crw-extract 0.3.5

HTML extraction and markdown conversion engine for the CRW web scraper
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332

use scraper::{Html, Selector};

/// When a priority selector is "too broad" (>90% of body), drill down into it
/// to find a narrower content element.
fn find_content_within(parent_el: &scraper::ElementRef, parent_len: usize) -> Option<String> {
    let inner_selectors = [
        ".main-page-content",
        ".article-content",
        ".post-content",
        ".entry-content",
        ".content-body",
        ".article-body",
        "[itemprop=\"articleBody\"]",
        "[itemprop=\"text\"]",
        ".mw-parser-output",
        "#mw-content-text",
        "#content",
        ".content",
        "article", // nested article inside broad main
    ];

    let mut best: Option<(String, f64)> = None;
    for sel_str in &inner_selectors {
        if let Ok(sel) = Selector::parse(sel_str) {
            for el in parent_el.select(&sel) {
                let content = el.html();
                if content.len() < 200 {
                    continue;
                }
                // Skip if still too broad relative to parent
                if content.len() as f64 / parent_len as f64 > 0.85 {
                    continue;
                }
                let score = text_density(&content) * (content.len() as f64).ln();
                if best.as_ref().is_none_or(|(_, s)| score > *s) {
                    best = Some((content, score));
                }
            }
        }
    }
    best.map(|(c, _)| c)
}

/// Extract the "main content" element from HTML.
///
/// Uses text-density scoring across candidate selectors to pick the richest element.
/// Falls back to the `<body>` if no scored candidate is found.
pub fn extract_main_content(html: &str) -> String {
    let document = Html::parse_document(html);

    // Priority candidates in order: well-known semantic selectors first.
    let priority_selectors = ["article", "main", "[role=\"main\"]"];

    // Compute body length once for ratio checks below.
    let body_len = Selector::parse("body")
        .ok()
        .and_then(|sel| document.select(&sel).next())
        .map(|b| b.html().len())
        .unwrap_or(html.len());

    // Try priority selectors first (no scoring needed — these are strong signals).
    for sel_str in &priority_selectors {
        if let Ok(sel) = Selector::parse(sel_str)
            && let Some(el) = document.select(&sel).next()
        {
            let content = el.html();
            // Only accept if the element has substantial text.
            if text_density(&content) > 0.1 && content.len() > 200 {
                // Skip if the element wraps nearly the entire document
                // (e.g. Wikipedia's <article> contains everything including sidebar).
                if body_len > 0 && content.len() as f64 / body_len as f64 > 0.9 {
                    // Try to find a narrower content element within the broad container.
                    if let Some(narrowed) = find_content_within(&el, content.len()) {
                        return narrowed;
                    }
                    continue; // Too broad — fall through to scoring
                }
                return content;
            }
        }
    }

    // Score all candidate selectors by text density and pick the best.
    let scored_selectors = [
        ".post-content",
        ".article-body",
        ".entry-content",
        ".article-content",
        ".post-body",
        ".story-body",
        ".content-body",
        "#main-content",
        "#article",
        "#content",
        ".content",
        ".main",
        "[itemprop=\"articleBody\"]",
        "[itemprop=\"text\"]",
        // MDN
        ".main-page-content",
        // StackOverflow
        ".js-post-body",
        ".s-prose",
        "#question",
        // Generic
        ".page-content",
        "#page-content",
        "[role=\"article\"]",
        // Wikipedia / MediaWiki
        ".mw-parser-output",
        "#mw-content-text",
        "#bodyContent",
        ".mw-body-content",
    ];

    let mut best: Option<(String, f64)> = None;
    for sel_str in &scored_selectors {
        if let Ok(sel) = Selector::parse(sel_str)
            && let Some(el) = document.select(&sel).next()
        {
            let content = el.html();
            if content.len() < 100 {
                continue;
            }
            // Skip selectors that wrap nearly the entire body (same as priority check).
            if body_len > 0 && content.len() as f64 / body_len as f64 > 0.9 {
                if let Some(narrowed) = find_content_within(&el, content.len()) {
                    return narrowed;
                }
                continue;
            }
            let score = text_density(&content) * (content.len() as f64).ln();
            if best.as_ref().is_none_or(|(_, s)| score > *s) {
                best = Some((content, score));
            }
        }
    }

    if let Some((content, _)) = best {
        return content;
    }

    // Last resort: return full body.
    if let Ok(sel) = Selector::parse("body")
        && let Some(body) = document.select(&sel).next()
    {
        return body.inner_html();
    }

    html.to_string()
}

/// Compute text-to-html ratio as a simple content density signal.
/// Returns a value in [0, 1]: higher = more text relative to markup.
fn text_density(html: &str) -> f64 {
    let doc = Html::parse_fragment(html);
    let text_len: usize = doc.root_element().text().map(|t| t.len()).sum();
    if html.is_empty() {
        return 0.0;
    }
    text_len as f64 / html.len() as f64
}

/// All extracted metadata from a page.
pub struct ExtractedMetadata {
    pub title: Option<String>,
    pub description: Option<String>,
    pub language: Option<String>,
    pub og_title: Option<String>,
    pub og_description: Option<String>,
    pub og_image: Option<String>,
    pub canonical_url: Option<String>,
}

