Module detect 

Format detection by content sniffing.

The file extension is a hint and is never consulted here. A .jpg that is really a PDF must be handled as a PDF or refused outright; handing it to the JPEG handler would produce a confident success message about a file that was never touched (docs/ARCHITECTURE.md §1, docs/THREAT_MODEL.md §5.4). Detection therefore takes bytes and nothing else — there is deliberately no way to pass it a path.

Detection is itself a hostile-input parser: it is the one piece of code that sees every byte of every file the user feeds in, including files no handler will ever accept. It reads through [crate::bytes::Reader] for the same reason the handlers do.

Why this is hand-written rather than a dependency

file-format and infer were both evaluated (docs/ARCHITECTURE.md §4). Phase 1 needed to discriminate exactly four supported formats plus a short list of formats worth naming in a refusal, which is under a hundred lines of magic-number matching. Taking a crate with broad magic tables for that would add supply-chain surface (ADR-0008) to save very little.

Phase 2 brought the ambiguity that comment anticipated, and it turned out not to be the kind a magic table solves. .docx, .xlsx, .pptx, and every OpenDocument file share one magic number, because they are all ZIP archives. Telling them apart means opening the container and reading the content type the package declares for its own main part — which no magic-number crate does either, and which the ZIP layer this crate already owns does directly (ADR-0027, ADR-0028).

Why detection opens the container

It would be cheaper to search the raw bytes for word/document.xml and be done. That is also how a file gets routed to the wrong handler: the string appears verbatim in any archive that merely contains a Word document, and an attacker can put it in a comment. Reading the declared content type is the format’s own answer to “what is this”, and it costs one central directory walk and one small inflate.

Enums

Format

A format strypt has a handler for.

Functions

detect

Identify data by content.