infer

infer determines a file’s real type by inspecting the first bytes of its content against a built-in table of magic-number signatures, covering images, video, audio, archives, documents, fonts, and executables. Because it reads the byte signature rather than trusting a filename extension or user-supplied MIME type, it is commonly used to validate uploaded files server-side where an attacker could otherwise rename a malicious file to look like an innocuous one.

The crate supports both a std mode (reading from a File or byte slice, including container formats like CFB via the cfb dependency) and a no_std/alloc mode for embedded or constrained environments, and exposes a small, extensible Matcher type so applications can register custom signatures for formats infer doesn’t recognize out of the box.

What You Get

• Magic-number-based type detection covering images, video, audio, archives, documents, fonts, and executables
• get() / get_from_path() functions returning a Type with extension and MIME type
• is_*() convenience predicates (is_image, is_video, is_archive, etc.) for common category checks
• Custom Matcher registration for detecting formats not covered by the built-in signature table
• no_std + alloc compatibility for use in embedded or WASM contexts

Common Use Cases

• Validating uploaded file content server-side so an attacker can’t bypass extension-based checks by renaming a malicious file
• Sniffing the real type of a file whose extension is missing, wrong, or untrusted
• Routing files to type-specific processing pipelines (image resizing, video transcoding, document parsing) based on detected type rather than extension
• Adding detection for a proprietary or uncommon binary format via a custom Matcher

Under The Hood

Architecture - The crate centers on a Type enum returned by get()/get_from_path(), backed by a matchers/ directory splitting signature-matching logic by category (image, video, audio, archive, doc, font, app, text, odf, book), each exposing small matcher functions checked in priority order against the byte buffer; map.rs wires these matchers to their corresponding extension/MIME-type metadata.

Tech Stack - Pure Rust, no_std-compatible by default with std/alloc as opt-in Cargo features; the only optional dependency is cfb (Compound File Binary format support, e.g. for legacy MS Office formats), pulled in only under the std feature so the no_std build stays dependency-free.

Code Quality - The tests/ and testdata/ directories contain real sample files per supported format used to assert correct detection, and the matcher functions are small, single-purpose, and easy to audit individually; the crate has maintained a steady, if modest, contribution and release cadence.

API Design - The primary entry points (infer::get(buf), infer::get_from_path(path), and category predicates like infer::is_image(buf)) require no configuration and return a simple Option<Type> or bool, making the common case a one-line call; extending detection with a custom Matcher is a small amount of additional code, keeping the API approachable for both the simple and the extensible use case.