/// Extract metadata (title, description, OG tags, canonical) from HTML.
pub fn extract_metadata(html: &str) -> ExtractedMetadata {
    let document = Html::parse_document(html);

    let title = select_text(&document, "title");

    let description = select_attr(&document, r#"meta[name="description"]"#, "content");

    let og_title = select_attr(&document, r#"meta[property="og:title"]"#, "content");
    let og_description = select_attr(&document, r#"meta[property="og:description"]"#, "content");
    let og_image = select_attr(&document, r#"meta[property="og:image"]"#, "content");

    let canonical_url = select_attr(&document, r#"link[rel="canonical"]"#, "href");

    // Extract language from <html lang="..."> attribute.
    let language = select_attr(&document, "html", "lang");

    ExtractedMetadata {
        title,
        description,
        language,
        og_title,
        og_description,
        og_image,
        canonical_url,
    }
}

fn select_text(doc: &Html, selector: &str) -> Option<String> {
    Selector::parse(selector)
        .ok()
        .and_then(|sel| doc.select(&sel).next())
        .map(|el| el.text().collect::<String>().trim().to_string())
        .filter(|s| !s.is_empty())
}

fn select_attr(doc: &Html, selector: &str, attr: &str) -> Option<String> {
    Selector::parse(selector)
        .ok()
        .and_then(|sel| doc.select(&sel).next())
        .and_then(|el| el.value().attr(attr).map(|s| s.to_string()))
        .filter(|s| !s.is_empty())
}

/// Extract all links from HTML.
pub fn extract_links(html: &str, base_url: &str) -> Vec<String> {
    let document = Html::parse_document(html);
    let sel = match Selector::parse("a[href]") {
        Ok(s) => s,
        Err(_) => return vec![],
    };

    let base = url::Url::parse(base_url).ok();

    document
        .select(&sel)
        .filter_map(|el| {
            let href = el.value().attr("href")?;
            if href.starts_with('#')
                || href.starts_with("javascript:")
                || href.starts_with("mailto:")
                || href.starts_with("data:")
                || href.starts_with("tel:")
                || href.starts_with("blob:")
            {
                return None;
            }
            if let Some(base) = &base {
                base.join(href).ok().map(|u| u.to_string())
            } else if href.starts_with("http") {
                Some(href.to_string())
            } else {
                None
            }
        })
        .collect()
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn extracts_article_content() {
        let html = r#"<html><body><nav>Nav</nav><article><p>Main content</p></article><footer>Foot</footer></body></html>"#;
        let content = extract_main_content(html);
        assert!(content.contains("Main content"));
    }

    #[test]
    fn extracts_title_and_description() {
        let html = r#"<html><head><title>Test Page</title><meta name="description" content="A test"></head><body></body></html>"#;
        let meta = extract_metadata(html);
        assert_eq!(meta.title.unwrap(), "Test Page");
        assert_eq!(meta.description.unwrap(), "A test");
    }

    #[test]
    fn extracts_og_metadata() {
        let html = r#"<html><head>
            <meta property="og:title" content="OG Title">
            <meta property="og:description" content="OG Desc">
            <meta property="og:image" content="https://img.com/pic.jpg">
            <link rel="canonical" href="https://example.com/canonical">
        </head><body></body></html>"#;
        let meta = extract_metadata(html);
        assert_eq!(meta.og_title.unwrap(), "OG Title");
        assert_eq!(meta.og_description.unwrap(), "OG Desc");
        assert_eq!(meta.og_image.unwrap(), "https://img.com/pic.jpg");
        assert_eq!(meta.canonical_url.unwrap(), "https://example.com/canonical");
    }

    #[test]
    fn skips_broad_article_picks_mw_parser_output() {
        // Simulate Wikipedia structure: <article> wraps everything,
        // but .mw-parser-output is the real content.
        let filler = "x".repeat(500);
        let html = format!(
            r#"<html><body>
            <article>
              <div id="mw-navigation">{filler}</div>
              <div id="content" role="main">
                <div id="bodyContent">
                  <div id="mw-content-text">
                    <div class="mw-parser-output">
                      <p>This is the real Wikipedia article content about web scraping. {filler}</p>
                    </div>
                  </div>
                </div>
              </div>
              <div class="catlinks">{filler}</div>
            </article>
            </body></html>"#
        );
        let content = extract_main_content(&html);
        assert!(
            content.contains("real Wikipedia article content"),
            "Should extract .mw-parser-output content"
        );
        // Should NOT contain the navigation or catlinks filler
        assert!(
            !content.contains("mw-navigation"),
            "Should not include navigation div"
        );
    }

    #[test]
    fn extracts_links() {
        let html = r##"<html><body><a href="/page1">P1</a><a href="https://other.com">O</a><a href="#top">T</a></body></html>"##;
        let links = extract_links(html, "https://example.com");
        assert_eq!(links.len(), 2);
        assert!(links.contains(&"https://example.com/page1".to_string()));
        assert!(
            links.contains(&"https://other.com".to_string())
                || links.contains(&"https://other.com/".to_string())
        );
    }
}