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        let msg = format!("{e:#}");
309        if msg.contains("timed out") {
310            Error::Timeout {
311                url: opts.url.clone(),
312                timeout: opts.timeout,
313            }
314        } else {
315            Error::Engine(msg)
316        }
317    })?;
318
319    let mut page = Page::from_servo(servo_page);
320    page.visibility_policy = opts.visibility;
321    if let Some(schema) = opts.extract_schema.as_ref() {
322        page.extracted = Some(schema.extract_from(&page.html));
323    }
324    Ok(page)
325}
326
327/// Fetch a URL and return readable Markdown.
328pub fn markdown(url: &str) -> crate::error::Result<String> {
329    fetch(FetchOptions::new(url))?.markdown_with_url(url)
330}
331
332/// Fetch a URL and return structured JSON.
333pub fn extract_json(url: &str) -> crate::error::Result<String> {
334    fetch(FetchOptions::new(url))?.extract_json_with_url(url)
335}
336
337/// Fetch a URL and return plain text (`document.body.innerText`).
338pub fn text(url: &str) -> crate::error::Result<String> {
339    Ok(fetch(FetchOptions::new(url))?.inner_text)
340}
341
342#[cfg(test)]
343mod tests {
344    use super::*;
345
346    #[test]
347    fn fetch_options_defaults() {
348        let opts = FetchOptions::new("https://example.com");
349        assert_eq!(opts.url, "https://example.com");
350        assert_eq!(opts.timeout, Duration::from_secs(30));
351        assert_eq!(opts.settle, Duration::ZERO);
352        assert!(matches!(opts.mode, FetchMode::Content));
353    }
354
355    #[test]
356    fn fetch_options_screenshot() {
357        let opts = FetchOptions::screenshot("https://example.com", true);
358        assert!(matches!(opts.mode, FetchMode::Screenshot { full_page: true }));
359    }
360
361    #[test]
362    fn fetch_options_javascript() {
363        let opts = FetchOptions::javascript("https://example.com", "document.title");
364        assert!(matches!(opts.mode, FetchMode::JavaScript(ref e) if e == "document.title"));
365    }
366
367    #[test]
368    fn fetch_options_chaining() {
369        let opts = FetchOptions::new("https://example.com")
370            .timeout(Duration::from_secs(60))
371            .settle(Duration::from_millis(500));
372        assert_eq!(opts.timeout, Duration::from_secs(60));
373        assert_eq!(opts.settle, Duration::from_millis(500));
374    }
375
376    #[test]
377    fn fetch_user_agent_set() {
378        let opts = FetchOptions::new("https://example.com").user_agent("MyBot/1.0");
379        assert_eq!(opts.user_agent.as_deref(), Some("MyBot/1.0"));
380    }
381
382    #[test]
383    fn fetch_user_agent_default_is_none() {
384        let opts = FetchOptions::new("https://example.com");
385        assert!(opts.user_agent.is_none());
386    }
387
388    #[test]
389    fn fetch_user_agent_sanitizes_crlf() {
390        let opts = FetchOptions::new("https://example.com").user_agent("Bot\r\nX-Evil: yes");
391        assert_eq!(opts.user_agent.as_deref(), Some("Bot  X-Evil: yes"));
392    }
393
394    #[test]
395    fn fetch_user_agent_sanitizes_null() {
396        let opts = FetchOptions::new("https://example.com").user_agent("Bot\0/1.0");
397        assert_eq!(opts.user_agent.as_deref(), Some("Bot /1.0"));
398    }
399
400    #[test]
401    fn fetch_user_agent_empty_string() {
402        let opts = FetchOptions::new("https://example.com").user_agent("");
403        assert_eq!(opts.user_agent.as_deref(), Some(""));
404    }
405
406    #[test]
407    fn page_markdown_from_html() {
408        let page = Page {
409            html: "<html><head><title>Test</title></head><body><p>hello world</p></body></html>".into(),
410            inner_text: "hello world".into(),
411            ..Page::default()
412        };
413        let md = page.markdown().unwrap();
414        assert!(md.contains("hello world"));
415    }
416
417    #[test]
418    fn page_extract_json_produces_valid_json() {
419        let page = Page {
420            html: "<html><head><title>Test</title></head><body><p>content</p></body></html>".into(),
421            inner_text: "content".into(),
422            ..Page::default()
423        };
424        let json = page.extract_json().unwrap();
425        let _: Value = serde_json::from_str(&json).expect("valid JSON");
426    }
427
428    #[test]
429    fn page_screenshot_png_none_by_default() {
430        let page = Page::default();
431        assert!(page.screenshot_png().is_none());
432    }
433
434    #[test]
435    fn page_markdown_with_selector_scopes_to_subtree() {
436        let page = Page {
