crw-renderer 0.13.4

HTTP and CDP browser rendering 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
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
use regex::Regex;
use std::sync::LazyLock;

/// Akamai bot-manager reference ID: `Reference #<digits>.<hex>.<digits>.<hex>`.
/// Used only in [`looks_like_vendor_block`] — see the Akamai arm there.
static AKAMAI_REF_RE: LazyLock<Regex> = LazyLock::new(|| {
    Regex::new(r"Reference #\d+\.[0-9a-f]+\.\d+\.[0-9a-f]+").expect("static regex")
});

/// Heuristic: does the HTML look like an SPA shell that needs JS rendering?
pub fn needs_js_rendering(html: &str) -> bool {
    // Check up to 500KB — some pages have huge <head> sections (CSS, preloaded data)
    // and the <body> may start well beyond 50KB.
    let check_len = html.len().min(500_000);
    let lower = html[..check_len].to_lowercase();
    let body_len = extract_body_text_len(&lower);

    // Very short body text + presence of JS framework indicators.
    // After stripping script/style, most SPA shells have very little actual text.
    if body_len < 200 {
        let spa_indicators = [
            "id=\"root\"",
            "id=\"app\"",
            "id=\"__next\"",
            "id=\"__nuxt\"",
            "id=\"__gatsby\"",
            "id=\"svelte\"",
            "ng-app",
            "data-reactroot",
            "data-reactid",
            "data-remix-run",
            "data-sveltekit",
            "data-astro-",
            "<script src",
            "window.__initial_state__",
            "__next_data__",
            "__nuxt__",
            "__sveltekit_data",
            "window.__remixcontext",
            "window.__astro",
            "gatsby-focus-wrapper",
        ];
        if spa_indicators.iter().any(|ind| lower.contains(ind)) {
            return true;
        }
    }

    // Noscript tag with meaningful content suggests JS is needed.
    if lower.contains("<noscript>") && lower.contains("enable javascript") {
        return true;
    }

    // Framer / Webflow / other site builder markers (often fully JS-rendered)
    if body_len < 500 {
        let builder_indicators = [
            "framerusercontent.com",
            "webflow.io",
            "wixsite.com",
            "squarespace.com/universal",
        ];
        if builder_indicators.iter().any(|ind| lower.contains(ind)) {
            return true;
        }
    }

    // Bundler-heavy modern SPA: short body + many script tags. Catches sites
    // that don't expose a recognizable framework marker but ship most of their
    // content via client-side hydration. Threshold is conservative (5+ scripts,
    // body <1000 chars) — 3 scripts is a normal load (analytics + ads + a font
    // loader) on minimal static pages, so we require more before escalating.
    if body_len < 1000 {
        let script_count = lower.matches("<script").count();
        if script_count >= 5 {
            return true;
        }
        let storybook_indicators = [
            "id=\"storybook-root\"",
            "id=\"storybook-docs\"",
            "__storybook",
            "?path=/docs/",
            "/iframe.html",
        ];
        if storybook_indicators.iter().any(|ind| lower.contains(ind)) {
            return true;
        }
    }

    false
}

/// Detect generic anti-bot interstitials (non-Cloudflare): tiny pages whose
/// visible body text consists of a "verifying you're human" / "security check"
/// message. Matched only on visible body text so a JS bundle containing one
/// of these strings cannot false-positive.
pub fn looks_like_generic_bot_wall(html: &str) -> bool {
    if html.len() > 80_000 {
        return false;
    }
    let lower = html.to_lowercase();
    let body_stripped = body_html_without_scripts_lower(&lower);
    let body_text = visible_text_from_stripped_html(&body_stripped);
    if body_text.chars().filter(|c| !c.is_whitespace()).count() > 600 {
        return false;
    }

    let phrases = [
        "performing security verification",
        "verify you are human",
        "checking your browser",
        "enable javascript and cookies",
        "security check",
        "access denied",
        "request blocked",
        // CloudFront / AWS WAF generic block page. Title is the giveaway;
        // body content varies (geo-block, WAF rule, distribution misconfig).
        // All variants render an identical 403 shell with these strings.
        "the request could not be satisfied",
        "generated by cloudfront",
        // Akamai / AWS-style geo-block phrasing — also surfaces on some
        // origin-side firewall pages that don't say "blocked" outright.
        "configured to block access",
    ];
    phrases.iter().any(|p| body_text.contains(p))
}

