mupdf_basic_text_extractor/
lib.rs

1use mupdf::{TextPageOptions};
2
3#[derive(Default, Debug, PartialEq, Clone)]
4pub struct Fragment {
5    pub text: String,
6    pub x: f64,
7    pub y: f64,
8    pub font_name: Option<String>,
9    pub font_size: f64,
10    pub bbox_width: f64,
11    pub bbox_height: f64
12
13}
14#[derive(Default, Debug, Clone)]
15pub struct Line {
16    pub text_fragments: Vec<Fragment>
17
18}
19#[derive(Default, Debug, Clone)]
20pub struct Page {
21    pub lines: Vec<Line>
22}
23#[derive(Default, Debug, Clone)]
24pub struct Doc {
25    pub pages: Vec<Page>
26}
27
28pub fn get_structured_document_from_filepath(path: String)-> Result<Doc, Box<dyn std::error::Error>> {
29    let filename: String = path;
30    let document = mupdf::Document::open(&filename)?;
31    let mut new_document: Doc = Doc::default();
32
33    for page in document.pages()? {
34        let pageref = page.as_ref();
35        let pagebounds = &pageref.unwrap().bounds()?;
36        let text_page = &page?.to_text_page(TextPageOptions::empty())?;
37        let text_blocks = text_page.blocks();
38
39        let mut raw_page_fragments = Vec::<Fragment>::default();
40
41        for block in text_blocks {
42            
43            let lines = block.lines();
44            for line in lines {
45                let chars = line.chars();
46                let mut current_working_fragment= Fragment::default();
47
48                let mut working_frag_reset = false;
49                let mut first_char = true;
50                for char in chars {
51
52                    if working_frag_reset {
53                        first_char = true;
54                        working_frag_reset = false;
55                    }
56
57                    if first_char {
58                        current_working_fragment.x = char.quad().ll.x as f64;
59                        current_working_fragment.y = pagebounds.height() as f64 - char.quad().ll.y as f64 ;
60
61                        current_working_fragment.font_size = char.size() as f64;
62                        current_working_fragment.bbox_width = (char.quad().lr.x - char.quad().ll.x) as f64;
63                        current_working_fragment.bbox_height = (char.quad().ur.y - char.quad().lr.y) as f64;
64                        
65                        first_char = false;
66                    }
67                    if let Some(ch) = char.char() {
68                        if ch.is_whitespace() {
69                            raw_page_fragments.push(current_working_fragment);
70                            current_working_fragment = Fragment::default();
71                            working_frag_reset = true;
72                        }
73                        current_working_fragment.text.push(ch);
74                    }
75                    
76                }
77                if current_working_fragment != Fragment::default() {
78                    raw_page_fragments.push(current_working_fragment);
79                }
80            }
81        }
82
83        let mut new_page = Page::default();
84
85        let mut current_working_newline = Line::default();
86        let mut working_newline_reset = false;
87        let mut last_y:f64 = 0.0;
88        if let Some(first_fr) = raw_page_fragments.first() {
89            last_y = first_fr.y.clone();
90        }
91
92        for fr in &raw_page_fragments {
93            if working_newline_reset {
94                working_newline_reset = false;
95            }
96            if fr.y != last_y {
97                new_page.lines.push(current_working_newline.clone());
98                current_working_newline = Line::default();
99                working_newline_reset = true;
100            }
101            current_working_newline.text_fragments.push(fr.clone());
102            last_y = fr.y;
103        }
104        if working_newline_reset {
105            new_page.lines.push(current_working_newline);
106        }
107
108        for line in &mut new_page.lines {
109            line.text_fragments.sort_by(|a, b| a.x.total_cmp(&b.x));
110        }
111
112        new_document.pages.push(new_page);
113    }
114
115    Ok(new_document)
116}
117
118#[cfg(test)]
119mod tests {
120
121    use std::fs;
122
123    use super::*;
124
125    #[test]
126    fn it_works() {
127        let structured_doc_extracted =
128        get_structured_document_from_filepath("test_pdfs/DraftTest_02.pdf".into());
129        
130        if let Ok(doc) = structured_doc_extracted {
131            for page in doc.pages {
132                println!("PAGE -----");
133                for line in page.lines {
134                    print!("     LINE: ----- ");
135                    if let Some(first_frag) = line.text_fragments.first() {
136                        print!("Y: {}, FONTSIZE: {}", first_frag.y, first_frag.font_size);
137                        println!("\n");
138                    }
139                    print!("          ");
140                    for frag in line.text_fragments {
141                        print!("{} |", frag.text);
142                    }
143                    println!("");
144                }
145            }
146        }
147
148    }
149}