deltachat/
dehtml.rs

1//! De-HTML.
2//!
3//! A module to remove HTML tags from the email text
4
5use std::io::BufRead;
6use std::sync::LazyLock;
7
8use quick_xml::{
9    Reader,
10    errors::Error as QuickXmlError,
11    events::{BytesEnd, BytesStart, BytesText},
12};
13
14use crate::simplify::{SimplifiedText, simplify_quote};
15
16#[derive(Default)]
17struct Dehtml {
18    strbuilder: String,
19    quote: String,
20    add_text: AddText,
21    last_href: Option<String>,
22    /// GMX wraps a quote in `<div name="quote">`. After a `<div name="quote">`, this count is
23    /// increased at each `<div>` and decreased at each `</div>`. This way we know when the quote ends.
24    /// If this is > `0`, then we are inside a `<div name="quote">`
25    divs_since_quote_div: u32,
26    /// Everything between `<div name="quote">` and `<div name="quoted-content">` is usually metadata
27    /// If this is > `0`, then we are inside a `<div name="quoted-content">`.
28    divs_since_quoted_content_div: u32,
29    /// `<div class="header-protection-legacy-display">` elements should be omitted, see
30    /// <https://www.rfc-editor.org/rfc/rfc9788.html#section-4.5.3.3>.
31    divs_since_hp_legacy_display: u32,
32    /// All-Inkl just puts the quote into `<blockquote> </blockquote>`. This count is
33    /// increased at each `<blockquote>` and decreased at each `</blockquote>`.
34    blockquotes_since_blockquote: u32,
35}
36
37impl Dehtml {
38    /// Returns true if HTML parser is currently inside the quote.
39    fn is_quote(&self) -> bool {
40        self.divs_since_quoted_content_div > 0 || self.blockquotes_since_blockquote > 0
41    }
42
43    /// Returns the buffer where the text should be written.
44    ///
45    /// If the parser is inside the quote, returns the quote buffer.
46    fn get_buf(&mut self) -> &mut String {
47        if self.is_quote() {
48            &mut self.quote
49        } else {
50            &mut self.strbuilder
51        }
52    }
53
54    fn get_add_text(&self) -> AddText {
55        // Everything between `<div name="quoted">` and `<div name="quoted_content">` is
56        // metadata which we don't want.
57        if self.divs_since_quote_div > 0 && self.divs_since_quoted_content_div == 0
58            || self.divs_since_hp_legacy_display > 0
59        {
60            AddText::No
61        } else {
62            self.add_text
63        }
64    }
65}
66
67#[derive(Debug, Default, PartialEq, Clone, Copy)]
68enum AddText {
69    /// Inside `<script>`, `<style>` and similar tags
70    /// which contents should not be displayed.
71    No,
72
73    #[default]
74    YesRemoveLineEnds,
75
76    /// Inside `<pre>`.
77    YesPreserveLineEnds,
78}
79
80pub(crate) fn dehtml(buf: &str) -> Option<SimplifiedText> {
81    let (s, quote) = dehtml_quick_xml(buf);
82    if !s.trim().is_empty() {
83        let text = dehtml_cleanup(s);
84        let top_quote = if !quote.trim().is_empty() {
85            Some(dehtml_cleanup(simplify_quote(&quote).0))
86        } else {
87            None
88        };
89        return Some(SimplifiedText {
90            text,
91            top_quote,
92            ..Default::default()
93        });
94    }
95    let s = dehtml_manually(buf);
96    if !s.trim().is_empty() {
97        let text = dehtml_cleanup(s);
98        return Some(SimplifiedText {
99            text,
100            ..Default::default()
101        });
102    }
103    None
104}
105
106fn dehtml_cleanup(mut text: String) -> String {
107    text.retain(|c| c != '\r');
108    let lines = text.trim().split('\n');
109    let mut text = String::new();
110    let mut linebreak = false;
111    for line in lines {
112        if line.chars().all(char::is_whitespace) {
113            linebreak = true;
114        } else {
115            if !text.is_empty() {
116                text += "\n";
117                if linebreak {
118                    text += "\n";
119                }
120            }
121            text += line.trim_end();
122            linebreak = false;
123        }
124    }
125    text
126}
127
128fn dehtml_quick_xml(buf: &str) -> (String, String) {
129    let buf = buf.trim().trim_start_matches("<!doctype html>");
130
131    let mut dehtml = Dehtml {
132        strbuilder: String::with_capacity(buf.len()),
133        ..Default::default()
134    };
135
136    let mut reader = quick_xml::Reader::from_str(buf);
137    reader.config_mut().check_end_names = false;
138
139    let mut buf = Vec::new();
140    let mut char_buf = String::with_capacity(4);
141
142    loop {
143        match reader.read_event_into(&mut buf) {
144            Ok(quick_xml::events::Event::Start(ref e)) => {
145                dehtml_starttag_cb(e, &mut dehtml, &reader)
146            }
147            Ok(quick_xml::events::Event::End(ref e)) => dehtml_endtag_cb(e, &mut dehtml),
148            Ok(quick_xml::events::Event::Text(ref e)) => dehtml_text_cb(e, &mut dehtml),
149            Ok(quick_xml::events::Event::CData(e)) => {
150                str_cb(&String::from_utf8_lossy(&e as &[_]), &mut dehtml)
151            }
152            Ok(quick_xml::events::Event::Empty(ref e)) => {
153                // Handle empty tags as a start tag immediately followed by end tag.
