microblog.pub/app/utils/highlight.py

import base64
import hashlib
from functools import lru_cache

from bs4 import BeautifulSoup  # type: ignore
from pygments import highlight as phighlight  # type: ignore
from pygments.formatters import HtmlFormatter  # type: ignore
from pygments.lexers import get_lexer_by_name  # type: ignore
from pygments.lexers import guess_lexer  # type: ignore

from app.config import CODE_HIGHLIGHTING_THEME

_FORMATTER = HtmlFormatter(style=CODE_HIGHLIGHTING_THEME)

HIGHLIGHT_CSS = _FORMATTER.get_style_defs()
HIGHLIGHT_CSS_HASH = base64.b64encode(
    hashlib.sha256(HIGHLIGHT_CSS.encode()).digest()
).decode()


@lru_cache(256)
def highlight(html: str) -> str:
    soup = BeautifulSoup(html, "html5lib")
    for code in soup.find_all("code"):
        if not code.parent.name == "pre":
            continue

        # Replace <br> tags with line breaks (Mastodon sends code like this)
        code_content = (
            code.encode_contents().decode().replace("<br>", "\n").replace("<br/>", "\n")
        )

        # If this comes from a microblog.pub instance we may have the language
        # in the class name
        if "class" in code.attrs and code.attrs["class"][0].startswith("language-"):
            try:
                lexer = get_lexer_by_name(
                    code.attrs["class"][0].removeprefix("language-")
                )
            except Exception:
                lexer = guess_lexer(code_content)
        else:
            lexer = guess_lexer(code_content)

        # Replace the code with Pygment output
        # XXX: the HTML escaping causes issue with Python type annotations
        code_content = code_content.replace(") -&gt; ", ") -> ")
        code.parent.replaceWith(
            BeautifulSoup(
                phighlight(code_content, lexer, _FORMATTER), "html5lib"
            ).body.next
        )

    return soup.body.encode_contents().decode()