Skip to main content

snapshot/
snapshot.rs

1//! Generate the PDF snapshot baseline.
2//!
3//! `cargo run --release -p fleischwolf-pdf --example snapshot -- <root> <outdir>`
4//!
5//! Recursively converts every `*.pdf` under `<root>` and writes its Markdown to
6//! `<outdir>/<relative-path>.md` (the pipeline's deterministic output, committed
7//! as the conformance baseline). Conversion errors are written verbatim as
8//! `ERROR: …` so failures are captured, never silently skipped.
9
10use std::path::{Path, PathBuf};
11
12use fleischwolf_pdf::Pipeline;
13
14const IMAGE_EXTS: &[&str] = &["png", "jpg", "jpeg", "tif", "tiff", "bmp", "gif", "webp"];
15
16fn is_supported(p: &Path) -> bool {
17    matches!(
18        p.extension().and_then(|e| e.to_str()),
19        Some(e) if e == "pdf" || e == "gz" || IMAGE_EXTS.contains(&e)
20    )
21}
22
23fn find_pdfs(dir: &Path, out: &mut Vec<PathBuf>) {
24    let Ok(entries) = std::fs::read_dir(dir) else {
25        return;
26    };
27    let mut entries: Vec<_> = entries.flatten().map(|e| e.path()).collect();
28    entries.sort();
29    for p in entries {
30        if p.is_dir() {
31            find_pdfs(&p, out);
32        } else if is_supported(&p) {
33            out.push(p);
34        }
35    }
36}
37
38fn main() {
39    let mut args = std::env::args().skip(1);
40    let root = PathBuf::from(args.next().expect("usage: snapshot <root> <outdir>"));
41    let outdir = PathBuf::from(args.next().expect("usage: snapshot <root> <outdir>"));
42
43    let mut pdfs = Vec::new();
44    find_pdfs(&root, &mut pdfs);
45
46    let mut pipeline = Pipeline::new().expect("load pipeline");
47    let (mut ok, mut err) = (0u32, 0u32);
48    for pdf in &pdfs {
49        let rel = pdf.strip_prefix(&root).unwrap_or(pdf);
50        let name = pdf.file_name().unwrap().to_string_lossy().to_string();
51        let md = match std::fs::read(pdf)
52            .map_err(|e| format!("read: {e}"))
53            .and_then(|bytes| {
54                let ext = pdf.extension().and_then(|e| e.to_str()).unwrap_or("");
55                let result = if ext == "gz" {
56                    fleischwolf_pdf::convert_mets_gbs(&bytes, &name)
57                } else if IMAGE_EXTS.contains(&ext) {
58                    pipeline.convert_image(&bytes, &name)
59                } else {
60                    pipeline.convert(&bytes, None, &name)
61                };
62                result
63                    .map(|d| d.export_to_markdown())
64                    .map_err(|e| e.to_string())
65            }) {
66            Ok(md) => {
67                ok += 1;
68                md
69            }
70            Err(e) => {
71                err += 1;
72                eprintln!("ERR {}: {e}", rel.display());
73                format!("ERROR: {e}\n")
74            }
75        };
76        // Groundtruth naming keeps the source extension: `<file>.<ext>.md`.
77        let mut dest = outdir.join(rel).into_os_string();
78        dest.push(".md");
79        let dest = PathBuf::from(dest);
80        std::fs::create_dir_all(dest.parent().unwrap()).expect("mkdir");
81        std::fs::write(&dest, md).expect("write snapshot");
82    }
83    eprintln!("snapshots: {} ok, {} error, {} total", ok, err, pdfs.len());
84}