154                // For example, `<p/>` is treated as `<p></p>`.
155                dehtml_starttag_cb(e, &mut dehtml, &reader);
156                dehtml_endtag_cb(
157                    &BytesEnd::new(String::from_utf8_lossy(e.name().as_ref())),
158                    &mut dehtml,
159                );
160            }
161            Ok(quick_xml::events::Event::GeneralRef(ref e)) => {
162                match e.resolve_char_ref() {
163                    Err(err) => eprintln!(
164                        "resolve_char_ref() error at position {}: {:?}",
165                        reader.buffer_position(),
166                        err,
167                    ),
168                    Ok(Some(ch)) => {
169                        char_buf.clear();
170                        char_buf.push(ch);
171                        str_cb(&char_buf, &mut dehtml);
172                    }
173                    Ok(None) => {
174                        let event_str = String::from_utf8_lossy(e);
175                        if let Some(s) = quick_xml::escape::resolve_html5_entity(&event_str) {
176                            str_cb(s, &mut dehtml);
177                        } else {
178                            // Nonstandard entity. Add escaped.
179                            str_cb(&format!("&{event_str};"), &mut dehtml);
180                        }
181                    }
182                }
183            }
184            Err(QuickXmlError::IllFormed(_)) => {
185                // This is probably not HTML at all and should be left as is.
186                str_cb(&String::from_utf8_lossy(&buf), &mut dehtml);
187            }
188            Err(e) => {
189                eprintln!(
190                    "Parse html error: Error at position {}: {:?}",
191                    reader.buffer_position(),
192                    e
193                );
194            }
195            Ok(quick_xml::events::Event::Eof) => break,
196            _ => (),
197        }
198        buf.clear();
199    }
200
201    (dehtml.strbuilder, dehtml.quote)
202}
203
204fn dehtml_text_cb(event: &BytesText, dehtml: &mut Dehtml) {
205    if dehtml.get_add_text() == AddText::YesPreserveLineEnds
206        || dehtml.get_add_text() == AddText::YesRemoveLineEnds
207    {
208        let event = event as &[_];
209        let event_str = std::str::from_utf8(event).unwrap_or_default();
210        str_cb(event_str, dehtml);
211    }
212}
213
214fn str_cb(event_str: &str, dehtml: &mut Dehtml) {
215    static LINE_RE: LazyLock<regex::Regex> =
216        LazyLock::new(|| regex::Regex::new(r"(\r?\n)+").unwrap());
217
218    let add_text = dehtml.get_add_text();
219    if add_text == AddText::YesRemoveLineEnds {
220        // Replace all line ends with spaces.
221        // E.g. `\r\n\r\n` is replaced with one space.
222        let event_str = LINE_RE.replace_all(event_str, " ");
223
224        // Add a space if `event_str` starts with a space
225        // and there is no whitespace at the end of the buffer yet.
226        // Trim the rest of leading whitespace from `event_str`.
