#!/usr/bin/python
# repofs.py - Export the contents of a package repo as a readonly filesystem.
# 
# Copyright 2009 Red Hat, Inc. GPLv2+ BOILERPLATE GOES HERE.
#
# Author: Will Woods <wwoods@redhat.com>
#
# Exports the contents of the repo(s) in a filesystem that looks like this:
# $PACKAGE_UID/[package contents]
# where $PACKAGE_UID is some unique package identifier (e.g. ENVRA in RPM)

import os
import glob
from stat import *
import errno
import subprocess
import time

import fuse
fuse.fuse_python_api = (0, 2)
fuse.feature_assert('stateful_files', 'has_init')
from fuse import Fuse

from sqlite3 import dbapi2 as sqlite

import bz2, gzip

import yum.repoMDObject

FILEMODE = S_IRUSR|S_IRGRP|S_IROTH
DIRMODE = FILEMODE|S_IFDIR|S_IXUSR|S_IXGRP|S_IXOTH

def bunzip(infile,outfile):
    (p,f) = os.path.split(outfile)
    if not os.path.isdir(p):
        os.makedirs(p)
    outf = open(outfile,"w")
    inf = bz2.BZ2File(infile)
    data = inf.read(4096)
    while data:
        outf.write(data)
        data = inf.read(4096)
    inf.close()
    outf.close()

def flag2mode(flags):
    md = {os.O_RDONLY: 'r', os.O_WRONLY: 'w', os.O_RDWR: 'w+'}
    m = md[flags & (os.O_RDONLY | os.O_WRONLY | os.O_RDWR)]
    if flags | os.O_APPEND:
        m = m.replace('w', 'a', 1)
    return m

def fuseclonestat(stat):
    s = fuse.Stat()
    (s.st_mode, s.st_ino, s.st_dev, s.st_nlink, s.st_uid, s.st_gid, s.st_size,
            s.st_atime, s.st_mtime, s.st_ctime) = stat
    return s

def leading_paths(dir):
    dirlist = []
    while dir:
        dirlist.append(dir)
        if dir == '/':
            break
        (dir, dummy) = os.path.split(dir)
    return dirlist
      
class SimpleYumRepo(object):
    def __init__(self, path=None, cachedir=None):
        self.path = path
        self.cachedir = cachedir
        self.pkgkey = {}
        if path:
            self.parse_repomd()

    def parse_repomd(self):
        repomd = os.path.join(self.path, "repodata/repomd.xml")
        repoXML = yum.repoMDObject.RepoMD(self.path, repomd)
        for t in ('primary_db', 'filelists_db'):
            if t in repoXML.fileTypes():
                d = repoXML.getData(t)
                (base,dbpath) = d.location
                dbfile = os.path.join(self.path,dbpath)
                # TODO check for existing db file
                if dbfile.endswith(".bz2"):
                    outfile = os.path.join(self.cachedir,".repodata",
                                           os.path.basename(dbfile)[:-4])
                    bunzip(dbfile,outfile)
                    dbfile = outfile
                # TODO: elif .gz, else...
                con = sqlite.connect(dbfile)
                setattr(self,t,con)
        for p in self.package_uids():
            pass

    # This is kind of stupid - we need this cached to use files_for_package..
    # Maybe this should populate the cache and package_uids() can just return
    # pkgkey.keys()
    def package_uids(self):
        '''Generator for package UIDs in this repo'''
        c = self.primary_db.cursor()
        c.execute("SELECT epoch, name, version, release, arch, pkgKey FROM packages")
        for (e,n,v,r,a,key) in c:
            nevra = "%s-%s:%s-%s.%s" % (n,e,v,r,a)
            self.pkgkey[nevra] = key
            yield nevra

    def files_for_package(self, packageuid):
        '''Generator for files in the given packageuid.
        Each item is a tuple of the form (abspath,type) where type is
        'd' or 'f' (dir/file).'''
        c = self.filelists_db.cursor()
        key = self.pkgkey.get(packageuid)
        if not key:
            raise StopIteration # like returning an empty list
        c.execute("SELECT dirname, filenames, filetypes FROM filelist "
                  "WHERE pkgKey=?",(key,))
        dirs = []
        for (dir,names,types) in c:
            if dir not in dirs:
                for d in leading_paths(dir): 
                    if d not in dirs:
                        dirs.append(d)
                        yield (d,'d')
            for (n,t) in zip(names.split('/'),types):
                f = os.path.join(dir,n)
                if t == 'd' and f in dirs:
                    continue
                else:
                    dirs.append(f)
                yield (f,t)

    def unpack(self,rpm,targetdir=None):
        if targetdir:
            if not os.path.isdir(targetdir):
                os.makedirs(targetdir)
        else:
            targetdir = self.cachedir

        inf = open(rpm)
        # Find RPM header and read compression algorithm
        # Skip forward to gzipped CPIO archive
        # FIXME: Awful. Just awful. At least use rpm2cpio.
        header = inf.read(409600)
        offset = header.index("\x1f\x8b")
        del header
        inf.seek(offset)
        gz = gzip.GzipFile(fileobj=inf, mode="rb")
        # Open a pipe to "cpio -iumd --quiet"
        cpio = subprocess.Popen(args=["cpio","-iumd","--quiet"],
                cwd=targetdir, stdin=subprocess.PIPE)
        data = gz.read(4096)
        while data:
            cpio.stdin.write(data)
            data = gz.read(4096)
        gz.close()
        inf.close()
        cpio.stdin.close()
        cpio.wait()

