Lib/tarfile.py (part 2)

Source:

cpython 3.14 @ ab2d84fe1023/Lib/tarfile.py ↗

This annotation covers extraction, the 3.12 security filter, and low-level header parsing. See lib_tarfile_detail for TarFile.open, add, and TarInfo construction.

Map

Lines	Symbol	Role
1-80	Compression detection	Magic bytes for gz/bz2/xz detection in open
81-200	TarInfo.fromtarfile	Parse a 512-byte header block
201-400	GNU/POSIX extensions	Long names (GNUTYPE_LONGNAME), sparse files, PAX headers
401-600	TarFile.extractall	Extract all members; apply filter
601-800	Extraction filter (3.12)	'fully_trusted', 'tar', 'data' security policies
801-1100	TarFile._extract_member	Per-member extraction: regular/symlink/hardlink/dir/device
1101-1400	TarFile.addfile	Write a member header + data

Reading

Header parsing

# CPython: Lib/tarfile.py:1120 TarInfo.fromtarfile
@classmethod
def fromtarfile(cls, tarfile):
    buf = tarfile.fileobj.read(BLOCKSIZE)  # 512 bytes
    obj = cls.frombuf(buf, tarfile.encoding, tarfile.errors)
    obj.offset = tarfile.fileobj.tell() - BLOCKSIZE
    return obj._proc_member(tarfile)

def frombuf(cls, buf, encoding, errors):
    # Parse POSIX ustar header fields:
    # name (100), mode (8), uid (8), gid (8), size (12), mtime (12),
    # chksum (8), type (1), linkname (100), magic (6), ...
    if len(buf) == 0:
        raise EmptyHeaderError('empty header')
    if buf.count(NUL) == BLOCKSIZE:
        raise EOFHeaderError('end of file header')
    tarinfo = cls()
    tarinfo.name  = nts(buf[0:100], encoding, errors)
    tarinfo.mode  = nti(buf[100:108])
    tarinfo.size  = nti(buf[124:136])
    ...

GNU long names

# CPython: Lib/tarfile.py:1350 TarInfo._proc_gnulong
def _proc_gnulong(self, tarfile):
    """Read a GNU long name/linkname extension block."""
    buf = tarfile.fileobj.read(self._block(self.size))
    # The next header contains the actual file entry
    next_hdr = TarInfo.fromtarfile(tarfile)
    next_hdr.name = nts(buf, tarfile.encoding, tarfile.errors)
    return next_hdr

GNU tar encodes names longer than 100 bytes in a preceding header block with type GNUTYPE_LONGNAME.

extractall with filter

# CPython: Lib/tarfile.py:2080 TarFile.extractall
def extractall(self, path='.', members=None, *, numeric_owner=False,
               filter=None):
    """Extract all members to path, applying filter to each TarInfo."""
    for tarinfo in members or self:
        if filter is not None:
            tarinfo = filter(tarinfo, path)
            if tarinfo is None:
                continue   # filter vetoed this member
        self._extract_member(tarinfo, os.path.join(path, tarinfo.name),
                             numeric_owner=numeric_owner)

Extraction filter (3.12)

# CPython: Lib/tarfile.py:420 _data_filter
def _data_filter(member, dest_path):
    """Rejects members with absolute paths, '..' components, or devices."""
    if os.path.isabs(member.name) or '..' in member.name.split('/'):
        raise AbsolutePathError(member)
    if member.isdev():
        raise SpecialFileError(member)
    if member.islnk() or member.issym():
        # Check symlink target doesn't escape dest_path
        ...
    return member

filter='data' applies _data_filter. filter='tar' is less restrictive (allows symlinks). filter='fully_trusted' skips all checks (legacy behavior).

gopy notes

tarfile is pure Python and uses gzip, bz2, and lzma for compression. TarFile.extractall uses os.makedirs and open() backed by gopy's file objects. The filter parameter uses Python callables invoked during extraction.