227        let buf = dehtml.get_buf();
228        if !buf.ends_with(' ') && !buf.ends_with('\n') && event_str.starts_with(' ') {
229            *buf += " ";
230        }
231
232        *buf += event_str.trim_start();
233    } else if add_text == AddText::YesPreserveLineEnds {
234        *dehtml.get_buf() += LINE_RE.replace_all(event_str, "\n").as_ref();
235    }
236}
237
238#[expect(clippy::arithmetic_side_effects)]
239fn dehtml_endtag_cb(event: &BytesEnd, dehtml: &mut Dehtml) {
240    let tag = String::from_utf8_lossy(event.name().as_ref())
241        .trim()
242        .to_lowercase();
243
244    match tag.as_str() {
245        "style" | "script" | "title" | "pre" => {
246            *dehtml.get_buf() += "\n\n";
247            dehtml.add_text = AddText::YesRemoveLineEnds;
248        }
249        "div" => {
250            pop_tag(&mut dehtml.divs_since_quote_div);
251            pop_tag(&mut dehtml.divs_since_quoted_content_div);
252            pop_tag(&mut dehtml.divs_since_hp_legacy_display);
253
254            *dehtml.get_buf() += "\n\n";
255            dehtml.add_text = AddText::YesRemoveLineEnds;
256        }
257        "a" => {
258            if let Some(ref last_href) = dehtml.last_href.take() {
259                let buf = dehtml.get_buf();
260                if buf.ends_with('[') {
261                    buf.truncate(buf.len() - 1);
262                } else {
263                    *buf += "](";
264                    *buf += last_href;
265                    *buf += ")";
266                }
267            }
268        }
269        "b" | "strong" if dehtml.get_add_text() != AddText::No => {
270            *dehtml.get_buf() += "*";
271        }
272        "i" | "em" if dehtml.get_add_text() != AddText::No => {
273            *dehtml.get_buf() += "_";
274        }
275        "blockquote" => pop_tag(&mut dehtml.blockquotes_since_blockquote),
276        _ => {}
277    }
278}
279
280#[expect(clippy::arithmetic_side_effects)]
281fn dehtml_starttag_cb<B: std::io::BufRead>(
282    event: &BytesStart,
283    dehtml: &mut Dehtml,
284    reader: &quick_xml::Reader<B>,
285) {
286    let tag = String::from_utf8_lossy(event.name().as_ref())
287        .trim()
288        .to_lowercase();
289
290    match tag.as_str() {
291        "p" | "table" | "td" => {
292            if !dehtml.strbuilder.is_empty() {
293                *dehtml.get_buf() += "\n\n";
294            }
295            dehtml.add_text = AddText::YesRemoveLineEnds;
296        }
297        #[rustfmt::skip]
298        "div" => {
299            maybe_push_tag(event, reader, "quote", &mut dehtml.divs_since_quote_div);
300            maybe_push_tag(event, reader, "quoted-content", &mut dehtml.divs_since_quoted_content_div);
301            maybe_push_tag(event, reader, "header-protection-legacy-display",
302                &mut dehtml.divs_since_hp_legacy_display);
303
304            *dehtml.get_buf() += "\n\n";
305            dehtml.add_text = AddText::YesRemoveLineEnds;
306        }
307        "br" => {
308            *dehtml.get_buf() += "\n";
309            dehtml.add_text = AddText::YesRemoveLineEnds;
310        }
311        "style" | "script" | "title" => {
312            dehtml.add_text = AddText::No;
313        }
314        "pre" => {
315            *dehtml.get_buf() += "\n\n";
316            dehtml.add_text = AddText::YesPreserveLineEnds;
317        }
318        "a" => {
319            if let Some(href) = event
320                .html_attributes()
321                .filter_map(|attr| attr.ok())
322                .find(|attr| {
323                    String::from_utf8_lossy(attr.key.as_ref())
324                        .trim()
325                        .to_lowercase()
326                        == "href"
327                })
328            {
329                let href = href
330                    .decode_and_unescape_value(reader.decoder())
331                    .unwrap_or_default()
332                    .to_string();
333
334                if !href.is_empty() {
335                    dehtml.last_href = Some(href);
336                    *dehtml.get_buf() += "[";
337                }
338            }
339        }
340        "b" | "strong" if dehtml.get_add_text() != AddText::No => {
341            *dehtml.get_buf() += "*";
342        }
343        "i" | "em" if dehtml.get_add_text() != AddText::No => {
344            *dehtml.get_buf() += "_";
345        }
346        "blockquote" => dehtml.blockquotes_since_blockquote += 1,
347        _ => {}
348    }
349}
350
351/// In order to know when a specific tag is closed, we need to count the opening and closing tags.
352/// The `counts`s are stored in the `Dehtml` struct.
353#[expect(clippy::arithmetic_side_effects)]
354fn pop_tag(count: &mut u32) {
355    if *count > 0 {
356        *count -= 1;
357    }
358}
359
360/// In order to know when a specific tag is closed, we need to count the opening and closing tags.
361/// The `counts`s are stored in the `Dehtml` struct.