/// Vendor-specific anti-bot block markers. Returns the matched vendor name
/// for logging/metrics, or `None` when no vendor signature is found.
///
/// Markers are curated durable signatures (owned CDN domains, public SDK
/// identifiers, vendor brand strings) — chosen for low false-positive risk
/// and resistance to cosmetic vendor updates. Pair with the catch-all
/// [`looks_like_generic_bot_wall`] phrase list, which handles vendors that
/// haven't been signature-mapped yet.
///
/// Scans only the first 15KB; vendor block pages are small and put their
/// markers in `<head>` or early body. Pages over 200KB return `None` —
/// real content dwarfs vendor block shells.
pub fn looks_like_vendor_block(html: &str) -> Option<&'static str> {
    if html.len() > 200_000 {
        return None;
    }
    let head = &html[..html.len().min(15_000)];
    let lower_head = head.to_lowercase();

    // Cloudflare: challenge form with cf-managed token, error code span, or
    // challenge-platform JS loader. All three are unique to CF's anti-bot.
    if (lower_head.contains("challenge-form") && lower_head.contains("__cf_chl_f_tk="))
        || lower_head.contains("cf-error-code")
        || lower_head.contains("/cdn-cgi/challenge-platform/")
    {
        return Some("cloudflare");
    }

    // Akamai: bot-manager reference IDs follow `Reference #<hex>.<hex>.<hex>.<hex>`.
    // "Pardon Our Interruption" is the canonical block page headline.
    if lower_head.contains("pardon our interruption") || AKAMAI_REF_RE.is_match(head) {
        return Some("akamai");
    }

    // PerimeterX: window._pxAppId SDK assignment, owned captcha CDN.
    if lower_head.contains("window._pxappid =") || lower_head.contains("captcha.px-cdn.net") {
        return Some("perimeterx");
    }

    // DataDome: owned captcha delivery domain.
    if lower_head.contains("captcha-delivery.com") {
        return Some("datadome");
    }

    // Imperva / Incapsula: resource marker + incident ID phrasing.
    if lower_head.contains("_incapsula_resource") || lower_head.contains("incapsula incident id") {
        return Some("imperva");
    }

    // Sucuri: WAF block page brand string.
    if lower_head.contains("sucuri website firewall") {
        return Some("sucuri");
    }

    // Kasada: SDK signature.
    if lower_head.contains("kpsdk.scriptstart = kpsdk.now()") {
        return Some("kasada");
    }

    // CloudFront / AWS WAF (geo-block, distribution misconfig). Already
    // partly in the catch-all phrase list — capture as a vendor here for
    // telemetry split.
    if lower_head.contains("generated by cloudfront")
        || lower_head.contains("the request could not be satisfied")
    {
        return Some("cloudfront");
    }

    None
}

/// Returns true when an HTTP response yielded effectively no visible text in
/// the body (post script/style strip). Used by the renderer to decide whether
/// to escalate a "successful" HTTP fetch to JS rendering when no SPA marker
/// was recognized.
///
/// Distinct from [`needs_js_rendering`]: that one is a pre-fetch heuristic
/// on raw markup, looking for framework shells. This one is purely about
/// outcome — does the page have *any* content for an extractor to chew on.
pub fn looks_like_thin_html(html: &str) -> bool {
    let check_len = html.len().min(500_000);
    let lower = html[..check_len].to_lowercase();
    extract_body_text_len(&lower) < 200
}

