servo_fetch/
fetch.rs

1//! Single-page fetching and rendered content extraction.
2
3use std::collections::HashMap;
4use std::fmt;
5use std::sync::Arc;
6use std::time::Duration;
7
8use serde_json::Value;
9use servo::accesskit::{Node, NodeId};
10
11use crate::error::Error;
12use crate::net::sanitize_user_agent;
13
14/// Rendered page returned by [`fetch`].
15#[derive(Debug, Clone, Default, serde::Serialize)]
16#[non_exhaustive]
17pub struct Page {
18    /// Fully rendered HTML after JavaScript execution.
19    pub html: String,
20    /// Plain text content (`document.body.innerText`).
21    pub inner_text: String,
22    /// Page title extracted from `<title>` tag.
23    pub title: Option<String>,
24    /// Parsed layout data from the injected CSS heuristics script.
25    #[serde(skip_serializing_if = "Option::is_none")]
26    pub layout_json: Option<String>,
27    /// Per-node visibility flags from the visibility-aware extraction pass.
28    #[serde(skip)]
29    visibility_json: Option<String>,
30    /// Result of JavaScript evaluation, if [`FetchOptions::javascript`] was used.
31    #[serde(skip_serializing_if = "Option::is_none")]
32    pub js_result: Option<String>,
33    /// Browser console messages captured during page load.
34    pub console_messages: Vec<ConsoleMessage>,
35    /// Accessibility tree (AccessKit), serialized as JSON, if requested.
36    #[serde(skip_serializing_if = "Option::is_none")]
37    pub accessibility_tree: Option<String>,
38    /// Structured data extracted via [`FetchOptions::schema`].
39    #[serde(skip_serializing_if = "Option::is_none")]
40    pub extracted: Option<Value>,
41    /// PNG-encoded screenshot bytes — read via [`Page::screenshot_png`].
42    #[serde(skip)]
43    screenshot_png: Option<Vec<u8>>,
44    /// Typed AccessKit tree, shared cheaply across [`Page`] clones.
45    #[serde(skip)]
46    a11y: Option<Arc<HashMap<NodeId, Node>>>,
47    /// Visibility policy that was active when this page was fetched.
48    #[serde(skip)]
49    visibility_policy: crate::visibility::VisibilityPolicy,
50}
51
52impl Page {
53    /// Extract readable Markdown from this page.
54    pub fn markdown(&self) -> crate::error::Result<String> {
55        self.markdown_with_url("")
56    }
57
58    /// Extract readable Markdown, using the original URL for link resolution.
59    pub fn markdown_with_url(&self, url: &str) -> crate::error::Result<String> {
60        Ok(crate::extract::extract_text(&self.extract_input(url, None))?)
61    }
62
63    /// Extract structured JSON from this page.
64    pub fn extract_json(&self) -> crate::error::Result<String> {
65        self.extract_json_with_url("")
66    }
67
68    /// Extract structured JSON, using the original URL for link resolution.
69    pub fn extract_json_with_url(&self, url: &str) -> crate::error::Result<String> {
70        Ok(crate::extract::extract_json(&self.extract_input(url, None))?)
71    }
72
73    /// Extract readable Markdown from the subtree matched by a CSS selector.
74    pub fn markdown_with_selector(&self, url: &str, selector: &str) -> crate::error::Result<String> {
75        Ok(crate::extract::extract_text(&self.extract_input(url, Some(selector)))?)
76    }
77
78    /// Extract structured JSON from the subtree matched by a CSS selector.
79    pub fn extract_json_with_selector(&self, url: &str, selector: &str) -> crate::error::Result<String> {
80        Ok(crate::extract::extract_json(&self.extract_input(url, Some(selector)))?)
81    }
82
83    /// PNG screenshot bytes, if captured via [`FetchOptions::screenshot`].
