#!/usr/bin/python
"""This tool reads a Debian package from stdin and emits a yaml stream on
stdout.  It does not access a database. Therefore it can be run in parallel and
on multiple machines. The generated yaml contains multiple documents. The first
document contains package metadata. Then a document is emitted for each file.
And finally a document consisting of the string "commit" is emitted."""

import hashlib
import optparse
import sys
import tarfile
import zlib

import lzma
import yaml

from dedup.arreader import ArReader
from dedup.debpkg import process_control, get_tar_hashes
from dedup.hashing import DecompressedHash, SuppressingHash, HashedStream, \
        HashBlacklistContent
from dedup.compression import GzipDecompressor, DecompressedStream
from dedup.image import GIFHash, PNGHash

boring_content = set(("", "\n"))

def sha512_nontrivial():
    return HashBlacklistContent(hashlib.sha512(), boring_content)

def gziphash():
    hashobj = DecompressedHash(GzipDecompressor(), hashlib.sha512())
    hashobj = SuppressingHash(hashobj, (ValueError, zlib.error))
    hashobj.name = "gzip_sha512"
    return HashBlacklistContent(hashobj, boring_content)

def pnghash():
    hashobj = PNGHash(hashlib.sha512())
    hashobj = SuppressingHash(hashobj, (ValueError,))
    hashobj.name = "png_sha512"
    return hashobj

def gifhash():
    hashobj = GIFHash(hashlib.sha512())
    hashobj = SuppressingHash(hashobj, (ValueError,))
    hashobj.name = "gif_sha512"
    return hashobj

def decompress_tar(filelike, extension):
    if extension in (".lzma", ".xz"):
        filelike = DecompressedStream(filelike, lzma.LZMADecompressor())
        extension = ""
    if extension not in ("", ".gz", ".bz2"):
        raise ValueError("unknown compression format with extension %r" %
                         extension)
    return tarfile.open(fileobj=filelike, mode="r|" + extension[1:])

def process_package(filelike, hash_functions):
    af = ArReader(filelike)
    af.read_magic()
    state = "start"
    while True:
        try:
            name = af.read_entry()
        except EOFError:
            raise ValueError("data.tar not found")
        if name.startswith("control.tar"):
            if state != "start":
                raise ValueError("unexpected control.tar")
            state = "control"
            tf = decompress_tar(af, name[11:])
            for elem in tf:
                if elem.name != "./control":
                    continue
                if state != "control":
                    raise ValueError("duplicate control file")
                state = "control_file"
                yield process_control(tf.extractfile(elem).read())
                break
            continue
        elif name.startswith("data.tar"):
            if state != "control_file":
                raise ValueError("missing control file")
            state = "data"
            tf = decompress_tar(af, name[8:])
            for name, size, hashes in get_tar_hashes(tf, hash_functions):
                # filenames are not actually iso-8859-1, but this decode
                # cannot cause UnicodeDecodeError
                name = name.decode("iso-8859-1")
                yield dict(name=name, size=size, hashes=hashes)
            yield "commit"
            break

def process_package_with_hash(filelike, hash_functions, sha256hash):
    hstream = HashedStream(filelike, hashlib.sha256())
    for elem in process_package(hstream, hash_functions):
        if elem == "commit":
            while hstream.read(4096):
                pass
            if hstream.hexdigest() != sha256hash:
                raise ValueError("hash sum mismatch")
            yield elem
            break
        yield elem

def main():
    parser = optparse.OptionParser()
    parser.add_option("-H", "--hash", action="store",
                      help="verify that stdin hash given sha256 hash")
    options, args = parser.parse_args()
    hash_functions = [sha512_nontrivial, gziphash, pnghash, gifhash]
    if options.hash:
        gen = process_package_with_hash(sys.stdin, hash_functions, options.hash)
    else:
        gen = process_package(sys.stdin, hash_functions)
    yaml.safe_dump_all(gen, sys.stdout)

if __name__ == "__main__":
    main()