/// Returns true when an extracted markdown is below the floor used by the
/// renderer to decide a fetch produced effectively no extractable content.
/// Pair with [`looks_like_thin_html`] for a full thin-content judgment.
pub fn is_thin_markdown(markdown_len: usize) -> bool {
    markdown_len < 100
}

/// Reason a rendered page is considered a failed render. Returned by
/// [`looks_like_failed_render`] so callers can include the cause in failover
/// warnings or telemetry.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum FailedRenderReason {
    /// Next.js error boundary HTML was injected into the document. Indicates
    /// the framework caught an unhandled exception during hydration or
    /// rendering — the page the user wanted is not present.
    NextJsClientError,
    /// React rendered its production "Minified React error" placeholder. Same
    /// failure class as Next.js, but framework-agnostic.
    ReactMinifiedError,
    /// Next.js root `<div id="__next">` is present but empty (no hydration
    /// took place). Distinct from a generic placeholder because it specifically
    /// indicates an SPA whose JS never executed.
    EmptyNextRoot,
}

impl FailedRenderReason {
    pub fn as_str(self) -> &'static str {
        match self {
            FailedRenderReason::NextJsClientError => "nextjs_client_error",
            FailedRenderReason::ReactMinifiedError => "react_minified_error",
            FailedRenderReason::EmptyNextRoot => "empty_next_root",
        }
    }
}

/// Detect framework-level render failures *in the HTML markup*. Only matches
/// DOM-specific markers (element ids, data attributes) — never visible body
/// text — to avoid false positives on pages that legitimately mention the
/// error string (e.g. a blog post about Next.js debugging).
///
/// Returns `None` when the page looks healthy.
pub fn looks_like_failed_render(html: &str) -> Option<FailedRenderReason> {
    // Bail out fast for very large pages — error boundary markup is small and
    // appears near the body root; scanning megabytes would cost more than it
    // gains. Most failed renders produce <30KB of HTML.
    if html.len() > 200_000 {
        return None;
    }
    let lower = html.to_lowercase();

    // Next.js App Router error boundary. Next renders an error UI with these
    // marker attributes when a client-side exception bubbles to the root.
    // Sources: next.js/packages/next/src/client/components/error-boundary.tsx
    if lower.contains("id=\"__next-error-") || lower.contains("data-nextjs-error") {
        return Some(FailedRenderReason::NextJsClientError);
    }

    // Next.js Pages Router error overlay (visible in dev) and the production
    // "Application error" fallback both render this id. The marker is a DOM
    // element id and only appears when Next chose to render its error path.
    if lower.contains("id=\"__next_error__\"") {
        return Some(FailedRenderReason::NextJsClientError);
    }

    // React production error: surfaces as a minified message with a numeric
    // code linking back to react.dev. Match the canonical anchor href since
    // that combination only exists when React rendered the error explainer.
    if lower.contains("https://react.dev/errors/")
        || lower.contains("https://reactjs.org/docs/error-decoder")
    {
        return Some(FailedRenderReason::ReactMinifiedError);
    }

    // Empty Next.js root shell: <div id="__next"></div> (or with whitespace).
    // The renderer returned the SSR shell but no hydration ran; the
    // user-visible content is missing.
    if let Some(start) = lower.find("id=\"__next\"") {
        let after_id = &lower[start..];
        if let Some(close) = after_id.find('>') {
            let tail = &after_id[close + 1..];
            if let Some(end) = tail.find("</div>") {
                let inner = tail[..end].trim();
                if inner.is_empty() {
                    return Some(FailedRenderReason::EmptyNextRoot);
                }
            }
        }
    }

