fastpylight 0.1.19

Installation

pip install fastpylight

fastpylight is a small Python/Rust wrapper around Lumis, a Tree-sitter syntax highlighter with Neovim themes. Language names, token scopes, and theme names follow Lumis.

Python

from fastpylight import highlight, highlight_spans, guess, theme_css

code = "def f(x):\n    return x + 1\n"

# Output for the CSS Highlight API.
html = highlight(code, "python")

# Static span output.
spans = highlight_spans(code, "python")

# CSS rules for the span output.
css = theme_css("github_light", "pre code")

highlight returns HTML with token ranges in UTF-16 code units, for use with the CSS Highlight API:

<hl-code toks='[[0,3,"keyword"],[4,5,"function"],[6,7,"variable"],[14,20,"keyword"],[21,22,"variable"],[23,24,"operator"],[25,26,"number"]]'><pre><code>def f(x):
    return x + 1
</code></pre></hl-code>

highlight_spans returns normal HTML spans:

<pre><code><span class="hl-keyword">def</span> <span class="hl-function">f</span>(<span class="hl-variable">x</span>):
    <span class="hl-keyword">return</span> <span class="hl-variable">x</span> <span class="hl-operator">+</span> <span class="hl-number">1</span>
</code></pre>

The exact classes depend on Lumis scopes. A complete GitHub Light CSS example for both output modes is in docs/github_light.css.

Languages

The default language set enables Lumis web, web-extra, system, and backend bundles, plus R, Julia, PowerShell, Lua, Swift, MATLAB, Perl, Pascal, Fortran, and Objective-C.

Markdown is treated as quotation: fenced code bodies and other injected languages flatten to the markup.raw.block scope instead of being highlighted as the embedded language, so markdown source specimens read as markdown. Markdown's own structure (headings, fence markers, info strings, inline code) keeps its scopes.

languages() lists every available name. The highlighting functions need an exact language name and raise ValueError on an unknown one. Use "plaintext" for un-highlighted output:

highlight(code, "plaintext")  # toks='[]', no tokens

To detect a language instead of specifying one, use guess, which takes the code plus an optional hint (a name, extension, or filename) and falls back to "plaintext" when nothing matches:

guess("fn main() {}", "rs")        # 'rust'  (extension hint)
guess("#!/usr/bin/env python\n")   # 'python' (shebang)
guess("just some prose")           # 'plaintext' (no match)

highlight(code, guess(code))       # detect, then highlight

CSS Highlight API

The highlight function is for browser code that applies token ranges with the CSS Highlight API. component_js() returns the reference component:

class HlCode extends HTMLElement {
  connectedCallback(){
    if(!CSS.highlights) return;
    setTimeout(() => {
      const d=this.getAttribute('toks'); if(!d) return;
      const tn=this.querySelector('code').firstChild, toks=JSON.parse(d);
      this._ranges=[];
      toks.forEach(([s,e,k])=>{
        const r=new Range(); r.setStart(tn,s); r.setEnd(tn,e);
        const h=CSS.highlights.get(k)||new Highlight();
        h.add(r); CSS.highlights.set(k,h);
        this._ranges.push([r,k]);
      });
      this.removeAttribute('toks');
    }, 0);
  }
  disconnectedCallback(){
    if(!this._ranges) return;
    this._ranges.forEach(([r,k])=>{
      const h=CSS.highlights.get(k); if(h) h.delete(r);
    });
    this._ranges=null;
  }
}
if(!customElements.get('hl-code')) customElements.define('hl-code',HlCode);

highlight needs ::highlight(...) rules. highlight_spans needs class rules. docs/github_light.css includes both for the default hl- prefix.

You can generate CSS at runtime:

highlight_css = theme_css("github_light")
span_css = theme_css("github_light", "pre code")

For non-CSS consumers (e.g. document converters), theme_colors("github_light") returns the same theme as data: a dict of dotted scope names (including normal, which the CSS omits) to {'fg', 'bg', 'bold', 'italic', 'underline', 'strikethrough'}.