displaystyle=True)
# >>> math(mt).toprettyxml()
# '\n  \n  \n'

class mrow(math):
    """Group sub-expressions as a horizontal row."""

    def close(self):
        """Close element and return first non-full parent or None.

        Remove , if it is single child and the parent infers an mrow
        or if it has only one child element.
        """
        parent = self.parent
        if isinstance(parent, MathRowSchema) and parent.nchildren == 1:
            parent.nchildren = len(parent.children)
            parent.children = self.children
            for child in self.children:
                child.parent = parent
            return parent.close()
        if len(self) == 1:
            try:
                parent.children[parent.children.index(self)] = self.children[0]
                self.children[0].parent = parent
            except (AttributeError, ValueError):
                return self.children[0]
        return super(mrow, self).close()

# >>> mrow(displaystyle=False)
# mrow(displaystyle=False)

# The elements , , , , , ,
# , , and  treat their contents as a single inferred mrow
# formed from all their children.
class MathRowSchema(math):
    """Base class for elements treating content as a single inferred mrow."""

class mtr(MathRowSchema): pass

class mtd(MathRowSchema): pass

class menclose(MathRowSchema):
    nchildren = 1 # \boxed expects one argument or a group

class mphantom(MathRowSchema):
    nchildren = 1 # \phantom expects one argument or a group

class msqrt(MathRowSchema):
    nchildren = 1 # \sqrt expects one argument or a group

class mstyle(MathRowSchema):
    nchildren = 1 # \mathrm, ... expect one argument or a group


class MathToken(math):
    """Token Element: contains textual data instead of children.

    Base class for mo, mi, and mn.
    """
    nchildren = 0

    def __init__(self, data, **attributes):
        self.data = data
        super(MathToken, self).__init__(**attributes)

    def _xml_body(self, level=0):
        return [unicode(self.data).translate(self.xml_entities)]

class mtext(MathToken): pass
class mi(MathToken): pass
class mo(MathToken): pass
class mn(MathToken): pass

# >>> mo(u'<')
# mo('<')
# >>> mo(u'<')._xml()
# ['', '<', '']

class MathSchema(math):
    """Base class for schemata expecting 2 or more children.

    The special attribute `switch` indicates that the last two child
    elements are in reversed order and must be switched before XML-export.
    """

    nchildren = 2

    def __init__(self, *children, **kwargs):
        self.switch = kwargs.pop('switch', False)
        math.__init__(self, *children, **kwargs)

    def append(self, child):
        current_node = super(MathSchema, self).append(child)
        # normalize order if full
        if self.switch and self.full():
            self.children[-1], self.children[-2] = self.children[-2], self.children[-1]
            self.switch = False
        return current_node

class msub(MathSchema): pass
class msup(MathSchema): pass
class msubsup(MathSchema):
    nchildren = 3

# >>> msub(mi('x'), mo('-'))
# msub(mi('x'), mo('-'))
# >>> msubsup(mi('base'), mi('sub'), mi('super'))
# msubsup(mi('base'), mi('sub'), mi('super'))
# >>> msubsup(mi('base'), mi('super'), mi('sub'), switch=True)
# msubsup(mi('base'), mi('sub'), mi('super'))

class munder(msub): pass
class mover(msup): pass

# >>> munder(mi('lim'), mo('-'), accent=False)
# munder(mi('lim'), mo('-'), accent=False)
# >>> mu = munder(mo('-'), accent=False, switch=True)
# >>> mu
# munder(mo('-'), switch=True, accent=False)
# >>> mu.append(mi('lim'))
# >>> mu
# munder(mi('lim'), mo('-'), accent=False)
# >>> mu.append(mi('lim'))
# Traceback (most recent call last):
# SyntaxError: Node munder(mi('lim'), mo('-'), accent=False) already full!
# >>> munder(mo('-'), mi('lim'), accent=False, switch=True).toprettyxml()
# '\n  lim\n  -\n'

class munderover(msubsup): pass

class mroot(MathSchema):
    nchildren = 2

class mfrac(math):
    nchildren = 2

class mspace(math):
    nchildren = 0


# LaTeX to MathML translation
# ---------------------------

# auxiliary functions
# ~~~~~~~~~~~~~~~~~~~

def tex_cmdname(string):
    """Return leading TeX command name and remainder of `string`.