362#[expect(clippy::arithmetic_side_effects)]
363fn maybe_push_tag(
364    event: &BytesStart,
365    reader: &Reader<impl BufRead>,
366    tag_name: &str,
367    count: &mut u32,
368) {
369    if *count > 0 || tag_contains_attr(event, reader, tag_name) {
370        *count += 1;
371    }
372}
373
374fn tag_contains_attr(event: &BytesStart, reader: &Reader<impl BufRead>, name: &str) -> bool {
375    event.attributes().any(|r| {
376        r.map(|a| {
377            a.decode_and_unescape_value(reader.decoder())
378                .map(|v| v == name)
379                .unwrap_or(false)
380        })
381        .unwrap_or(false)
382    })
383}
384
385pub fn dehtml_manually(buf: &str) -> String {
386    // Just strip out everything between "<" and ">"
387    let mut strbuilder = String::new();
388    let mut show_next_chars = true;
389    for c in buf.chars() {
390        match c {
391            '<' => show_next_chars = false,
392            '>' => show_next_chars = true,
393            _ => {
394                if show_next_chars {
395                    strbuilder.push(c)
396                }
397            }
398        }
399    }
400    strbuilder
401}
402
403#[cfg(test)]
404mod tests {
405    use super::*;
406
407    #[test]
408    fn test_dehtml() {
409        let cases = vec![
410            (
411                "<a href='https://example.com'> Foo </a>",
412                "[ Foo ](https://example.com)",
413            ),
414            ("<b> bar </b>", "* bar *"),
415            ("<i>foo</i>", "_foo_"),
416            ("<b> bar <i> foo", "* bar _ foo"),
417            ("&amp; bar", "& bar"),
418            // Despite missing ', this should be shown:
419            ("<a href='/foo.png>Hi</a> ", "Hi"),
420            ("No link: <a href='https://get.delta.chat/'/>", "No link:"),
421            (
422                "No link: <a href='https://get.delta.chat/'></a>",
423                "No link:",
424            ),
425            ("<!doctype html>\n<b>fat text</b>", "*fat text*"),
426            // Invalid html (at least DC should show the text if the html is invalid):
427            ("<!some invalid html code>\n<b>some text</b>", "some text"),
428        ];
429        for (input, output) in cases {
430            assert_eq!(dehtml(input).unwrap().text, output);
431        }
432        let none_cases = vec!["<html> </html>", ""];
433        for input in none_cases {
434            assert_eq!(dehtml(input), None);
435        }
436    }
437
438    #[test]
439    fn test_dehtml_parse_br() {
440        let html = "line1<br>line2";
441        let plain = dehtml(html).unwrap().text;
442        assert_eq!(plain, "line1\nline2");
443
444        let html = "line1<br> line2";
445        let plain = dehtml(html).unwrap().text;
446        assert_eq!(plain, "line1\nline2");
447
448        let html = "line1  <br><br> line2";
449        let plain = dehtml(html).unwrap().text;
450        assert_eq!(plain, "line1\n\nline2");
451
452        let html = "\r\r\nline1<br>\r\n\r\n\r\rline2<br/>line3\n\r";
453        let plain = dehtml(html).unwrap().text;
454        assert_eq!(plain, "line1\nline2\nline3");
455    }
456
457    #[test]
458    fn test_dehtml_parse_span() {
459        assert_eq!(dehtml("<span>Foo</span>bar").unwrap().text, "Foobar");
460        assert_eq!(dehtml("<span>Foo</span> bar").unwrap().text, "Foo bar");
461        assert_eq!(dehtml("<span>Foo </span>bar").unwrap().text, "Foo bar");
462        assert_eq!(dehtml("<span>Foo</span>\nbar").unwrap().text, "Foo bar");
463        assert_eq!(dehtml("\n<span>Foo</span> bar").unwrap().text, "Foo bar");
464        assert_eq!(dehtml("<span>Foo</span>\n\nbar").unwrap().text, "Foo bar");
465        assert_eq!(dehtml("Foo\n<span>bar</span>").unwrap().text, "Foo bar");
466        assert_eq!(dehtml("Foo<span>\nbar</span>").unwrap().text, "Foo bar");
467    }
468
469    #[test]
470    fn test_dehtml_parse_p() {
471        let html = "<p>Foo</p><p>Bar</p>";
472        let plain = dehtml(html).unwrap().text;
473        assert_eq!(plain, "Foo\n\nBar");
474
475        let html = "<p>Foo<p>Bar";
476        let plain = dehtml(html).unwrap().text;
477        assert_eq!(plain, "Foo\n\nBar");
478
479        let html = "<p>Foo</p><p>Bar<p>Baz";