437            html: "<html><body><article>keep</article><aside>drop</aside></body></html>".into(),
438            ..Page::default()
439        };
440        let md = page.markdown_with_selector("https://example.com", "article").unwrap();
441        assert!(md.contains("keep"));
442        assert!(!md.contains("drop"));
443    }
444
445    #[test]
446    fn page_extract_json_with_selector_includes_url() {
447        let page = Page {
448            html: "<html><body><article>scoped</article></body></html>".into(),
449            ..Page::default()
450        };
451        let json = page
452            .extract_json_with_selector("https://example.com/page", "article")
453            .unwrap();
454        let parsed: Value = serde_json::from_str(&json).expect("valid JSON");
455        assert_eq!(parsed["url"].as_str(), Some("https://example.com/page"));
456        assert!(parsed["text_content"].as_str().unwrap().contains("scoped"));
457    }
458
459    #[test]
460    fn page_markdown_with_selector_no_match_returns_empty() {
461        let page = Page {
462            html: "<html><body><article>x</article></body></html>".into(),
463            ..Page::default()
464        };
465        let md = page.markdown_with_selector("", ".nonexistent").unwrap();
466        assert!(md.is_empty());
467    }
468
469    #[test]
470    fn page_markdown_with_invalid_selector_returns_error() {
471        let page = Page {
472            html: "<html><body><p>x</p></body></html>".into(),
473            ..Page::default()
474        };
475        let err = page.markdown_with_selector("", "###invalid[[[").unwrap_err();
476        assert!(err.to_string().contains("invalid CSS selector"));
477    }
478
479    #[test]
480    fn page_markdown_with_empty_selector_returns_error() {
481        let page = Page {
482            html: "<html><body><p>x</p></body></html>".into(),
483            ..Page::default()
484        };
485        assert!(page.markdown_with_selector("", "").is_err());
486    }
487
488    #[test]
489    fn fetch_rejects_invalid_url() {
490        let result = fetch(FetchOptions::new("not a url"));
491        assert!(result.is_err());
492        let err = result.unwrap_err();
493        assert!(matches!(err, Error::InvalidUrl { .. }));
494    }
495
496    #[test]
497    fn fetch_rejects_private_ip() {
498        let result = fetch(FetchOptions::new("http://127.0.0.1/"));
499        assert!(result.is_err());
500    }
501
502    #[test]
503    fn fetch_rejects_file_scheme() {
504        let result = fetch(FetchOptions::new("file:///etc/passwd"));
505        assert!(result.is_err());
506    }
507
508    mod page_from_servo {
509        use crate::bridge;
510        use crate::fetch::{ConsoleLevel, Page};
511
512        fn synthetic_image(w: u32, h: u32) -> image::RgbaImage {
513            image::RgbaImage::from_pixel(w, h, image::Rgba([255, 0, 0, 255]))
514        }
515
516        fn empty_servo_page() -> bridge::ServoPage {
517            bridge::ServoPage::default()
518        }
519
520        #[test]
521        fn extracts_title_from_html() {
522            let mut sp = empty_servo_page();
523            sp.html = "<html><head><title>Hello World</title></head></html>".into();
524            let page = Page::from_servo(sp);
525            assert_eq!(page.title.as_deref(), Some("Hello World"));
526        }
527
528        #[test]
529        fn title_is_none_when_tag_missing() {
530            let mut sp = empty_servo_page();
531            sp.html = "<html><body>no title here</body></html>".into();
532            let page = Page::from_servo(sp);
533            assert!(page.title.is_none());
534        }
535
536        #[test]
537        fn title_is_none_when_tag_empty() {
538            let mut sp = empty_servo_page();
539            sp.html = "<html><head><title></title></head></html>".into();
540            let page = Page::from_servo(sp);
541            assert!(page.title.is_none());
542        }
543
544        #[test]
545        fn title_is_none_for_empty_html() {
546            let page = Page::from_servo(empty_servo_page());
547            assert!(page.title.is_none());
548        }
549
550        #[test]
551        fn inner_text_none_becomes_empty_string() {
552            let sp = empty_servo_page();
553            assert!(sp.inner_text.is_none());
554            let page = Page::from_servo(sp);
555            assert_eq!(page.inner_text, "");
556        }
557
558        #[test]
559        fn screenshot_is_encoded_as_png() {
560            let mut sp = empty_servo_page();
561            sp.screenshot = Some(synthetic_image(8, 8));
562            let page = Page::from_servo(sp);