    >>> tex_cmdname('mymacro2') # up to first non-letter
    ('mymacro', '2')
    >>> tex_cmdname('name 2') # strip trailing whitespace
    ('name', '2')
    >>> tex_cmdname('_2') # single non-letter character
    ('_', '2')

    """
    m = re.match(r'([a-zA-Z]+) *(.*)', string)
    if m is None:
        m = re.match(r'(.?)(.*)', string)
    return m.group(1), m.group(2)

# Test:
#
# >>> tex_cmdname('name_2') # first non-letter terminates
# ('name', '_2')
# >>> tex_cmdname(' next') # leading whitespace is returned
# (' ', 'next')
# >>> tex_cmdname('1 2') # whitespace after non-letter is kept
# ('1', ' 2')
# >>> tex_cmdname('') # empty string
# ('', '')

def tex_number(string):
    """Return leading number literal and remainder of `string`.

    >>> tex_number('123.4')
    ('123.4', '')

    """
    m = re.match(r'([0-9.,]*[0-9]+)(.*)', string)
    if m is None:
        return '', string
    return m.group(1), m.group(2)

# Test:
#
# >>> tex_number(' 23.4b') # leading whitespace -> no number
# ('', ' 23.4b')
# >>> tex_number('23,400/2') # comma separator included
# ('23,400', '/2')
# >>> tex_number('23. 4/2') # trailing separator not included
# ('23', '. 4/2')
# >>> tex_number('4, 2') # trailing separator not included
# ('4', ', 2')
# >>> tex_number('1 000.4')
# ('1', ' 000.4')

def tex_token(string):
    """Return first simple TeX token and remainder of `string`.

    >>> tex_token('\\command{without argument}')
    ('\\command', '{without argument}')
    >>> tex_token('or first character')
    ('o', 'r first character')

    """
    m = re.match(r"""((?P\\[a-zA-Z]+)\s* # TeX command, skip whitespace
                      |(?P\\.)          # one-character TeX command
                      |(?P.?))            # first character (or empty)
                     (?P.*$)    # remaining part of string
                 """, string, re.VERBOSE)
    cmd, chcmd, ch, remainder = m.group('cmd', 'chcmd', 'ch', 'remainder')
    return cmd or chcmd or ch, remainder

# Test:
#
# >>> tex_token('{opening bracket of group}')
# ('{', 'opening bracket of group}')
# >>> tex_token('\\skip whitespace after macro name')
# ('\\skip', 'whitespace after macro name')
# >>> tex_token('. but not after single char')
# ('.', ' but not after single char')
# >>> tex_token('') # empty string.
# ('', '')
# >>> tex_token('\{escaped bracket')
# ('\\{', 'escaped bracket')

def tex_group(string):
    """Return first TeX group or token and remainder of `string`.

    >>> tex_group('{first group} returned without brackets')
    ('first group', ' returned without brackets')

    """
    split_index = 0
    nest_level = 0   # level of {{nested} groups}
    escape = False   # the next character is escaped (\)

    if not string.startswith('{'):
        # special case: there is no group, return first token and remainder
        return string[:1], string[1:]
    for c in string:
        split_index += 1
        if escape:
            escape = False
        elif c == '\\':
            escape = True
        elif c == '{':
            nest_level += 1
        elif c == '}':
            nest_level -= 1
        if nest_level == 0:
            break
    else:
        raise SyntaxError('Group without closing bracket')
    return string[1:split_index-1], string[split_index:]



# >>> tex_group('{} empty group')
# ('', ' empty group')
# >>> tex_group('{group with {nested} group} ')
# ('group with {nested} group', ' ')
# >>> tex_group('{group with {nested group}} at the end')
# ('group with {nested group}', ' at the end')
# >>> tex_group('{{group} {with {{complex }nesting}} constructs}')
# ('{group} {with {{complex }nesting}} constructs', '')
# >>> tex_group('{group with \\{escaped\\} brackets}')
# ('group with \\{escaped\\} brackets', '')
# >>> tex_group('{group followed by closing bracket}} from outer group')
# ('group followed by closing bracket', '} from outer group')
# >>> tex_group('No group? Return first character.')
# ('N', 'o group? Return first character.')
# >>> tex_group(' {also whitespace}')
# (' ', '{also whitespace}')


def tex_token_or_group(string):
    """Return first TeX group or token and remainder of `string`.