    None
}

/// Check if rendered HTML is dominated by loading placeholders, spinners,
/// or chat-widget-only content. Used *after* JS rendering to detect cases
/// where the renderer returned early before the real content appeared
/// (common with slow React/Vite SPAs on underpowered renderers).
///
/// Markers are matched against *visible* body text only (tags and attributes
/// stripped) to avoid false positives from e.g. `<img alt="Loading...">` on
/// a page that actually has real content.
pub fn looks_like_loading_placeholder(html: &str) -> bool {
    // Bail out fast for large pages — real content dwarfs loading markers.
    if html.len() > 80_000 {
        return false;
    }
    let lower = html.to_lowercase();
    let body_stripped = body_html_without_scripts_lower(&lower);
    let body_text = visible_text_from_stripped_html(&body_stripped);
    let body_text_len = body_text.chars().filter(|c| !c.is_whitespace()).count();

    if body_text_len == 0 {
        return true;
    }

    // Short body + explicit loading text in VISIBLE text.
    if body_text_len < 400 {
        let loading_markers = [
            "loading...",
            "loading…",
            "please wait",
            "just a moment",
            "initializing",
            "preparing",
            "one moment",
        ];
        if loading_markers.iter().any(|m| body_text.contains(m)) {
            return true;
        }
    }

    // Very short body + spinner/loader DOM markers. Matched against body HTML
    // with <script>/<style> stripped, so inline JS like `'class="spinner"'`
    // does not trigger a false positive.
    if body_text_len < 200 {
        let spinner_markers = [
            "class=\"spinner",
            "class=\"loader",
            "class=\"loading",
            "class=\"preloader",
            "id=\"loader",
            "id=\"preloader",
            "aria-label=\"loading\"",
        ];
        if spinner_markers.iter().any(|m| body_stripped.contains(m)) {
            return true;
        }
    }

    false
}

/// Return the `<body>` of a lowercased HTML document with `<script>` and
/// `<style>` blocks removed. Remaining tags (and their attributes) are
/// preserved. Returns an empty string if no `<body>` is found.
fn body_html_without_scripts_lower(lower: &str) -> String {
    let body_start = lower
        .find("<body")
        .and_then(|i| lower[i..].find('>').map(|j| i + j + 1));
    let body_end = lower.rfind("</body>");

    let body = match (body_start, body_end) {
        (Some(start), Some(end)) if start < end => &lower[start..end],
        _ => return String::new(),
    };

    let stripped = strip_tag_blocks(body, "script");
    strip_tag_blocks(&stripped, "style")
}

/// Strip all HTML tags (open/close, with attributes) from an already
/// script/style-stripped HTML fragment. Whitespace is collapsed.
fn visible_text_from_stripped_html(stripped: &str) -> String {
    let mut text = String::with_capacity(stripped.len());
    let mut in_tag = false;
    let mut prev_ws = true;
    for ch in stripped.chars() {
        if ch == '<' {
            in_tag = true;
        } else if ch == '>' {
            in_tag = false;
        } else if !in_tag {
            if ch.is_whitespace() {
                if !prev_ws {
                    text.push(' ');
                    prev_ws = true;
                }
            } else {
                text.push(ch);
                prev_ws = false;
            }
        }
    }
    text
}

/// Rough estimate of non-whitespace text length inside `<body>` of a
/// lowercased HTML document. Returns `1000` as a "probably has content"
/// fallback if no `<body>` is found.
fn extract_body_text_len(lower: &str) -> usize {
    if !lower.contains("<body") {
        return 1000;
    }
    let stripped = body_html_without_scripts_lower(lower);
    visible_text_from_stripped_html(&stripped)
        .chars()
        .filter(|c| !c.is_whitespace())
        .count()
}

/// Remove all `<tag ...>...</tag>` blocks from HTML. The input is assumed
/// to be already lowercased (callers pass lowercased HTML).
fn strip_tag_blocks(html: &str, tag: &str) -> String {
    let mut result = String::with_capacity(html.len());
    let open = format!("<{}", tag);
    let close = format!("</{}>", tag);
    let mut remaining = html;

    while let Some(start) = remaining.find(&open) {
        result.push_str(&remaining[..start]);
        let after_open = &remaining[start..];
        if let Some(end) = after_open.find(&close) {
            remaining = &after_open[end + close.len()..];
        } else {
            remaining = "";
            break;
        }
    }
    result.push_str(remaining);
    result
}