563            let bytes = page.screenshot_png().expect("screenshot encoded");
564            assert_eq!(&bytes[..8], b"\x89PNG\r\n\x1a\n", "PNG magic bytes");
565        }
566
567        #[test]
568        fn console_messages_empty_by_default() {
569            let page = Page::from_servo(empty_servo_page());
570            assert!(page.console_messages.is_empty());
571        }
572
573        #[test]
574        fn console_messages_preserve_all_six_levels() {
575            let cases = [
576                (bridge::ConsoleLevel::Log, ConsoleLevel::Log),
577                (bridge::ConsoleLevel::Debug, ConsoleLevel::Debug),
578                (bridge::ConsoleLevel::Info, ConsoleLevel::Info),
579                (bridge::ConsoleLevel::Warn, ConsoleLevel::Warn),
580                (bridge::ConsoleLevel::Error, ConsoleLevel::Error),
581                (bridge::ConsoleLevel::Trace, ConsoleLevel::Trace),
582            ];
583            for (src, expected) in cases {
584                let mut sp = empty_servo_page();
585                sp.console_messages = vec![bridge::ConsoleMessage {
586                    level: src,
587                    message: "msg".into(),
588                }];
589                let page = Page::from_servo(sp);
590                assert_eq!(
591                    page.console_messages.len(),
592                    1,
593                    "console message lost for source level {src:?}",
594                );
595                assert_eq!(
596                    page.console_messages[0].level, expected,
597                    "level mapping wrong for source {src:?}",
598                );
599            }
600        }
601
602        #[test]
603        fn console_messages_preserve_ordering_across_levels() {
604            let mut sp = empty_servo_page();
605            sp.console_messages = vec![
606                bridge::ConsoleMessage {
607                    level: bridge::ConsoleLevel::Info,
608                    message: "first".into(),
609                },
610                bridge::ConsoleMessage {
611                    level: bridge::ConsoleLevel::Error,
612                    message: "second".into(),
613                },
614                bridge::ConsoleMessage {
615                    level: bridge::ConsoleLevel::Warn,
616                    message: "third".into(),
617                },
618            ];
619            let page = Page::from_servo(sp);
620            assert_eq!(page.console_messages.len(), 3);
621            assert_eq!(page.console_messages[0].message, "first");
622            assert_eq!(page.console_messages[1].message, "second");
623            assert_eq!(page.console_messages[2].message, "third");
624            assert_eq!(page.console_messages[0].level, ConsoleLevel::Info);
625            assert_eq!(page.console_messages[1].level, ConsoleLevel::Error);
626            assert_eq!(page.console_messages[2].level, ConsoleLevel::Warn);
627        }
628
629        #[test]
630        fn extracted_starts_as_none_until_schema_applied() {
631            let page = Page::from_servo(empty_servo_page());
632            assert!(page.extracted.is_none());
633        }
634
635        #[test]
636        fn full_round_trip_preserves_every_field() {
637            let sp = bridge::ServoPage {
638                html: "<html><head><title>T</title></head><body>B</body></html>".into(),
639                inner_text: Some("B".into()),
640                layout_json: Some("[]".into()),
641                visibility_json: Some("[]".into()),
642                screenshot: Some(synthetic_image(2, 2)),
643                js_result: Some("42".into()),
644                accessibility_tree: Some("{}".into()),
645                a11y: None,
646                console_messages: vec![bridge::ConsoleMessage {
647                    level: bridge::ConsoleLevel::Log,
648                    message: "x".into(),
649                }],
650            };
651            let page = Page::from_servo(sp);
652            assert_eq!(page.html, "<html><head><title>T</title></head><body>B</body></html>");
653            assert_eq!(page.inner_text, "B");
654            assert_eq!(page.title.as_deref(), Some("T"));
655            assert_eq!(page.layout_json.as_deref(), Some("[]"));
656            assert_eq!(page.js_result.as_deref(), Some("42"));
657            assert_eq!(page.accessibility_tree.as_deref(), Some("{}"));
658            assert_eq!(page.console_messages.len(), 1);
659            assert!(page.screenshot_png().is_some());
660            assert!(page.extracted.is_none());
661        }
662    }
663}
servo_fetch/fetch.rs

servo_fetch/
fetch.rs
