~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2005 Canonical Ltd

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

# This program is distributed in the hope that it will be useful,

# but WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

# GNU General Public License for more details.

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

import sys, os

import bzrlib

from bzrlib.trace import mutter, note

from bzrlib.osutils import isdir, quotefn, compact_date, rand_bytes, splitpath, \

     sha_file, appendpath, file_kind

from bzrlib.errors import BzrError

BZR_BRANCH_FORMAT = "Bazaar-NG branch, format 0.0.4\n"

## TODO: Maybe include checks for common corruption of newlines, etc?

def find_branch(f, **args):

    from transport import transport

    from local_transport import LocalTransport

    t = transport(f)

    # FIXME: This is a hack around transport so that

    #        We can search the local directories for

    #        a branch root.

    if args.has_key('init') and args['init']:

        # Don't search if we are init-ing

        return Branch(t, **args)

    if isinstance(t, LocalTransport):

        root = find_branch_root(f)

        if root != f:

            t = transport(root)

    return Branch(t, **args)

def _relpath(base, path):

    """Return path relative to base, or raise exception.

    The path may be either an absolute path or a path relative to the

    current working directory.

    Lifted out of Branch.relpath for ease of testing.

    os.path.commonprefix (python2.4) has a bad bug that it works just

    on string prefixes, assuming that '/u' is a prefix of '/u2'.  This

    avoids that problem."""

    rp = os.path.abspath(path)

    s = []

    head = rp

    while len(head) >= len(base):

        if head == base:

            break

        head, tail = os.path.split(head)

        if tail:

            s.insert(0, tail)

    else:

        from errors import NotBranchError

        raise NotBranchError("path %r is not within branch %r" % (rp, base))

    return os.sep.join(s)

def find_branch_root(f=None):

    """Find the branch root enclosing f, or pwd.

    f may be a filename or a URL.

    It is not necessary that f exists.

    Basically we keep looking up until we find the control directory or

    run into the root."""

    if f == None:

        f = os.getcwd()

    else:

        f = os.path.realpath(f)

    if not os.path.exists(f):

        raise BzrError('%r does not exist' % f)

    orig_f = f

    while True:

        if os.path.exists(os.path.join(f, bzrlib.BZRDIR)):

            return f

        head, tail = os.path.split(f)

        if head == f:

            # reached the root, whatever that may be

            raise BzrError('%r is not in a branch' % orig_f)

        f = head

class DivergedBranches(Exception):

    def __init__(self, branch1, branch2):

        self.branch1 = branch1

        self.branch2 = branch2

        Exception.__init__(self, "These branches have diverged.")

class NoSuchRevision(BzrError):

    def __init__(self, branch, revision):

        self.branch = branch

        self.revision = revision

        msg = "Branch %s has no revision %d" % (branch, revision)

        BzrError.__init__(self, msg)

######################################################################

# branch objects

class Branch(object):

    """Branch holding a history of revisions.

    base

        Base directory of the branch.

    _lock_mode

        None, or 'r' or 'w'

    _lock_count

        If _lock_mode is true, a positive count of the number of times the

        lock has been taken.

    _lock

        Lock object from bzrlib.lock.

    """

    base = None

    _lock_mode = None

    _lock_count = None

    _lock = None

    cache_root = None

    # Map some sort of prefix into a namespace

    # stuff like "revno:10", "revid:", etc.

    # This should match a prefix with a function which accepts

    REVISION_NAMESPACES = {}

    def __init__(self, transport, init=False):

        """Create new branch object at a particular location.

        transport -- A Transport object, defining how to access files.

                (If a string, transport.transport() will be used to

                create a Transport object)

        init -- If True, create new control files in a previously

             unversioned directory.  If False, the branch must already

             be versioned.

        In the test suite, creation of new trees is tested using the

        `ScratchBranch` class.

