mailrs-mime

RFC 2045 / 2046 MIME body parser. Walks the multipart tree, decodes Content-Transfer-Encoding (base64, quoted-printable, 7bit, 8bit, binary), exposes text / HTML extraction + find-by-content-type + attachment iteration.

Layered: built on top of mailrs-rfc5322 (headers), mailrs-rfc2231 (params), and mailrs-rfc2047 is available for downstream callers who need encoded-word header decode.

Replaces residual mail-parser usage in inbound code paths where only the MIME tree shape matters (find calendar invite part, extract attachments, get the text body for indexing).

Quickstart

use mailrs_mime::parse;

let raw = b"Content-Type: multipart/alternative; boundary=\"x\"\r\n\
\r\n\
--x\r\n\
Content-Type: text/plain\r\n\
\r\n\
plain version\r\n\
--x\r\n\
Content-Type: text/html\r\n\
\r\n\
<p>html version</p>\r\n\
--x--\r\n";

let root = parse(raw);
assert!(root.content_type.is_multipart());

// Find the HTML alternative:
let html = root.find_by_content_type("text/html").unwrap();
assert_eq!(html.body_text().as_deref(), Some("<p>html version</p>"));

// Iterate all parts depth-first:
for part in root.walk() {
    println!("{}", part.content_type.mime_type());
}

What this crate does

• parse(raw) → Part — top-level entry. Returns a tree matching the message's multipart structure.
• Part::walk() — depth-first iterator over self + descendants
• Part::find_by_content_type("text/calendar") — exact-match lookup, useful for iTIP / iMIP invite extraction
• Part::body_text() — for text/* leaves, decodes per the part's charset= (via encoding_rs)
• Part::attachments() + Part::attachment_filename() — finds parts marked attachment (by Content-Disposition or filename parameter)
• Content-Transfer-Encoding decoders: base64, quoted-printable, identity (7bit/8bit/binary), unknown (passes through)
• RFC 2231 parameter decoding built in — filename*=UTF-8''… forms surface as plain UTF-8 via Part::attachment_filename()

What this crate does not

• No RFC 2047 encoded-word decode in headers — that's mailrs-rfc2047. Use it on From / Subject / display names after extracting via mailrs-rfc5322.
• No body autodetection beyond charset — if the message lies about its charset, we trust the lie. (Real parsers occasionally apply heuristics for mojibake recovery; out of v1.0 scope.)
• No DKIM / SPF / DMARC — separate stones (mailrs-dkim, mailrs-spf, mailrs-dmarc).
• No address-list structured parse — From: returns raw bytes via mailrs-rfc5322; building (name, addr) from it is downstream.
• No outbound MIME builder — this crate is read-only. Use a builder crate for composing outbound mail.

Layering

mailrs-mime                       ← this crate (MIME body tree)
   ├── mailrs-rfc5322             ← header lookup, lazy
   ├── mailrs-rfc2231             ← parameter encode/decode
   └── encoding_rs + base64       ← foundational

Each layer is independently published; downstream callers can pull whichever subset they need without the rest.

Performance

Measured (criterion, M-series Mac, release):

Compared to mail-parser 0.11 on the same inputs:

The simple case is close to a wash; the multipart walk path is where mail-parser's optimized internals still win. Future polish rounds may close that gap. v1.0 ships at shape parity — the data extracted is the same; perf is acceptable for inbound where parse cost is dwarfed by DNS / disk / DB calls.

Reproduce: cargo bench -p mailrs-mime --bench mime.

License

Apache-2.0 OR MIT.