84    #[must_use]
85    pub fn screenshot_png(&self) -> Option<&[u8]> {
86        self.screenshot_png.as_deref()
87    }
88
89    fn extract_input<'a>(&'a self, url: &'a str, selector: Option<&'a str>) -> crate::extract::ExtractInput<'a> {
90        crate::extract::ExtractInput::new(&self.html, url)
91            .with_layout_json(self.layout_json.as_deref())
92            .with_visibility_json(self.visibility_json.as_deref())
93            .with_a11y(self.a11y.as_deref())
94            .with_inner_text(Some(&self.inner_text))
95            .with_selector(selector)
96            .with_visibility(self.visibility_policy)
97    }
98
99    pub(crate) fn from_servo(page: crate::bridge::ServoPage) -> Self {
100        let title = {
101            let doc = dom_query::Document::from(page.html.as_str());
102            let t = doc.select("title").text().to_string();
103            if t.is_empty() { None } else { Some(t) }
104        };
105        let screenshot_png = page.screenshot.and_then(|img| {
106            let mut buf = std::io::Cursor::new(Vec::new());
107            img.write_to(&mut buf, image::ImageFormat::Png).ok()?;
108            Some(buf.into_inner())
109        });
110        Self {
111            html: page.html,
112            inner_text: page.inner_text.unwrap_or_default(),
113            title,
114            layout_json: page.layout_json,
115            visibility_json: page.visibility_json,
116            js_result: page.js_result,
117            console_messages: page
118                .console_messages
119                .into_iter()
120                .map(|m| ConsoleMessage {
121                    level: match m.level {
122                        crate::bridge::ConsoleLevel::Log => ConsoleLevel::Log,
123                        crate::bridge::ConsoleLevel::Debug => ConsoleLevel::Debug,
124                        crate::bridge::ConsoleLevel::Info => ConsoleLevel::Info,
125                        crate::bridge::ConsoleLevel::Warn => ConsoleLevel::Warn,
126                        crate::bridge::ConsoleLevel::Error => ConsoleLevel::Error,
127                        crate::bridge::ConsoleLevel::Trace => ConsoleLevel::Trace,
128                    },
129                    message: m.message,
130                })
131                .collect(),
132            screenshot_png,
133            accessibility_tree: page.accessibility_tree,
134            a11y: page.a11y.map(Arc::new),
135            extracted: None,
136            visibility_policy: crate::visibility::VisibilityPolicy::default(),
137        }
138    }
139}
140
141/// Browser console message captured during page load.
142#[derive(Debug, Clone, PartialEq, Eq, serde::Serialize)]
143#[non_exhaustive]
144pub struct ConsoleMessage {
145    /// Severity level.
146    pub level: ConsoleLevel,
147    /// Message text.
148    pub message: String,
149}
150
151/// Console message severity.
152#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize)]
153#[serde(rename_all = "lowercase")]
154#[non_exhaustive]
155pub enum ConsoleLevel {
156    /// General log message.
157    Log,
158    /// Debug-level message.
159    Debug,
160    /// Informational message.
161    Info,
162    /// Warning message.
163    Warn,
164    /// Error message.
165    Error,
166    /// Trace-level message.
167    Trace,
168}
169
170impl ConsoleLevel {
171    /// Returns the string representation of this level.
172    #[must_use]
173    pub fn as_str(&self) -> &'static str {
174        match self {
175            Self::Log => "log",
176            Self::Debug => "debug",
177            Self::Info => "info",
178            Self::Warn => "warn",
179            Self::Error => "error",
180            Self::Trace => "trace",
181        }
182    }
183}
184
185impl fmt::Display for ConsoleLevel {
186    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
187        f.pad(self.as_str())
188    }
189}
190
191#[derive(Debug, Clone, Default)]
192pub(crate) enum FetchMode {
193    #[default]
194    Content,
195    Screenshot {
196        full_page: bool,
197    },
198    JavaScript(String),
199}
200
201/// Options for a single page fetch.
202#[must_use = "options do nothing until passed to fetch()"]
203#[derive(Debug, Clone)]
204pub struct FetchOptions {
205    pub(crate) url: String,
206    pub(crate) timeout: Duration,
207    pub(crate) settle: Duration,
208    pub(crate) mode: FetchMode,
209    pub(crate) user_agent: Option<String>,
210    pub(crate) extract_schema: Option<crate::schema::ExtractSchema>,
211    pub(crate) visibility: crate::visibility::VisibilityPolicy,
212}
213
214impl FetchOptions {
215    /// Fetch rendered content (default mode).