        """

        if isinstance(transport, basestring):

            from transport import transport as get_transport

            transport = get_transport(transport)

        self._transport = transport

        if init:

            self._make_control()

        self._check_format()

    def __str__(self):

        return '%s(%r)' % (self.__class__.__name__, self._transport.base)

    __repr__ = __str__

    def __del__(self):

        if self._lock_mode or self._lock:

            from warnings import warn

            warn("branch %r was not explicitly unlocked" % self)

            self._lock.unlock()

        # TODO: It might be best to do this somewhere else,

        # but it is nice for a Branch object to automatically

        # cache it's information.

        # Alternatively, we could have the Transport objects cache requests

        # See the earlier discussion about how major objects (like Branch)

        # should never expect their __del__ function to run.

        if self.cache_root is not None:

            #from warnings import warn

            #warn("branch %r auto-cleanup of cache files" % self)

            try:

                import shutil

                shutil.rmtree(self.cache_root)

            except:

                pass

            self.cache_root = None

    def _get_base(self):

        if self._transport:

            return self._transport.base

        return None

    base = property(_get_base)

    def lock_write(self):

        # TODO: Upgrade locking to support using a Transport,

        # and potentially a remote locking protocol

        if self._lock_mode:

            if self._lock_mode != 'w':

                from errors import LockError

                raise LockError("can't upgrade to a write lock from %r" %

                                self._lock_mode)

            self._lock_count += 1

        else:

            self._lock = self._transport.lock_write(

                    self._rel_controlfilename('branch-lock'))

            self._lock_mode = 'w'

            self._lock_count = 1

    def lock_read(self):

        if self._lock_mode:

            assert self._lock_mode in ('r', 'w'), \

                   "invalid lock mode %r" % self._lock_mode

            self._lock_count += 1

        else:

            self._lock = self._transport.lock_read(

                    self._rel_controlfilename('branch-lock'))

            self._lock_mode = 'r'

            self._lock_count = 1

    def unlock(self):

        if not self._lock_mode:

            from errors import LockError

            raise LockError('branch %r is not locked' % (self))

        if self._lock_count > 1:

            self._lock_count -= 1

        else:

            self._lock.unlock()

            self._lock = None

            self._lock_mode = self._lock_count = None

    def abspath(self, name):

        """Return absolute filename for something in the branch"""

        return self._transport.abspath(name)

    def relpath(self, path):

        """Return path relative to this branch of something inside it.

        Raises an error if path is not in this branch."""

        return self._transport.relpath(path)

    def _rel_controlfilename(self, file_or_path):

        if isinstance(file_or_path, basestring):

            file_or_path = [file_or_path]

        return [bzrlib.BZRDIR] + file_or_path

    def controlfilename(self, file_or_path):

        """Return location relative to branch."""

        return self._transport.abspath(self._rel_controlfilename(file_or_path))

    def controlfile(self, file_or_path, mode='r'):

        """Open a control file for this branch.

        There are two classes of file in the control directory: text

        and binary.  binary files are untranslated byte streams.  Text

        control files are stored with Unix newlines and in UTF-8, even

        if the platform or locale defaults are different.

        Controlfiles should almost never be opened in write mode but

        rather should be atomically copied and replaced using atomicfile.

        """

        import codecs

        relpath = self._rel_controlfilename(file_or_path)

        #TODO: codecs.open() buffers linewise, so it was overloaded with

        # a much larger buffer, do we need to do the same for getreader/getwriter?

        # TODO: Try to use transport.put() rather than branch.controlfile(mode='w')

        if mode == 'rb': 

            return self._transport.get(relpath)

        elif mode == 'wb':

            raise BzrError("Branch.controlfile(mode='wb') is not supported, use put_controlfile")

        elif mode == 'r':

            return self._transport.get(relpath, decode=True)

        elif mode == 'w':

            raise BzrError("Branch.controlfile(mode='w') is not supported, use put_controlfile")

            #return codecs.getwriter('utf-8')(

            #        self._transport.open(relpath), errors='replace')

        else:

            raise BzrError("invalid controlfile mode %r" % mode)

    def put_controlfile(self, file_or_path, f, encode=True):

        """Write an entry as a controlfile.

        :param file_or_path: This is the sub-path underneath the bzr control directory

        :param f: A file-like or string object whose contents should be placed

                  in the appropriate location.