// ── Cloudflare challenge detection ───────────────────────────────────

/// Detect a Cloudflare anti-bot challenge / interstitial in the response.
///
/// Strategy: a single weak marker is not enough — most marketing pages
/// reference Cloudflare somewhere. We require either a *strong* marker
/// (uniquely tied to the challenge interstitial) or a *combination* of
/// two weak markers. This keeps false positives low while still catching
/// the JS-challenge HTML LightPanda fails to solve.
///
/// Pair with [`is_cloudflare_mitigated_header`] which uses the
/// `cf-mitigated` response header — that signal is independent of body
/// content and is the most reliable indicator.
pub fn looks_like_cloudflare_challenge(html: &str) -> bool {
    if html.len() > 80_000 {
        return false;
    }
    let lower = html.to_lowercase();

    // Strong markers: appear ONLY on the interstitial.
    let strong = [
        "cf-browser-verification",
        "cf-challenge-running",
        "/cdn-cgi/challenge-platform/",
        "_cf_chl_opt",
        "__cf_chl_managed_tk__",
        "window._cf_chl_opt",
    ];
    if strong.iter().any(|m| lower.contains(m)) {
        return true;
    }

    // Weak markers: each can appear on legitimate Cloudflare-fronted pages.
    // "ray id:" + "cloudflare" co-occur on most CF-fronted error pages and
    // would false-positive a real page with a CF footer; require challenge-
    // specific phrasing instead.
    let weak = [
        "just a moment",
        "checking your browser",
        "attention required",
        "performance &amp; security by cloudflare",
        "performance & security by cloudflare",
    ];
    let weak_hits = weak.iter().filter(|m| lower.contains(*m)).count();
    weak_hits >= 2
}

/// Returns true when the `cf-mitigated` response header indicates the
/// request was challenged or blocked by Cloudflare. Independent of the
/// HTTP status code — Cloudflare may return 200 with this header set.
///
/// `header_value` is the raw header value (case-sensitive on the right
/// side; we lower-case here for safety).
pub fn is_cloudflare_mitigated_header(header_value: &str) -> bool {
    let lower = header_value.trim().to_ascii_lowercase();
    matches!(lower.as_str(), "challenge" | "block")
}

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

    #[test]
    fn detects_spa_shell() {
        let html = r#"<html><head></head><body><div id="root"></div><script src="/app.js"></script></body></html>"#;
        assert!(needs_js_rendering(html));
    }

    #[test]
    fn static_page_no_js_needed() {
        let html = r#"<html><body><article><h1>Hello World</h1><p>This is a long article with plenty of text content to read and enjoy. It has multiple paragraphs and lots of useful information.</p></article></body></html>"#;
        assert!(!needs_js_rendering(html));
    }

    #[test]
    fn detects_loading_placeholder_text() {
        let html =
            r#"<html><body><div><p>Loading...</p><p>Hi! Ask me anything.</p></div></body></html>"#;
        assert!(looks_like_loading_placeholder(html));
    }

    #[test]
    fn detects_spinner_only_body() {
        let html = r#"<html><body><div class="spinner"></div></body></html>"#;
        assert!(looks_like_loading_placeholder(html));
    }

    #[test]
    fn real_content_not_placeholder() {
        let html = r#"<html><body><article><h1>Welcome to my creative space</h1><p>Waqar Bin Abrar is a full stack developer specializing in MERN stack and Flutter apps, building scalable digital solutions for clients worldwide.</p><p>With years of experience delivering production applications, he combines technical expertise with design sensibility.</p></article></body></html>"#;
        assert!(!looks_like_loading_placeholder(html));
    }