216    pub fn new(url: &str) -> Self {
217        Self {
218            url: url.into(),
219            timeout: Duration::from_secs(30),
220            settle: Duration::ZERO,
221            mode: FetchMode::Content,
222            user_agent: None,
223            extract_schema: None,
224            visibility: crate::visibility::VisibilityPolicy::default(),
225        }
226    }
227
228    /// Capture a PNG screenshot.
229    pub fn screenshot(url: &str, full_page: bool) -> Self {
230        Self {
231            mode: FetchMode::Screenshot { full_page },
232            ..Self::new(url)
233        }
234    }
235
236    /// Execute a JavaScript expression and return the result.
237    pub fn javascript(url: &str, expression: impl Into<String>) -> Self {
238        Self {
239            mode: FetchMode::JavaScript(expression.into()),
240            ..Self::new(url)
241        }
242    }
243
244    /// Page load timeout (default: 30s).
245    pub fn timeout(mut self, timeout: Duration) -> Self {
246        self.timeout = timeout;
247        self
248    }
249
250    /// Extra wait after load event for SPA hydration (default: 0).
251    pub fn settle(mut self, settle: Duration) -> Self {
252        self.settle = settle;
253        self
254    }
255
256    /// Override the User-Agent string for this request.
257    pub fn user_agent(mut self, ua: impl Into<String>) -> Self {
258        self.user_agent = Some(sanitize_user_agent(ua.into()));
259        self
260    }
261
262    /// Extract structured data from the rendered page using the given schema.
263    pub fn schema(mut self, schema: crate::schema::ExtractSchema) -> Self {
264        self.extract_schema = Some(schema);
265        self
266    }
267
268    /// Visibility-filtering policy applied during extraction.
269    pub fn visibility(mut self, policy: crate::visibility::VisibilityPolicy) -> Self {
270        self.visibility = policy;
271        self
272    }
273}
274
275/// Fetch a single page via the embedded Servo engine.
276#[allow(clippy::needless_pass_by_value)]
277pub fn fetch(opts: FetchOptions) -> crate::error::Result<Page> {
278    crate::net::ensure_crypto_provider();
279
280    crate::net::validate_url(&opts.url)?;
281
282    if matches!(opts.mode, FetchMode::Content)
283        && let Some(bytes) = crate::pdf::probe(&opts.url, opts.timeout.as_secs().max(1))
284    {
285        let text = crate::extract::extract_pdf(&bytes);
286        return Ok(Page {
287            html: String::new(),
288            inner_text: text,
289            ..Page::default()
290        });
291    }
292
293    let bridge_opts = crate::bridge::FetchOptions {
294        url: &opts.url,
295        timeout_secs: opts.timeout.as_secs().max(1),
296        settle_ms: u64::try_from(opts.settle.as_millis()).unwrap_or(u64::MAX),
297        user_agent: opts.user_agent.as_deref(),
298        mode: match opts.mode {
299            FetchMode::Content => crate::bridge::FetchMode::Content { include_a11y: false },
300            FetchMode::Screenshot { full_page } => crate::bridge::FetchMode::Screenshot { full_page },
301            FetchMode::JavaScript(ref expr) => crate::bridge::FetchMode::ExecuteJs {
302                expression: expr.clone(),
303            },
304        },
305    };
306
307    let servo_page = crate::bridge::fetch_page(bridge_opts).map_err(|e| {
308        if format!("{e:#}").contains("timed out") {
309            Error::Timeout {
310                url: opts.url.clone(),
311                timeout: opts.timeout,
312            }
313        } else {
314            Error::engine(e, Some(opts.url.clone()))
315        }
316    })?;
317
318    let mut page = Page::from_servo(servo_page);
319    page.visibility_policy = opts.visibility;
320    if let Some(schema) = opts.extract_schema.as_ref() {
321        page.extracted = Some(schema.extract_from(&page.html));
322    }
323    Ok(page)
324}
325
326/// Fetch a URL and return readable Markdown.
327pub fn markdown(url: &str) -> crate::error::Result<String> {
328    fetch(FetchOptions::new(url))?.markdown_with_url(url)
329}
330
331/// Fetch a URL and return structured JSON.