480        let plain = dehtml(html).unwrap().text;
481        assert_eq!(plain, "Foo\n\nBar\n\nBaz");
482    }
483
484    #[test]
485    fn test_dehtml_parse_href() {
486        let html = "<a href=url>text</a>";
487        let plain = dehtml(html).unwrap().text;
488
489        assert_eq!(plain, "[text](url)");
490    }
491
492    #[test]
493    fn test_dehtml_case_sensitive_link() {
494        let html = "<html><A HrEf=\"https://foo.bar/Data\">case in URLs matter</A></html>";
495        let plain = dehtml(html).unwrap().text;
496        assert_eq!(plain, "[case in URLs matter](https://foo.bar/Data)");
497    }
498
499    #[test]
500    fn test_dehtml_bold_text() {
501        let html = "<!DOCTYPE name [<!DOCTYPE ...>]><!-- comment -->text <b><?php echo ... ?>bold</b><![CDATA[<>]]>";
502        let plain = dehtml(html).unwrap().text;
503
504        assert_eq!(plain, "text *bold*<>");
505    }
506
507    #[test]
508    fn test_dehtml_html_encoded() {
509        let html = "&lt;&gt;&quot;&apos;&amp; &auml;&Auml;&ouml;&Ouml;&uuml;&Uuml;&szlig; foo&AElig;&ccedil;&Ccedil; &diams;&lrm;&rlm;&zwnj;&noent;&zwj;";
510
511        let plain = dehtml(html).unwrap().text;
512
513        assert_eq!(
514            plain,
515            "<>\"\'& äÄöÖüÜß fooÆçÇ \u{2666}\u{200e}\u{200f}\u{200c}&noent;\u{200d}"
516        );
517    }
518
519    #[test]
520    fn test_unclosed_tags() {
521        let input = r##"
522        <!DOCTYPE HTML PUBLIC '-//W3C//DTD HTML 4.01 Transitional//EN'
523        'http://www.w3.org/TR/html4/loose.dtd'>
524        <html>
525        <head>
526        <title>Hi</title>
527        <meta http-equiv='Content-Type' content='text/html; charset=iso-8859-1'>						
528        </head>
529        <body>
530        lots of text
531        </body>
532        </html>
533        "##;
534        let txt = dehtml(input).unwrap();
535        assert_eq!(txt.text.trim(), "lots of text");
536    }
537
538    #[test]
539    fn test_pre_tag() {
540        let input = "<html><pre>\ntwo\nlines\n</pre></html>";
541        let txt = dehtml(input).unwrap();
542        assert_eq!(txt.text.trim(), "two\nlines");
543    }
544
545    #[test]
546    fn test_hp_legacy_display() {
547        let input = r#"
548<html><head><title></title></head><body>
549<div class="header-protection-legacy-display">
550<pre>Subject: Dinner plans</pre>
551</div>
552<p>
553Let's meet at Rama's Roti Shop at 8pm and go to the park
554from there.
555</p>
556</body>
557</html>
558        "#;
559        let txt = dehtml(input).unwrap();
560        assert_eq!(
561            txt.text.trim(),
562            "Let's meet at Rama's Roti Shop at 8pm and go to the park from there."
563        );
564    }
565
566    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
567    async fn test_quote_div() {
568        let input = include_str!("../test-data/message/gmx-quote-body.eml");
569        let dehtml = dehtml(input).unwrap();
570        let SimplifiedText {
571            text,
572            is_forwarded,
573            is_cut,
574            top_quote,
575            footer,
576        } = dehtml;
577        assert_eq!(text, "Test");
578        assert_eq!(is_forwarded, false);
579        assert_eq!(is_cut, false);
580        assert_eq!(top_quote.as_deref(), Some("test"));
581        assert_eq!(footer, None);
582    }
583
584    #[test]
585    fn test_spaces() {
586        let input = include_str!("../test-data/spaces.html");
587        let txt = dehtml(input).unwrap();
588        assert_eq!(
589            txt.text,
590            "Welcome back to Strolling!\n\nHey there,\n\nWelcome back! Use this link to securely sign in to your Strolling account:\n\nSign in to Strolling\n\nFor your security, the link will expire in 24 hours time.\n\nSee you soon!\n\nYou can also copy & paste this URL into your browser:\n\nhttps://strolling.rosano.ca/members/?token=XXX&action=signin&r=https%3A%2F%2Fstrolling.rosano.ca%2F\n\nIf you did not make this request, you can safely ignore this email.\n\nThis message was sent from [strolling.rosano.ca](https://strolling.rosano.ca/) to [alice@example.org](mailto:alice@example.org)"
591        );
592    }
593}
deltachat/dehtml.rs

deltachat/
dehtml.rs