    #[test]
    fn logo_alt_loading_on_real_page_not_placeholder() {
        // Regression: "Loading..." inside an img alt attribute must NOT trigger
        // placeholder detection when the page has real visible content.
        let html = r#"<html><body>
            <header><img alt="Loading..." src="/logo.png"/></header>
            <article>
                <h1>Software Engineering Blog</h1>
                <p>Thoughts on distributed systems, programming languages, and the craft of writing software that lasts. New posts weekly.</p>
                <p>This site covers topics from Rust ownership to Kubernetes operators.</p>
            </article>
        </body></html>"#;
        assert!(!looks_like_loading_placeholder(html));
    }

    #[test]
    fn empty_body_is_placeholder() {
        let html = r#"<html><body></body></html>"#;
        assert!(looks_like_loading_placeholder(html));
    }

    #[test]
    fn large_page_never_placeholder() {
        let filler = "x".repeat(100_000);
        let html = format!("<html><body><p>Loading...</p>{filler}</body></html>");
        assert!(!looks_like_loading_placeholder(&html));
    }

    #[test]
    fn detects_nextjs_app_router_error_boundary() {
        let html = r#"<html><body><div id="__next-error-0"><h2>Application error: a client-side exception has occurred.</h2></div></body></html>"#;
        assert_eq!(
            looks_like_failed_render(html),
            Some(FailedRenderReason::NextJsClientError)
        );
    }

    #[test]
    fn detects_nextjs_pages_router_error() {
        let html = r#"<html><body><div id="__next_error__">oops</div></body></html>"#;
        assert_eq!(
            looks_like_failed_render(html),
            Some(FailedRenderReason::NextJsClientError)
        );
    }

    #[test]
    fn detects_react_minified_error() {
        let html = r#"<html><body><a href="https://react.dev/errors/418">Minified React error #418</a></body></html>"#;
        assert_eq!(
            looks_like_failed_render(html),
            Some(FailedRenderReason::ReactMinifiedError)
        );
    }

    #[test]
    fn detects_legacy_react_error_decoder_url() {
        let html = r#"<html><body><a href="https://reactjs.org/docs/error-decoder.html?invariant=31">React</a></body></html>"#;
        assert_eq!(
            looks_like_failed_render(html),
            Some(FailedRenderReason::ReactMinifiedError)
        );
    }

    #[test]
    fn blog_post_about_error_is_not_failed_render() {
        // Regression: a blog post that *describes* the Next.js error must NOT
        // be flagged as a failed render. The string appears only in body text
        // (and not in a __next-error- element id), so the detector must let
        // it through.
        let html = r#"<html><body><article><h1>Debugging Next.js</h1>
            <p>When you see "Application error: a client-side exception has occurred",
            it usually means a hydration mismatch.</p>
            <pre><code>console.log('debug')</code></pre>
        </article></body></html>"#;
        assert!(looks_like_failed_render(html).is_none());
    }

    #[test]
    fn healthy_page_is_not_failed_render() {
        let html =
            r#"<html><body><main><h1>Hello</h1><p>Real content here.</p></main></body></html>"#;
        assert!(looks_like_failed_render(html).is_none());
    }

    #[test]
    fn huge_page_is_not_scanned() {
        // Pages over 200KB are exempt. Even with a marker that would normally
        // trigger, the function must short-circuit to None.
        let mut html = String::from(r#"<html><body><div id="__next-error-0"></div>"#);
        html.push_str(&"<p>filler</p>".repeat(20_000));
        html.push_str("</body></html>");
        assert!(html.len() > 200_000);
        assert!(looks_like_failed_render(&html).is_none());
    }

    #[test]
    fn cf_strong_marker_detected() {
        let html =
            r#"<html><body><div id="cf-browser-verification">Just a moment...</div></body></html>"#;
        assert!(looks_like_cloudflare_challenge(html));
    }