332pub fn extract_json(url: &str) -> crate::error::Result<String> {
333    fetch(FetchOptions::new(url))?.extract_json_with_url(url)
334}
335
336/// Fetch a URL and return plain text (`document.body.innerText`).
337pub fn text(url: &str) -> crate::error::Result<String> {
338    Ok(fetch(FetchOptions::new(url))?.inner_text)
339}
340
341#[cfg(test)]
342mod tests {
343    use super::*;
344
345    #[test]
346    fn fetch_options_defaults() {
347        let opts = FetchOptions::new("https://example.com");
348        assert_eq!(opts.url, "https://example.com");
349        assert_eq!(opts.timeout, Duration::from_secs(30));
350        assert_eq!(opts.settle, Duration::ZERO);
351        assert!(matches!(opts.mode, FetchMode::Content));
352    }
353
354    #[test]
355    fn fetch_options_screenshot() {
356        let opts = FetchOptions::screenshot("https://example.com", true);
357        assert!(matches!(opts.mode, FetchMode::Screenshot { full_page: true }));
358    }
359
360    #[test]
361    fn fetch_options_javascript() {
362        let opts = FetchOptions::javascript("https://example.com", "document.title");
363        assert!(matches!(opts.mode, FetchMode::JavaScript(ref e) if e == "document.title"));
364    }
365
366    #[test]
367    fn fetch_options_chaining() {
368        let opts = FetchOptions::new("https://example.com")
369            .timeout(Duration::from_secs(60))
370            .settle(Duration::from_millis(500));
371        assert_eq!(opts.timeout, Duration::from_secs(60));
372        assert_eq!(opts.settle, Duration::from_millis(500));
373    }
374
375    #[test]
376    fn fetch_user_agent_set() {
377        let opts = FetchOptions::new("https://example.com").user_agent("MyBot/1.0");
378        assert_eq!(opts.user_agent.as_deref(), Some("MyBot/1.0"));
379    }
380
381    #[test]
382    fn fetch_user_agent_default_is_none() {
383        let opts = FetchOptions::new("https://example.com");
384        assert!(opts.user_agent.is_none());
385    }
386
387    #[test]
388    fn fetch_user_agent_sanitizes_crlf() {
389        let opts = FetchOptions::new("https://example.com").user_agent("Bot\r\nX-Evil: yes");
390        assert_eq!(opts.user_agent.as_deref(), Some("Bot  X-Evil: yes"));
391    }
392
393    #[test]
394    fn fetch_user_agent_sanitizes_null() {
395        let opts = FetchOptions::new("https://example.com").user_agent("Bot\0/1.0");
396        assert_eq!(opts.user_agent.as_deref(), Some("Bot /1.0"));
397    }
398
399    #[test]
400    fn fetch_user_agent_empty_string() {
401        let opts = FetchOptions::new("https://example.com").user_agent("");
402        assert_eq!(opts.user_agent.as_deref(), Some(""));
403    }
404
405    #[test]
406    fn page_markdown_from_html() {
407        let page = Page {
408            html: "<html><head><title>Test</title></head><body><p>hello world</p></body></html>".into(),
409            inner_text: "hello world".into(),
410            ..Page::default()
411        };
412        let md = page.markdown().unwrap();
413        assert!(md.contains("hello world"));
414    }
415
416    #[test]
417    fn page_extract_json_produces_valid_json() {
418        let page = Page {
419            html: "<html><head><title>Test</title></head><body><p>content</p></body></html>".into(),
420            inner_text: "content".into(),
421            ..Page::default()
422        };
423        let json = page.extract_json().unwrap();
424        let _: Value = serde_json::from_str(&json).expect("valid JSON");
425    }
426
427    #[test]
428    fn page_screenshot_png_none_by_default() {
429        let page = Page::default();
430        assert!(page.screenshot_png().is_none());
431    }
432
433    #[test]
434    fn page_markdown_with_selector_scopes_to_subtree() {
435        let page = Page {
436            html: "<html><body><article>keep</article><aside>drop</aside></body></html>".into(),
437            ..Page::default()
438        };
439        let md = page.markdown_with_selector("https://example.com", "article").unwrap();
440        assert!(md.contains("keep"));
441        assert!(!md.contains("drop"));
442    }
443
444    #[test]
445    fn page_extract_json_with_selector_includes_url() {
446        let page = Page {
447            html: "<html><body><article>scoped</article></body></html>".into(),
448            ..Page::default()
449        };
