This project provides functionality to read and write a variety of tar archive formats. Currently supported are the POSIX ustar, PAX (ustar with a few new entry types), GNU, and v7 (very old) formats.

This project is rather low level and is focused exclusively on reading and writing physical tar file entries using streams. Therefore, it contains no functionality to automatically build archives from a set of files on the filesystem or write the contents of a file to the filesystem. Additionally, there are no smarts that read multiple physical entries and combine them into a single logical entry (e.g., with PAX extended headers or GNU long link/path name support). This approach is taken to both improve portability (the CL spec has no facility for things like detecting symlinks or modifying/reading file owners, etc.) and properly separate concerns.

For a higher level library that reads and writes logical entries, as well as includes file system integration, see cl-tar.

This project is a fork of Nathan Froyd's archive library. Much code remains, but the non-portable bits have been stripped, better support for multiple archive types (and autodetection) has been added, better blocking support added (tar readers/writers are supposed to read/write in multiples of 512 bytes), cpio support removed, and a test suite added, along with other miscellaneous fixes and improvements.

One major user visible difference between this library and the original archive is that there is no need to discard entries when you are finished with them and support has been added for seeking within a stream (using FILE-POSITION under the hood). This means you can do something like iterate over all entries in one go and then get a stream containing the contents of an arbitrary entry.

However, care must be taken if you hop around in a tar file as it will only work if the underlying stream containing the tar archive is seekable. Typically, streams backed by files are seekable, but all other input streams (including stdin) are not. Therefore if there is any chance your code that uses tar-file will read files from non-seekable streams, you should process entries sequentially and completely before moving on to the next one. Alternatively, you can set the blocking factor high enough that the entire file is read into memory at once (yikes!).