HTML rewriting

use html_rewrite.qweruse html_absolute_links, html_set_attribute from html_rewrite      // Quillo 0.6: only these

lol_html = "1.2"url = "2"

// A blog post's HTML prepared for its RSS feed and for the page: links made absolute, images// lazy, outside links marked, scripts and trackers removed.////   quillo run html_rewrite.qwe      then open http://localhost:8080/feed-htmluse "html_rewrite.qwer"let POST = `<article><p>Read <a href="/docs/routes">the routes guide</a> or <a href="https://github.com/quillolang">the code</a>.</p><img src="../img/cover.jpg" srcset="../img/cover.jpg 1x, ../img/cover@2x.jpg 2x" alt="Cover"><script src="/js/tracker.js"></script><p class="ad">Buy now!</p><p>Questions? <a href="#comments">Comment</a> or <a href="mailto:ada@example.com">mail me</a>.</p></article>`route GET /feed-html {    let clean = html_remove(html_remove(POST, "script", false)?, ".ad", false)?    respond(200, html_absolute_links(clean, "https://blog.example.com/posts/2026/hello")?, { "content-type": "text/plain; charset=utf-8" })}route GET /page = respond(200, html_rewrite(html_external_links(POST, "blog.example.com")?, [    { selector: "img", set: { loading: "lazy", decoding: "async" } },    { selector: "script, .ad", remove: true },    { selector: "article", append_html: "<footer>Thanks for reading</footer>" },])?, { "content-type": "text/plain; charset=utf-8" })route GET /bad = { error: html_remove(POST, "p:nth-child(2n+1 x", false).error }