    >>> tex_token_or_group('\\command{without argument}')
    ('\\command', '{without argument}')
    >>> tex_token_or_group('first character')
    ('f', 'irst character')
    >>> tex_token_or_group(' also whitespace')
    (' ', 'also whitespace')
    >>> tex_token_or_group('{first group} keep rest')
    ('first group', ' keep rest')

    """
    arg, remainder = tex_token(string)
    if arg == '{':
        arg, remainder = tex_group(string.lstrip())
    return arg, remainder

# >>> tex_token_or_group('\{no group but left bracket')
# ('\\{', 'no group but left bracket')

def tex_optarg(string):
    """Return optional argument and remainder.

    >>> tex_optarg('[optional argument] returned without brackets')
    ('optional argument', ' returned without brackets')
    >>> tex_optarg('{empty string, if there is no optional arg}')
    ('', '{empty string, if there is no optional arg}')

    """
    m = re.match(r"""\s*                            # leading whitespace
                 \[(?P(\\]|[^\[\]]|\\])*)\] # [group] without nested groups
                 (?P.*$)
                 """, string, re.VERBOSE)
    if m is None and not string.startswith('['):
        return '', string
    try:
        return m.group('optarg'), m.group('remainder')
    except AttributeError:
        raise SyntaxError('Could not extract optional argument from %r' % string)

# Test:
# >>> tex_optarg(' [optional argument] after whitespace')
# ('optional argument', ' after whitespace')
# >>> tex_optarg('[missing right bracket')
# Traceback (most recent call last):
# SyntaxError: Could not extract optional argument from '[missing right bracket'
# >>> tex_optarg('[group with [nested group]]')
# Traceback (most recent call last):
# SyntaxError: Could not extract optional argument from '[group with [nested group]]'


def parse_latex_math(node, string):
    """Append MathML conversion of `string` to `node` and return it.

    >>> parse_latex_math(math(), r'\alpha')
    math(mi('α'))
    >>> parse_latex_math(mrow(), r'x_{n}')
    mrow(msub(mi('x'), mi('n')))

    """
    # Normalize white-space:
    string = ' '.join(string.split())
    tree = node

    while len(string) > 0:
        # Take off first character:
        c, string = string[0], string[1:]

        if c == ' ':
            continue  # whitespace is ignored in LaTeX math mode
        if c == '\\': # start of a LaTeX macro
            cmdname, string = tex_cmdname(string)
            node, string = handle_cmd(cmdname, node, string)
        elif c in "_^":
            node = handle_script_or_limit(node, c)
        elif c == '{':
            new_node = mrow()
            node.append(new_node)
            node = new_node
        elif c == '}':
            node = node.close()
        elif c == '&':
            new_node = mtd()
            node.close().append(new_node)
            node = new_node
        elif c.isalpha():
            node = node.append(mi(c))
        elif c.isdigit():
            number, string = tex_number(string)
            node = node.append(mn(c+number))
        elif c in anomalous_chars:
            # characters with a special meaning in LaTeX math mode
            # fix spacing before "unary" minus.
            attributes = {}
            if c == '-' and node.children:
                previous_node = node.children[-1]
                if (getattr(previous_node, 'data', '-') in '([='
                    or previous_node.get('class') == 'mathopen'):
                    attributes['form'] = 'prefix'
            node = node.append(mo(anomalous_chars[c], **attributes))
        elif c in "/()[]|":
            node = node.append(mo(c, stretchy=False))
        elif c in "+*=<>,.!?`';@":
            node = node.append(mo(c))
        else:
            raise SyntaxError(u'Unsupported character: "%s"' % c)
    return tree

