pdfminer.six ported to Rust, with a Node.js binding.
The port produces the same output as pdfminer.six: the same text, the same layout objects with bit-identical coordinates, and the same errors on broken files. It follows a pinned pdfminer.six release (now 20260107) module by module, so new upstream releases can be ported as small diffs.
const { extractPages, extractText, extractTextAsync, convert, PdfminerError } = require("@fuelfwd/pdfminer");
// Sync and async (`extractPagesAsync`, ... run on a worker thread) variants of each function.
// Input: a path, Buffer, Uint8Array or ArrayBuffer.
const pages = extractPages("file.pdf", { laparams: { charMargin: 2.0 }, pageNumbers: [0, 1] });
for (const page of pages) {
for (const item of page.items) {
if (item.type === "LTTextBoxHorizontal") console.log(item.index, item.bbox);
}
}
const text = await extractTextAsync(fs.readFileSync("file.pdf"), { password: "secret" });
const xml = convert("file.pdf", { outputType: "xml" }); // Buffer, as pdf2txt.py writes it
try {
extractText("broken.pdf");
} catch (err) {
if (err instanceof PdfminerError) console.log(err.pythonClass, err.pythonMessage); // e.g. "PDFSyntaxError"
}Pages are plain objects with the same class names and attributes as pdfminer's layout objects
(LTPage, LTTextBoxHorizontal, LTTextLineHorizontal, LTChar, LTAnno, LTCurve,
LTRect, LTLine, LTImage, LTFigure): type, bbox, items, and per type text,
fontname, size, matrix, graphicstate, ... See crates/pdfminer-node/index.d.ts.
pdfminer-rs pdf2txt [pdf2txt.py options] file.pdf # same options and output as pdf2txt.py
pdfminer-rs dumppdf [dumppdf.py options] file.pdf # same as dumppdf.py
pdfminer-rs dump file.pdf # the full layout tree as JSON lines| Path | What |
|---|---|
upstream/pdfminer.six |
The pinned upstream release (git submodule). |
crates/pdfminer |
The port. src/<module>.rs is pdfminer/<module>.py. src/pyrt/ reproduces the Python semantics the results depend on. |
crates/pdfminer-cli |
The pdfminer-rs command (pdf2txt, dumppdf, dump). |
crates/pdfminer-node |
The Node binding (napi-rs), package @fuelfwd/pdfminer. |
tools/ |
difftest.py (Python vs Rust on a corpus), oracle.py (the Python reference dump), upstream.py (upstream sync status), codegen.py (generated tables), gen_testdata/ (differential test data per module). |
UPSTREAM.toml |
For every upstream file: its Rust counterpart and the upstream version it matches. |
docs/ |
PORTING.md (the rules), UPDATING.md (how to move to a new release), DIVERGENCES.md (the complete list of known differences). |
Requirements: Rust (pinned in rust-toolchain.toml), uv (it
installs the reference Python and pdfminer.six from the submodule), Node 20+ for the binding.
git submodule update --init
uv sync
cargo test # unit tests and the ported upstream tests
uv run python tools/difftest.py # must print "0 different"Add your own PDFs to the comparison with PDFMINER_CORPUS=/dir1:/dir2.
To move to a new pdfminer.six release, follow docs/UPDATING.md.
MIT, like pdfminer.six. See LICENSE.