450        let json = page
451            .extract_json_with_selector("https://example.com/page", "article")
452            .unwrap();
453        let parsed: Value = serde_json::from_str(&json).expect("valid JSON");
454        assert_eq!(parsed["url"].as_str(), Some("https://example.com/page"));
455        assert!(parsed["text_content"].as_str().unwrap().contains("scoped"));
456    }
457
458    #[test]
459    fn page_markdown_with_selector_no_match_returns_empty() {
460        let page = Page {
461            html: "<html><body><article>x</article></body></html>".into(),
462            ..Page::default()
463        };
464        let md = page.markdown_with_selector("", ".nonexistent").unwrap();
465        assert!(md.is_empty());
466    }
467
468    #[test]
469    fn page_markdown_with_invalid_selector_returns_error() {
470        let page = Page {
471            html: "<html><body><p>x</p></body></html>".into(),
472            ..Page::default()
473        };
474        let err = page.markdown_with_selector("", "###invalid[[[").unwrap_err();
475        assert!(err.to_string().contains("invalid CSS selector"));
476    }
477
478    #[test]
479    fn page_markdown_with_empty_selector_returns_error() {
480        let page = Page {
481            html: "<html><body><p>x</p></body></html>".into(),
482            ..Page::default()
483        };
484        assert!(page.markdown_with_selector("", "").is_err());
485    }
486
487    #[test]
488    fn fetch_rejects_invalid_url() {
489        let result = fetch(FetchOptions::new("not a url"));
490        assert!(result.is_err());
491        let err = result.unwrap_err();
492        assert!(matches!(err, Error::InvalidUrl { .. }));
493    }
494
495    #[test]
496    fn fetch_rejects_private_ip() {
497        let result = fetch(FetchOptions::new("http://127.0.0.1/"));
498        assert!(result.is_err());
499    }
500
501    #[test]
502    fn fetch_rejects_file_scheme() {
503        let result = fetch(FetchOptions::new("file:///etc/passwd"));
504        assert!(result.is_err());
505    }
506
507    mod page_from_servo {
508        use crate::bridge;
509        use crate::fetch::{ConsoleLevel, Page};
510
511        fn synthetic_image(w: u32, h: u32) -> image::RgbaImage {
512            image::RgbaImage::from_pixel(w, h, image::Rgba([255, 0, 0, 255]))
513        }
514
515        fn empty_servo_page() -> bridge::ServoPage {
516            bridge::ServoPage::default()
517        }
518
519        #[test]
520        fn extracts_title_from_html() {
521            let mut sp = empty_servo_page();
522            sp.html = "<html><head><title>Hello World</title></head></html>".into();
523            let page = Page::from_servo(sp);
524            assert_eq!(page.title.as_deref(), Some("Hello World"));
525        }
526
527        #[test]
528        fn title_is_none_when_tag_missing() {
529            let mut sp = empty_servo_page();
530            sp.html = "<html><body>no title here</body></html>".into();
531            let page = Page::from_servo(sp);
532            assert!(page.title.is_none());
533        }
534
535        #[test]
536        fn title_is_none_when_tag_empty() {
537            let mut sp = empty_servo_page();
538            sp.html = "<html><head><title></title></head></html>".into();
539            let page = Page::from_servo(sp);
540            assert!(page.title.is_none());
541        }
542
543        #[test]
544        fn title_is_none_for_empty_html() {
545            let page = Page::from_servo(empty_servo_page());
546            assert!(page.title.is_none());
547        }
548
549        #[test]
550        fn inner_text_none_becomes_empty_string() {
551            let sp = empty_servo_page();
552            assert!(sp.inner_text.is_none());
553            let page = Page::from_servo(sp);
554            assert_eq!(page.inner_text, "");
555        }
556
557        #[test]
558        fn screenshot_is_encoded_as_png() {
559            let mut sp = empty_servo_page();
560            sp.screenshot = Some(synthetic_image(8, 8));
561            let page = Page::from_servo(sp);
562            let bytes = page.screenshot_png().expect("screenshot encoded");
563            assert_eq!(&bytes[..8], b"\x89PNG\r\n\x1a\n", "PNG magic bytes");
564        }
565
566        #[test]
567        fn console_messages_empty_by_default() {
568            let page = Page::from_servo(empty_servo_page());