class FuseRO(Fuse):
    '''A Fuse subclass for implementing readonly filesystems.'''
    # chmod chown create link mkdir mknod rename rmdir setxattr(?) symlink
    # truncate unlink write
    def __rofs(self, *args):
        '''Raises OSError(EROFS,"Read-only filesystem")'''
        raise OSError(errno.EROFS, "Read-only filesystem")
    chmod = __rofs
    chown = __rofs
    link = __rofs
    mkdir = __rofs
    rename = __rofs
    rmdir = __rofs
    setxattr = __rofs # You might override this so you can store xattrs
    symlink = __rofs
    truncate = __rofs
    unlink = __rofs
    def write(self, *args):
        '''write() function that raises IOError(EBADF)
        You can't open files for writing; this is a readonly filesystem!'''
        raise IOError(errno.EBADF, "write() on readonly filesystem")
    def _check_open(self, flags):
        '''checks the open() flags, and returns False if write access
        was requested. Returns True otherwise.'''
        accmode = os.O_RDONLY | os.O_WRONLY | os.O_RDWR
        return (flags & accmode) == os.O_RDONLY

class Repofs(FuseRO):
    def __init__(self, *args, **kw):
        Fuse.__init__(self, *args, **kw)

    # FIXME: this logging is terrible
    def log(self,message):
        self.logfile.write("%s %s\n" % (time.asctime(), message))
        
    def fsinit(self):
        if not os.path.isdir(self.cachedir):
            os.makedirs(self.cachedir)
        # FIXME this logging is awful
        self.logfile = open(os.path.join(self.cachedir,".log"),"a",0)
        self.log("fsinit(path=%s). Hang on.." % self.repopath)
        # TODO: figure out the repo type (Yum, etc) and use the right class.
        # That way we can support other distros. Yay!
        self.repos = []
        for rp in self.repopath.split(":"):
            r = SimpleYumRepo(path=rp, cachedir=self.cachedir)
            self.repos.append(r)
            self.log(" cachedir=%s, repopath=%s" % (r.cachedir, r.path))

    def _splitpath(self, path):
        '''Split an absolute path into (packageuid, path)'''
        path = path.lstrip('/')
        p = path.split('/',1)
        if len(p) == 1:
            p.append('')
        p[1] = '/' + p[1]
        return p

    def readdir(self, path, offset):
        self.log("readdir('%s', %s)" % (path, str(offset)))
        for repo in self.repos:
            if path == "/":
                for uid in repo.package_uids():
                    d = fuse.Direntry(str(uid))
                    d.type = S_IFDIR
                    yield d
            else:
                (packageuid, path) = self._splitpath(path)
                for (f,t) in repo.files_for_package(packageuid):
                    (dir, basename) = os.path.split(f)
                    if dir == path and basename:
                        d = fuse.Direntry(str(basename))
                        if t == 'd':
                            d.type = S_IFDIR
                        else:
                            d.type = S_IFREG
                        yield d

    def getattr(self, path):
        self.log("getattr('%s')" % path)
        attr = fuse.Stat()
        attr.st_mode = None
        attr.st_nlink = 2 # sure, why not
        if (path == '/'):
            attr.st_mode = DIRMODE
            return attr
        (packageuid, path) = self._splitpath(path)
        for repo in self.repos:
            for (f,t) in repo.files_for_package(packageuid):
                if f == path:
                    # found it! set mode to -r--r--r--
                    attr.st_mode = FILEMODE
                    if t == 'd': # change mode to dr-xr-xr-x
                        attr.st_mode = DIRMODE
                    # TODO: set some more attributes
                    return attr
        #raise OSError(errno.ENOENT, "No such file or directory")
        # SourceForge FUSE Python reference says to use this instead:
        return -errno.ENOENT

#    # FIXME actually use the provided arg
#    def utime(self, path, times):
#        self.log("utime(%s, %s)" % path, str(times))
#        return os.utime(self._cachefile(path))
#    def access(self, path, mode):
#        self.log("access(%s, %s)" % path, mode)
#        if not os.access(self._cachefile(path)):
#            return -errno.EACCES
#
    def statfs(self):
        #self.log("statfs()")
        s = fuse.StatVFS()
        local_s = os.statvfs(self.cachedir)
        # FIXME modify s using info from local_s
        return local_s
#
#    # XXX explicitly declare other functions that return proper error codes?
#    
#    # FIXME: use file objects instead?
#
#    def open(self, path, flags):
#        self.log("open(%s,%s)" % (path,flags))
#        return open(self._cachefile(path),flag2mode(flags))
#
#    def read(self, path, length, offset, fh=None):
#        self.log("read(%s,%s,%s)" % (path,length,offset))
#        fh.seek(offset)
#        return fh.read(length)
#
#    def release(self, path, fh=None):
#        self.log("release(%s)" % path)
#        fh.close()
#
#    def fgetattr(self, path, fh=None):
#        self.log("fgetattr(%s)" % path)
#        return os.fstat(fh.fileno())

    def main(self, *a, **kw):
        return Fuse.main(self, *a, **kw)


def main():
    usage = 'Repofs: mount a package repo and export all the files in the packages.\n\n' + Fuse.fusage
    server = Repofs(version="%prog " + fuse.__version__, usage=usage,
                    dash_s_do='setsingle')
    server.parser.add_option(mountopt="repo", metavar="PATH", dest="repopath",
            help="Package repo to mount")
    server.parser.add_option(mountopt="cachedir", metavar="PATH",
            help="Cache dir for expanded packages")
    server.parse(values=server, errex=1)
    server.main()

if __name__ == '__main__':
    main()