//! HTML rewritten as it streams, with CSS selectors (Cloudflare's lol-html): links and images made absolute, attributes set or removed (loading="lazy", rel="noopener"), elements removed or unwrapped, text replaced. Fast on big pages, keeps the rest of the markup as it was.//! version: 1.0.0//!//! [dependencies]//! lol_html = "1.2"//! url = "2"use lol_html::html_content::{ContentType, Element};use lol_html::{ElementContentHandlers, HandlerResult, RewriteStrSettings, Selector};use serde_json::Value;use std::borrow::Cow;const MAX_HTML: usize = 32 << 20;fn sel(s: &str) -> Result<Selector, String> {    s.parse::<Selector>().map_err(|e| format!("`{}` is not a selector lol-html takes: {}", s, e))}type Handler<'h> = (Cow<'h, Selector>, ElementContentHandlers<'h>);fn on<'h>(selector: &str, f: impl FnMut(&mut Element) -> HandlerResult + 'h) -> Result<Handler<'h>, String> {    Ok((Cow::Owned(sel(selector)?), ElementContentHandlers::default().element(f)))}fn run(html: &str, handlers: Vec<Handler<'_>>) -> Result<String, String> {    if html.len() > MAX_HTML {        return Err("the HTML is at most 32 MB".into());    }    lol_html::rewrite_str(html, RewriteStrSettings { element_content_handlers: handlers, ..RewriteStrSettings::default() }).map_err(|e| e.to_string())}fn absolute(base: &url::Url, v: &str) -> String {    let t = v.trim();    if t.is_empty() || t.starts_with('#') || t.starts_with("data:") || t.starts_with("mailto:") || t.starts_with("tel:") || t.starts_with("javascript:") {        return v.to_string();    }    base.join(t).map(|u| u.to_string()).unwrap_or_else(|_| v.to_string())}/// Every link and source made absolute against `base_url` (the page's address): `href` of `a`, `link` and `area`, `src` of `img`, `script`, `iframe`, `source`, `video`, `audio`, `track`,/// `srcset` entries, `poster` and form `action`s. For feeds, emails and pages shown elsewhere. Anchors (`#top`), `data:` and `mailto:` are left alone.pub fn html_absolute_links(html: &str, base_url: &str) -> Result<String, String> {    let base = url::Url::parse(base_url).map_err(|e| format!("`{}` is not a URL: {}", base_url, e))?;    // Each handler owns its copy of the base URL.    let attr = |name: &'static str| {        let b = base.clone();        move |el: &mut Element| -> HandlerResult {            if let Some(v) = el.get_attribute(name) {                el.set_attribute(name, &absolute(&b, &v))?;            }            Ok(())        }    };    let b = base.clone();    let srcset = move |el: &mut Element| -> HandlerResult {        if let Some(v) = el.get_attribute("srcset") {            let parts: Vec<String> = v                .split(',')                .map(|p| {                    let p = p.trim();                    match p.split_once(char::is_whitespace) {                        Some((u, d)) => format!("{} {}", absolute(&b, u), d.trim()),                        None => absolute(&b, p),                    }                })                .collect();            el.set_attribute("srcset", &parts.join(", "))?;        }        Ok(())    };    run(html, vec![        on("a[href], link[href], area[href]", attr("href"))?,        on("img[src], script[src], iframe[src], source[src], video[src], audio[src], track[src], embed[src]", attr("src"))?,        on("img[srcset], source[srcset]", srcset)?,        on("video[poster]", attr("poster"))?,        on("form[action]", attr("action"))?,    ])}/// Sets attribute `name` to `value` on every element matching `selector` (`img`, `a[target=_blank]`, `table.data`): `loading="lazy"` on images, `rel="noopener"` on new-tab links./// `only_if_missing` keeps a value already there.pub fn html_set_attribute(html: &str, selector: &str, name: &str, value: &str, only_if_missing: bool) -> Result<String, String> {    run(html, vec![on(selector, |el: &mut Element| {        if !(only_if_missing && el.has_attribute(name)) {            el.set_attribute(name, value)?;        }        Ok(())    })?])}/// Removes every element matching `selector` (`script`, `iframe`, `.ad`, `[onclick]`): with all it holds, or, when `keep_content`, only its tags (unwrapping a `<font>` or a `<span class=x>`).pub fn html_remove(html: &str, selector: &str, keep_content: bool) -> Result<String, String> {    run(html, vec![on(selector, |el: &mut Element| {        if keep_content { el.remove_and_keep_content() } else { el.remove() }        Ok(())    })?])}/// Several changes in one pass: `rules` are `[{ selector, set: { attr: value }, remove_attributes: [...], remove: true, unwrap: true, text: "...", html: "...", append_html: "..." }]`,/// applied in order to the elements each selector matches (`text` replaces the content with escaped text, `html` with markup, `append_html` adds at the end).pub fn html_rewrite(html: &str, rules: Vec<serde_json::Value>) -> Result<String, String> {    if rules.len() > 100 {        return Err("at most 100 rules".into());    }    let mut hs = Vec::new();    for r in &rules {        let selector = r["selector"].as_str().ok_or("each rule has a `selector`")?;        let r = r.clone();        hs.push(on(selector, move |el: &mut Element| {            if let Some(set) = r["set"].as_object() {                for (k, v) in set {                    let v = match v {                        Value::String(s) => s.clone(),                        Value::Null => continue,                        other => other.to_string(),                    };                    el.set_attribute(k, &v)?;                }            }            if let Some(rm) = r["remove_attributes"].as_array() {                for a in rm.iter().filter_map(Value::as_str) {                    el.remove_attribute(a);                }            }            if let Some(t) = r["text"].as_str() {                el.set_inner_content(t, ContentType::Text);            }            if let Some(h) = r["html"].as_str() {                el.set_inner_content(h, ContentType::Html);            }            if let Some(h) = r["append_html"].as_str() {                el.append(h, ContentType::Html);            }            if r["remove"] == true {                el.remove();            } else if r["unwrap"] == true {                el.remove_and_keep_content();            }            Ok(())        })?);    }    run(html, hs)}/// Links to other sites marked as such: `rel="nofollow noopener"` and `target="_blank"` on every `a[href]` whose host is not `own_host` (or one of its subdomains). For user content and outbound tracking.pub fn html_external_links(html: &str, own_host: &str) -> Result<String, String> {    let own = own_host.trim().trim_start_matches("www.").to_lowercase();    run(html, vec![on("a[href]", move |el: &mut Element| {        let href = el.get_attribute("href").unwrap_or_default();        if let Ok(u) = url::Url::parse(href.trim()) {            let host = u.host_str().unwrap_or("").trim_start_matches("www.").to_lowercase();            let external = (u.scheme() == "http" || u.scheme() == "https") && host != own && !host.ends_with(&format!(".{}", own));            if external {                el.set_attribute("rel", "nofollow noopener")?;                el.set_attribute("target", "_blank")?;            }        }        Ok(())    })?])}