processors_rs/
docx_processor.rs

1use crate::markdown_processor::MarkdownProcessor;
2use crate::processor::{Document, DocumentProcessor, FileProcessor};
3use docx_parser::MarkdownDocument;
4use std::path::Path;
5use text_splitter::ChunkConfigError;
6
7/// A struct for processing PDF files.
8pub struct DocxProcessor {
9    markdown_processor: MarkdownProcessor,
10}
11
12impl DocxProcessor {
13    pub fn new(chunk_size: usize, overlap: usize) -> Result<DocxProcessor, ChunkConfigError> {
14        let markdown_processor = MarkdownProcessor::new(chunk_size, overlap)?;
15        Ok(DocxProcessor { markdown_processor })
16    }
17}
18
19impl FileProcessor for DocxProcessor {
20    fn process_file(&self, path: impl AsRef<Path>) -> anyhow::Result<Document> {
21        // `docx-parser::MarkdownDocument::from_file` uses `panic!` instead of returning
22        // `Result` when the file is missing, corrupt, or not a valid DOCX/ZIP archive.
23        // We catch that panic here and convert it into a proper anyhow error so callers
24        // get a clean Err(…) rather than a process-level abort.
25        let path = path.as_ref().to_owned();
26        let docs =
27            std::panic::catch_unwind(move || MarkdownDocument::from_file(&path)).map_err(|e| {
28                let msg = if let Some(s) = e.downcast_ref::<String>() {
29                    s.clone()
30                } else if let Some(s) = e.downcast_ref::<&str>() {
31                    s.to_string()
32                } else {
33                    "unknown panic".to_string()
34                };
35                anyhow::anyhow!("docx_parser panicked while opening file: {}", msg)
36            })?;
37        let markdown = docs.to_markdown(false);
38        self.markdown_processor.process_document(&markdown)
39    }
40}
41
42#[cfg(test)]
43mod tests {
44    use super::*;
45    #[test]
46    fn test_extract_text() {
47        let txt_file = "../test_files/test.docx";
48        let processor = DocxProcessor::new(128, 0).unwrap();
49
50        let text = processor.process_file(txt_file).unwrap();
51        assert!(text
52            .chunks
53            .contains(&"This is a docx file test".to_string()));
54    }
55
56    // Returns an error if the file path is invalid.
57    #[test]
58    #[should_panic(
59        expected = "Error processing file: IO(Os { code: 2, kind: NotFound, message: \"No such file or directory\" })"
60    )]
61    fn test_extract_text_invalid_file_path() {
62        let invalid_file_path = "this_file_definitely_does_not_exist.docx";
63        let processor = DocxProcessor::new(128, 0).unwrap();
64        processor.process_file(invalid_file_path).unwrap();
65    }
66}
processors_rs/docx_processor.rs

processors_rs/
docx_processor.rs
