memvid_core/reader/
docx.rs

1use std::io::{Cursor, Read};
2
3use quick_xml::Reader as XmlReader;
4use quick_xml::events::Event;
5use zip::ZipArchive;
6
7use crate::{
8    DocumentFormat, DocumentReader, PassthroughReader, ReaderDiagnostics, ReaderHint, ReaderOutput,
9    Result,
10};
11
12const DOC_XML_PATH: &str = "word/document.xml";
13
14pub struct DocxReader;
15
16impl DocxReader {
17    fn extract_text(bytes: &[u8]) -> Result<String> {
18        let cursor = Cursor::new(bytes);
19        let mut archive =
20            ZipArchive::new(cursor).map_err(|err| crate::MemvidError::ExtractionFailed {
21                reason: format!("failed to open docx archive: {err}").into(),
22            })?;
23
24        let mut file =
25            archive
26                .by_name(DOC_XML_PATH)
27                .map_err(|err| crate::MemvidError::ExtractionFailed {
28                    reason: format!("docx missing document.xml: {err}").into(),
29                })?;
30        let mut xml = String::new();
31        file.read_to_string(&mut xml)
32            .map_err(|err| crate::MemvidError::ExtractionFailed {
33                reason: format!("failed to read document.xml: {err}").into(),
34            })?;
35
36        Ok(extract_plain_text(&xml, b"w:p"))
37    }
38}
39
40impl DocumentReader for DocxReader {
41    fn name(&self) -> &'static str {
42        "docx"
43    }
44
45    fn supports(&self, hint: &ReaderHint<'_>) -> bool {
46        matches!(hint.format, Some(DocumentFormat::Docx))
47            || hint
48                .mime
49                .map(|mime| {
50                    mime.eq_ignore_ascii_case(
51                        "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
52                    )
53                })
54                .unwrap_or(false)
55    }
56
57    fn extract(&self, bytes: &[u8], hint: &ReaderHint<'_>) -> Result<ReaderOutput> {
58        match Self::extract_text(bytes) {
59            Ok(text) => {
60                if text.trim().is_empty() {
61                    let mut output = PassthroughReader.extract(bytes, hint)?;
62                    output.reader_name = self.name().to_string();
63                    output.diagnostics.mark_fallback();
64                    output.diagnostics.record_warning(
65                        "docx reader produced empty text; falling back to default extractor",
66                    );
67                    Ok(output)
68                } else {
69                    let mut base = PassthroughReader.extract(bytes, hint)?;
70                    base.reader_name = self.name().to_string();
71                    base.document.text = Some(text);
72                    base.document.mime_type = Some(
73                        "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
74                            .to_string(),
75                    );
76                    base.diagnostics = ReaderDiagnostics::default();
77                    Ok(base)
78                }
79            }
80            Err(err) => {
81                let mut fallback = PassthroughReader.extract(bytes, hint)?;
82                fallback.reader_name = self.name().to_string();
83                fallback.diagnostics.mark_fallback();
84                fallback
85                    .diagnostics
86                    .record_warning(format!("docx reader error: {err}"));
87                Ok(fallback)
88            }
89        }
90    }
91}
92
93fn extract_plain_text(xml: &str, block_tag: &[u8]) -> String {
94    let mut reader = XmlReader::from_str(xml);
95    reader.trim_text(true);
96    let mut buf = Vec::new();
97    let mut text = String::new();
98    let mut first_block = true;
99
100    loop {
101        match reader.read_event_into(&mut buf) {
102            Ok(Event::Start(e)) => {
103                if e.name().as_ref().ends_with(block_tag) {
104                    if !first_block {
105                        text.push('\n');
106                    }
107                    first_block = false;
108                }
109            }
110            Ok(Event::Text(t)) => {
111                if let Ok(content) = t.unescape() {
112                    if !content.trim().is_empty() {
113                        text.push_str(content.trim());
114                        text.push(' ');
115                    }
116                }
117            }
118            Ok(Event::Eof) => break,
119            Err(_) => break,
120            _ => (),
121        }
122        buf.clear();
123    }
124
125    text.trim().to_string()
126}
memvid_core/reader/docx.rs

memvid_core/reader/
docx.rs
