enjoysmath
commented
5 years ago Here's the code using Lark:
from lark import Lark, Transformer
from lark.lexer import Token
class Parser:
grammar = """
?start: string
?string: (latex_string | any_str)+
?latex_string: block_latex
| inline_latex
?inline_latex: "$" inner_latex+ "$"
?block_latex: "$$" inner_latex+ "$$"
?inner_latex: variable
| integer
| text_block
| any_str
?variable: (LATIN | greek)
?text_block: "\\\\" TEXT_COMMAND "{" ANY_STR "}"
integer: SIGNED_INT
?greek: GREEK_LOWER -> greek
| GREEK_UPPER -> greek
GREEK_LOWER: /\\\\alpha|\\\\beta|\\\\gamma|\\\\delta|\\\\epsilon|\\\\zeta|\\\\eta|\\\\theta/
| /\\\\iota|\\\\kappa|\\\\lambda|\\\\mu|\\\\xi|\\\\omicron|\\\\pi|\\\\rho|\\\\sigma/
| /\\\\tau|\\\\upsilon|\\\\phi|\\\\psi|\\\\chi|\\\\omega/
GREEK_UPPER: /\\\\Alpha|\\\\Beta|\\\\Gamma|\\\\Delta|\\\\Epsilon|\\\\Zeta|\\\\Eta|\\\\Theta/
| /\\\\Iota|\\\\Kappa|\\\\Lambda|\\\\Mu|\\\\Xi|\\\\Omicron|\\\\Pi|\\\\Rho|\\\\Sigma/
| /\\\\Tau|\\\\Upsilon|\\\\Phi|\\\\Psi|\\\\Chi|\\\\Omega/
TEXT_COMMAND: /text|textbf|operatorname/
LATIN: /[a-zA-Z]/
any_str: ANY_STR
ANY_STR: /[^{}$]+/
%import common.SIGNED_INT
%import common.WS_INLINE
%ignore WS_INLINE
"""
def __init__(self, new=True):
if new is None:
new = True
if new:
self._transformer = None
self._parser = None
Parser.setup(self)
def setup(self):
self._transformer = ParseTreeTransformer()
self._parser = Lark(self.grammar)
def parse_to_list(self, s):
return self._transformer.transform(self._parser.parse(s))
class ParseTreeTransformer(Transformer):
pass
if __name__ == '__main__':
parser = Parser()
while True:
s = input("s=")
result = parser.parse_to_list(s)
print(result)Feel free to run it, and type in "Let $G$ be a group." It will parse, except only recognize $G$ as an any_str and not a variable. Therefore there's some bug in Lark that's causing any_str to take priority over my more detailed rules.
Any ideas?
erezsh
MegaIng
As you can see the code is becoming more complex than necessary given that Lark exists.
$\text{abc} a$ => ["\text{abc} ", Var("a")] is the result I want.
Anything surrounded by $ or $$ in MathJax is considered LaTeX. Anything latin or greek (\alpha e.g.) character I encounter inside dollar signs will be a variable. Except \text{...} might occur inside of dollar signs, in which case anthing inside of ... should not be parsed as a variable, but the entire string \text{...} is left alone. Same goes for \textbf and \operatorname.
So as you can see, using regexes has led me to the above code, in fact the simplest way you could do it, and the code is not finished (it doesn't run because I didn't finish it). Therefore because of all this complicatedness, I think Lark suits me best. However, I've tried several times to do this in Lark and failed for one reason or another.
I will be posting the Lark version in about 30 mins. So keep this open, please.