    #[test]
    fn cf_managed_token_detected() {
        let html = r#"<html><body><script>window._cf_chl_opt={cvId:'2'};</script></body></html>"#;
        assert!(looks_like_cloudflare_challenge(html));
    }

    #[test]
    fn cf_single_weak_marker_not_enough() {
        // A page that just mentions "Cloudflare" should not trigger.
        let html = r#"<html><body><article><h1>Why we use Cloudflare</h1><p>Performance benefits.</p></article></body></html>"#;
        assert!(!looks_like_cloudflare_challenge(html));
    }

    #[test]
    fn cf_two_weak_markers_trigger() {
        // Two challenge-specific phrases must co-occur. "ray id:" was
        // removed from the weak set because legitimate CF-fronted error
        // pages also include both "ray id" and "cloudflare".
        let html =
            r#"<html><body><h1>Just a moment...</h1><p>Checking your browser...</p></body></html>"#;
        assert!(looks_like_cloudflare_challenge(html));
    }

    #[test]
    fn cf_ray_id_alone_does_not_trigger() {
        // Pre-fix this would false-positive: "ray id" + "cloudflare" both
        // appear on benign CF-fronted pages.
        let html = r#"<html><body><h1>About</h1><p>Hosted via Cloudflare.</p><footer>Ray ID: abc123</footer></body></html>"#;
        assert!(!looks_like_cloudflare_challenge(html));
    }

    #[test]
    fn cf_mitigated_header_challenge() {
        assert!(is_cloudflare_mitigated_header("challenge"));
        assert!(is_cloudflare_mitigated_header(" CHALLENGE "));
        assert!(is_cloudflare_mitigated_header("block"));
    }

    #[test]
    fn cf_mitigated_header_other_values() {
        assert!(!is_cloudflare_mitigated_header(""));
        assert!(!is_cloudflare_mitigated_header("ok"));
        assert!(!is_cloudflare_mitigated_header("verified"));
    }

    #[test]
    fn cf_huge_page_not_scanned() {
        let mut html = String::from(r#"<html><body><div id="cf-browser-verification">"#);
        html.push_str(&"<p>x</p>".repeat(20_000));
        html.push_str("</div></body></html>");
        assert!(html.len() > 80_000);
        assert!(!looks_like_cloudflare_challenge(&html));
    }

    #[test]
    fn cloudfront_403_block_page_is_bot_wall() {
        // Real-world CloudFront geo-block page (americastire.com from EU egress).
        // Two strong markers: title and footer attribution. Engine must escalate
        // through the renderer chain (e.g. to chrome_proxy) instead of returning
        // this 403 shell as success.
        let html = r#"<html><head><title>ERROR: The request could not be satisfied</title></head>
            <body><h1>403 ERROR</h1>
            <h3>The request could not be satisfied.</h3>
            <p>The Amazon CloudFront distribution is configured to block access from your country.</p>
            <hr><i>Generated by cloudfront (CloudFront)</i></body></html>"#;
        assert!(looks_like_generic_bot_wall(html));
    }

    #[test]
    fn generic_403_with_block_phrasing_is_bot_wall() {
        // Origin-side / WAF block page that uses "configured to block access"
        // without naming a vendor — still a clear block signal.
        let html = r#"<html><body><h1>403</h1>
            <p>Our firewall is configured to block access from this region.</p></body></html>"#;
        assert!(looks_like_generic_bot_wall(html));
    }

    #[test]
    fn legitimate_blog_about_cloudfront_is_not_bot_wall() {
        // Regression guard: a long article about CloudFront must NOT trip the
        // bot-wall heuristic. visible body text length cap (600 chars) is the
        // existing safeguard — exceed it here to assert it still applies.
        let mut html = String::from(r#"<html><body><article>"#);
        html.push_str("<p>An article about CloudFront and how distributions are configured to block access by country. </p>".repeat(20).as_str());
        html.push_str("</article></body></html>");
        assert!(!looks_like_generic_bot_wall(&html));
    }