# Test:

# >>> print(parse_latex_math(math(), ''))
# math()
# >>> parse_latex_math(math(), ' \\sqrt{ \\alpha}')
# math(msqrt(mi('α')))
# >>> parse_latex_math(math(), '23.4x')
# math(mn('23.4'), mi('x'))
# >>> parse_latex_math(math(), '\\sqrt 2 \\ne 3')
# math(msqrt(mn('2')), mo('≠'), mn('3'))
# >>> parse_latex_math(math(), '\\sqrt{2 + 3} < 3')
# math(msqrt(mn('2'), mo('+'), mn('3')), mo('<'), mn('3'))
# >>> parse_latex_math(math(), '\\sqrt[3]{2 + 3}')
# math(mroot(mrow(mn('2'), mo('+'), mn('3')), mn('3')))
# >>> parse_latex_math(math(), '\max_x') # function takes limits
# math(munder(mo('max', movablelimits=True), mi('x')))
# >>> parse_latex_math(math(), 'x^j_i') # ensure correct order: base, sub, sup
# math(msubsup(mi('x'), mi('i'), mi('j')))
# >>> parse_latex_math(math(), '\int^j_i') # ensure correct order
# math(msubsup(mo('∫'), mi('i'), mi('j')))
# >>> parse_latex_math(math(), 'x_{\\alpha}')
# math(msub(mi('x'), mi('α')))
# >>> parse_latex_math(math(), 'x_\\text{in}')
# math(msub(mi('x'), mtext('in')))

def handle_cmd(name, node, string):
    """Process LaTeX command `name` followed by `string`.

    Append result to `node`.
    If needed, parse `string` for command argument.
    Return new current node and remainder of `string`:

    >>> handle_cmd('hbar', math(), r' \frac')
    (math(mi('ℏ')), ' \\frac')
    >>> handle_cmd('hspace', math(), r'{1ex} (x)')
    (math(mspace(width='1ex')), ' (x)')

    """

    # Token elements
    # ==============

    # identifier  ->  

    if name in letters:
        new_node = mi(letters[name])
        if name in greek_capitals:
            # upright in "TeX style" but MathML sets them italic ("ISO style").
            # CSS styling does not change the font style in Firefox 78.
            # Use 'mathvariant="normal"'?
            new_node['class'] = 'capital-greek'
        node = node.append(new_node)
        return node, string

    if name in functions:
        # use  followed by invisible function applicator character
        # (see https://www.w3.org/TR/MathML3/chapter3.html#presm.mi)
        if name == 'operatorname':
            # custom function name, e.g. ``\operatorname{abs}(x)``
            # TODO: \operatorname* -> with limits
            arg, string = tex_token_or_group(string)
            new_node = mi(arg, mathvariant='normal')
        else:
            new_node = mi(functions[name])
        # embellished function names:
        if name == 'varliminf':    # \underline\lim
            new_node = munder(new_node, mo(u'_'))
        elif name == 'varlimsup':  # \overline\lim
            new_node = mover(new_node, mo(u'¯'), accent=False)
        elif name == 'varprojlim': # \underleftarrow\lim
            new_node = munder(new_node, mo(u'\u2190'))
        elif name == 'varinjlim':  # \underrightarrow\lim
            new_node = munder(new_node, mo(u'\u2192'))

        node = node.append(new_node)
        # add ApplyFunction when appropriate (not \sin^2(x), say)
        # cf. https://www.w3.org/TR/MathML3/chapter3.html#presm.mi
        if string and string[0] not in ('^', '_'):
            node = node.append(mo(u'\u2061')) # ⁡
        return node, string

    if name in math_alphabets:
        if name == 'boldsymbol':
            attributes = {'class': 'boldsymbol'}
        else:
            attributes = {'mathvariant': math_alphabets[name]}
        if name == 'mathscr':
            attributes['class'] = 'mathscr'
        # Check for single symbol (letter, name, or ⅀)
        arg, remainder = tex_token_or_group(string)
        if arg.startswith('\\'):
            # convert single letters (so the isalpha() test below works).
            # TODO: convert all LICRs in a group (\matrm{\mu\Omega})
            arg = letters.get(arg[1:], arg)
        if name == 'mathbb':
            # mathvariant="double-struck" is ignored for Greek letters
            # (tested in Firefox 78). Use literal Unicode characters.
            arg = mathbb.get(arg, arg)
        if arg.isalpha() or arg == u'\u2140':
            node = node.append(mi(arg, **attributes))
            return node, remainder
        # Wrap in