569            assert!(page.console_messages.is_empty());
570        }
571
572        #[test]
573        fn console_messages_preserve_all_six_levels() {
574            let cases = [
575                (bridge::ConsoleLevel::Log, ConsoleLevel::Log),
576                (bridge::ConsoleLevel::Debug, ConsoleLevel::Debug),
577                (bridge::ConsoleLevel::Info, ConsoleLevel::Info),
578                (bridge::ConsoleLevel::Warn, ConsoleLevel::Warn),
579                (bridge::ConsoleLevel::Error, ConsoleLevel::Error),
580                (bridge::ConsoleLevel::Trace, ConsoleLevel::Trace),
581            ];
582            for (src, expected) in cases {
583                let mut sp = empty_servo_page();
584                sp.console_messages = vec![bridge::ConsoleMessage {
585                    level: src,
586                    message: "msg".into(),
587                }];
588                let page = Page::from_servo(sp);
589                assert_eq!(
590                    page.console_messages.len(),
591                    1,
592                    "console message lost for source level {src:?}",
593                );
594                assert_eq!(
595                    page.console_messages[0].level, expected,
596                    "level mapping wrong for source {src:?}",
597                );
598            }
599        }
600
601        #[test]
602        fn console_messages_preserve_ordering_across_levels() {
603            let mut sp = empty_servo_page();
604            sp.console_messages = vec![
605                bridge::ConsoleMessage {
606                    level: bridge::ConsoleLevel::Info,
607                    message: "first".into(),
608                },
609                bridge::ConsoleMessage {
610                    level: bridge::ConsoleLevel::Error,
611                    message: "second".into(),
612                },
613                bridge::ConsoleMessage {
614                    level: bridge::ConsoleLevel::Warn,
615                    message: "third".into(),
616                },
617            ];
618            let page = Page::from_servo(sp);
619            assert_eq!(page.console_messages.len(), 3);
620            assert_eq!(page.console_messages[0].message, "first");
621            assert_eq!(page.console_messages[1].message, "second");
622            assert_eq!(page.console_messages[2].message, "third");
623            assert_eq!(page.console_messages[0].level, ConsoleLevel::Info);
624            assert_eq!(page.console_messages[1].level, ConsoleLevel::Error);
625            assert_eq!(page.console_messages[2].level, ConsoleLevel::Warn);
626        }
627
628        #[test]
629        fn extracted_starts_as_none_until_schema_applied() {
630            let page = Page::from_servo(empty_servo_page());
631            assert!(page.extracted.is_none());
632        }
633
634        #[test]
635        fn full_round_trip_preserves_every_field() {
636            let sp = bridge::ServoPage {
637                html: "<html><head><title>T</title></head><body>B</body></html>".into(),
638                inner_text: Some("B".into()),
639                layout_json: Some("[]".into()),
640                visibility_json: Some("[]".into()),
641                screenshot: Some(synthetic_image(2, 2)),
642                js_result: Some("42".into()),
643                accessibility_tree: Some("{}".into()),
644                a11y: None,
645                console_messages: vec![bridge::ConsoleMessage {
646                    level: bridge::ConsoleLevel::Log,
647                    message: "x".into(),
648                }],
649            };
650            let page = Page::from_servo(sp);
651            assert_eq!(page.html, "<html><head><title>T</title></head><body>B</body></html>");
652            assert_eq!(page.inner_text, "B");
653            assert_eq!(page.title.as_deref(), Some("T"));
654            assert_eq!(page.layout_json.as_deref(), Some("[]"));
655            assert_eq!(page.js_result.as_deref(), Some("42"));
656            assert_eq!(page.accessibility_tree.as_deref(), Some("{}"));
657            assert_eq!(page.console_messages.len(), 1);
658            assert!(page.screenshot_png().is_some());
659            assert!(page.extracted.is_none());
660        }
661    }
662}
servo_fetch/fetch.rs

servo_fetch/
fetch.rs
