ZacharyZhang-NY
b7fca0146d
ElyDoc, ElyTable, ElyShow (LibreOfficeKit), ElyPDF (MuPDF), ElyTree (Git), ElyMark (Milkdown and CodeMirror in CEF). ElyMail and ElyAgent follow.
92 lines
3.5 KiB
Rust
92 lines
3.5 KiB
Rust
//! OCR: read pages, then lay their text in hidden (PDF-08).
|
|
|
|
use std::{
|
|
io,
|
|
path::{Path, PathBuf},
|
|
};
|
|
|
|
use ely_ipc::Unpack;
|
|
use serde_json::{Value, json};
|
|
|
|
use crate::client::{Doc, Worker};
|
|
use crate::proto::tag;
|
|
|
|
/// A line Tesseract read, in page points.
|
|
#[derive(Clone, Debug, PartialEq)]
|
|
pub struct OcrLine {
|
|
pub rect: [f32; 4],
|
|
pub text: String,
|
|
}
|
|
|
|
/// Beside the app, in its macOS bundle, or this build's.
|
|
fn beside(dir: &Path) -> PathBuf {
|
|
[dir.join("tessdata"), dir.join("../Resources/tessdata")]
|
|
.into_iter()
|
|
.find(|found| found.is_dir())
|
|
.unwrap_or_else(|| Path::new(env!("CARGO_MANIFEST_DIR")).join("../../engines/tessdata"))
|
|
}
|
|
|
|
/// The language data: `ELY_TESSDATA`, the app's, or this build's.
|
|
pub fn tessdata() -> io::Result<PathBuf> {
|
|
let found = match std::env::var_os("ELY_TESSDATA") {
|
|
Some(dir) => PathBuf::from(dir),
|
|
None => {
|
|
let exe = std::env::current_exe()?;
|
|
beside(exe.parent().unwrap_or(Path::new(".")))
|
|
}
|
|
};
|
|
match found.join("eng.traineddata").is_file() {
|
|
true => Ok(found),
|
|
false => Err(io::Error::new(
|
|
io::ErrorKind::NotFound,
|
|
format!("no OCR data in {} (run scripts/fetch-tessdata.sh or set ELY_TESSDATA)", found.display()),
|
|
)),
|
|
}
|
|
}
|
|
|
|
fn line(value: &Value) -> io::Result<OcrLine> {
|
|
let bad = || io::Error::other(format!("a bad OCR line {value}"));
|
|
let list = value["rect"].as_array().filter(|list| list.len() == 4).ok_or_else(bad)?;
|
|
let mut rect = [0f32; 4];
|
|
for (at, side) in list.iter().enumerate() {
|
|
rect[at] = side.as_f64().ok_or_else(bad)? as f32;
|
|
}
|
|
Ok(OcrLine { rect, text: value["text"].as_str().ok_or_else(bad)?.to_string() })
|
|
}
|
|
|
|
impl Worker {
|
|
/// Reads `page` in `langs` (as "eng+chi_sim") at `dpi`.
|
|
pub fn ocr_page(&mut self, doc: Doc, page: usize, langs: &str, data: &Path, dpi: f32) -> io::Result<Vec<OcrLine>> {
|
|
let data = data.to_str().ok_or_else(|| io::Error::other("a data path that is not UTF-8"))?;
|
|
let ask = self.on(&doc)?.u64(page as u64).str(langs).str(data).u64(u64::from(dpi.to_bits()));
|
|
let reply: Value = serde_json::from_slice(&self.request(tag::OCR_PAGE, &ask.bytes())?).map_err(io::Error::other)?;
|
|
reply.as_array().ok_or_else(|| io::Error::other("an OCR reply that is not a list"))?.iter().map(line).collect()
|
|
}
|
|
|
|
/// Lays `pages` in hidden text, one step; the mark.
|
|
pub fn text_layer(&mut self, doc: Doc, pages: &[(usize, Vec<OcrLine>)], replace: bool) -> io::Result<String> {
|
|
let spec: Vec<Value> = (pages.iter())
|
|
.map(|(page, lines)| json!({ "page": page, "lines": lines.iter().map(|l| json!({ "rect": l.rect, "text": l.text })).collect::<Vec<_>>() }))
|
|
.collect();
|
|
let ask = self.on(&doc)?.str(&Value::Array(spec).to_string()).u64(u64::from(replace));
|
|
Unpack::new(&self.request(tag::TEXT_LAYER, &ask.bytes())?).str()
|
|
}
|
|
}
|
|
|
|
#[cfg(test)]
|
|
mod tests {
|
|
use super::beside;
|
|
|
|
#[test]
|
|
fn a_bundle_finds_its_own_data() {
|
|
let app = tempfile::tempdir().expect("a temp dir");
|
|
let (exe, data) = (app.path().join("Contents/MacOS"), app.path().join("Contents/Resources/tessdata"));
|
|
std::fs::create_dir_all(&exe).expect("MacOS");
|
|
assert!(beside(&exe).ends_with("engines/tessdata"));
|
|
std::fs::create_dir_all(&data).expect("tessdata");
|
|
assert_eq!(beside(&exe), exe.join("../Resources/tessdata"));
|
|
std::fs::create_dir(exe.join("tessdata")).expect("beside");
|
|
assert_eq!(beside(&exe), exe.join("tessdata"));
|
|
}
|
|
}
|