NEWS.md
In zumime: Media Type Detection from Bytes and File Names

zumime 0.1.0

Initial CRAN submission.

zumime tells what a file is from its bytes and from its name, with no dependency and no system library.

mime_sniff() runs the WHATWG MIME Sniffing Standard over a raw vector or paths, with a supplied type, the no-sniff flag and a sniffing context, and reports the charset the bytes declare and their line endings.
Beyond the standard, in the default context, container readers tell Office Open XML (with its macro-enabled, template and slideshow variants), OpenDocument, EPUB, JAR and APK from plain ZIP; Word, Excel, PowerPoint, Outlook and installer files from OLE2; MP4, M4A, M4V, QuickTime, 3GP, HEIC, HEIF and AVIF by their brands; WebM from Matroska; and Ogg codecs. R's own RDS and RData files are recognised through gzip, as are Parquet, Arrow, fst, qs2, SQLite, DuckDB, HDF5, NetCDF, Stata, SPSS and SAS files, executables and common archives.
Structured text is typed at "probable": XML dialects by root element and namespace, JSON and its dialects by top-level keys, NDJSON, line formats such as vCard, iCalendar, WebVTT and PEM, and scripts by their shebang.
mime_type() types paths from bytes and name in one C loop, says whether the two agree, and reports what a consumer needs before opening a file: the type inside a compression wrapper, encryption, completeness (with complete = TRUE), line endings and a script's interpreter. Directories, FIFOs, devices and sockets are typed inode/* without being opened.
mime_check() answers whether files are one of a set of allowed types from their bytes alone, with the same answer as mime_type().
mime_sniffer(), mime_push() and mime_finish() sniff a stream chunk by chunk and answer as soon as the bytes decide the type.
A type registry generated from IANA's media types registry and Apache httpd's mime.types: mime_types(), mime_guess(), mime_extensions(), mime_parent() and mime_is(). Every type has a parent, ending at application/octet-stream; zumime_info()$table_version dates the sources.
mime_parse() and mime_format() parse and serialise a MIME type as the WHATWG standard defines it.
mime_register() and mime_unregister() add a type for the session, with a data rule matched in C or a function rule called from R. Either refines only what zumime sniffs in the default context.
Optional help from the family, used only when asked: inner = "decompress" (zukomp), text = "family" (zucsv, zujson, zutoml) and charset = "detect" (zuutf).