    #[test]
    fn vendor_cloudflare_challenge_form_detected() {
        let html = r#"<html><body><form class="challenge-form" action="/?__cf_chl_f_tk=abc123">
            </form></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("cloudflare"));
    }

    #[test]
    fn vendor_cloudflare_error_code_detected() {
        let html = r#"<html><body><span class="cf-error-code">1020</span></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("cloudflare"));
    }

    #[test]
    fn vendor_cloudflare_challenge_platform_detected() {
        let html = r#"<html><head><script src="/cdn-cgi/challenge-platform/h/g/orchestrate/chl_page/v1?ray=abc"></script></head></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("cloudflare"));
    }

    #[test]
    fn vendor_akamai_reference_id_detected() {
        let html = r#"<html><body><p>Access Denied</p>
            <p>Reference #18.2d351ab8.1557333295.a4e16ab</p></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("akamai"));
    }

    #[test]
    fn vendor_akamai_pardon_our_interruption_detected() {
        let html = r#"<html><body><h1>Pardon Our Interruption</h1>
            <p>As you were browsing, something about your browser made us think you were a bot.</p>
            </body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("akamai"));
    }

    #[test]
    fn vendor_perimeterx_pxappid_detected() {
        let html = r#"<html><head><script>window._pxAppId = 'PXabc123';</script></head></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("perimeterx"));
    }

    #[test]
    fn vendor_datadome_captcha_domain_detected() {
        let html = r#"<html><body><iframe src="https://geo.captcha-delivery.com/captcha/?initialCid=xyz"></iframe></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("datadome"));
    }

    #[test]
    fn vendor_imperva_incapsula_resource_detected() {
        let html = r#"<html><body><script src="/_Incapsula_Resource?SWJIYLWA=blah"></script></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("imperva"));
    }

    #[test]
    fn vendor_sucuri_firewall_brand_detected() {
        let html = r#"<html><body><h1>Sucuri WebSite Firewall - Access Denied</h1></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("sucuri"));
    }

    #[test]
    fn vendor_cloudfront_block_detected() {
        let html = r#"<html><head><title>ERROR: The request could not be satisfied</title></head>
            <body><h1>403 ERROR</h1>
            <hr><i>Generated by cloudfront (CloudFront)</i></body></html>"#;
        assert_eq!(looks_like_vendor_block(html), Some("cloudfront"));
    }

    #[test]
    fn vendor_legit_blog_about_cloudflare_is_none() {
        // A 50KB legit page mentioning Cloudflare but with no challenge markers
        // must NOT be flagged as a vendor block.
        let mut html = String::from("<html><body><article><h1>Why we picked Cloudflare</h1>");
        html.push_str(
            &"<p>Cloudflare gives us DDoS protection and a global anycast network.</p>".repeat(400),
        );
        html.push_str("</article></body></html>");
        assert!(html.len() > 15_000);
        assert!(looks_like_vendor_block(&html).is_none());
    }

    #[test]
    fn vendor_block_oversized_page_returns_none() {
        let big = "x".repeat(300_000);
        assert!(looks_like_vendor_block(&big).is_none());
    }

    #[test]
    fn vendor_block_clean_page_returns_none() {
        let html = r#"<html><body><main><h1>Hello</h1><p>Real content.</p></main></body></html>"#;
        assert!(looks_like_vendor_block(html).is_none());
    }

    #[test]
    fn spinner_class_in_script_body_ignored() {
        // class="spinner" inside a <script> block must not trigger spinner detection,
        // since scripts are stripped before text-length measurement.
        let html = r#"<html><body><article><h1>Real Article</h1><p>This is a real article with substantial content about the topic at hand, providing useful information.</p><script>const x = 'class="spinner"';</script></article></body></html>"#;
        assert!(!looks_like_loading_placeholder(html));
    }
}