        :param encode:  If true, encode the contents as utf-8

        """

        self._transport.put(self._rel_controlfilename(file_or_path), f, encode=encode)

    def _make_control(self):

        from bzrlib.inventory import Inventory

        from bzrlib.xml import pack_xml

        from cStringIO import StringIO

        self._transport.mkdir(self.controlfilename([]))

        self._transport.put(self._rel_controlfilename('README'),

            "This is a Bazaar-NG control directory.\n"

            "Do not change any files in this directory.\n")

        self._transport.put(self._rel_controlfilename('branch-format'),

            BZR_BRANCH_FORMAT)

        for d in ('text-store', 'inventory-store', 'revision-store'):

            self._transport.mkdir(self._rel_controlfilename(d))

        for f in ('revision-history', 'merged-patches',

                  'pending-merged-patches', 'branch-name',

                  'branch-lock',

                  'pending-merges'):

            self._transport.put(self._rel_controlfilename(f), '')

        mutter('created control directory in ' + self._transport.base)

        # TODO: Try and do this with self._transport.put() instead

        sio = StringIO()

        pack_xml(Inventory(), sio)

        sio.seek(0)

        self.put_controlfile('inventory', sio, encode=False)

    def _check_format(self):

        """Check this branch format is supported.

        The current tool only supports the current unstable format.

        In the future, we might need different in-memory Branch

        classes to support downlevel branches.  But not yet.

        """

        # This ignores newlines so that we can open branches created

        # on Windows from Linux and so on.  I think it might be better

        # to always make all internal files in unix format.

        fmt = self.controlfile('branch-format', 'r').read()

        fmt.replace('\r\n', '')

        if fmt != BZR_BRANCH_FORMAT:

            raise BzrError('sorry, branch format %r not supported' % fmt,

                           ['use a different bzr version',

                            'or remove the .bzr directory and "bzr init" again'])

        # We know that the format is the currently supported one.

        # So create the rest of the entries.

        from bzrlib.store import CompressedTextStore

        if self._transport.is_remote():

            import tempfile

            self.cache_root = tempfile.mkdtemp()

            mutter('Branch %r using caching in %r' % (self, self.cache_root))

        else:

            self.cache_root = None

        def get_store(name):

            relpath = self._rel_controlfilename(name)

            store = CompressedTextStore(self._transport.clone(relpath))

            if self._transport.is_remote():

                from meta_store import CachedStore

                cache_path = os.path.join(self.cache_root, name)

                os.mkdir(cache_path)

                store = CachedStore(store, cache_path)

            return store

        self.text_store = get_store('text-store')

        self.revision_store = get_store('revision-store')

        self.inventory_store = get_store('inventory-store')

    def read_working_inventory(self):

        """Read the working inventory."""

        from bzrlib.inventory import Inventory

        from bzrlib.xml import unpack_xml

        from time import time

        before = time()

        self.lock_read()

        try:

            # ElementTree does its own conversion from UTF-8, so open in

            # binary.

            inv = unpack_xml(Inventory,

                                  self.controlfile('inventory', 'rb'))

            mutter("loaded inventory of %d items in %f"

                   % (len(inv), time() - before))

            return inv

        finally:

            self.unlock()

    def _write_inventory(self, inv):

        """Update the working inventory.

        That is to say, the inventory describing changes underway, that

        will be committed to the next revision.

        """

        from bzrlib.xml import pack_xml

        from cStringIO import StringIO

        self.lock_write()

        try:

            # Transport handles atomicity

            sio = StringIO()

            pack_xml(inv, sio)

            sio.seek(0)

            self._transport.put(self._rel_controlfilename('inventory'), sio)

        finally:

            self.unlock()

        mutter('wrote working inventory')

    inventory = property(read_working_inventory, _write_inventory, None,

                         """Inventory for the working copy.""")

    def add(self, files, verbose=False, ids=None):

        """Make files versioned.

        Note that the command line normally calls smart_add instead.

        This puts the files in the Added state, so that they will be

        recorded by the next commit.

        files

            List of paths to add, relative to the base of the tree.

        ids

            If set, use these instead of automatically generated ids.

            Must be the same length as the list of files, but may

            contain None for ids that are to be autogenerated.

        TODO: Perhaps have an option to add the ids even if the files do

              not (yet) exist.

        TODO: Perhaps return the ids of the files?  But then again it

              is easy to retrieve them if they're needed.

        TODO: Adding a directory should optionally recurse down and

              add all non-ignored children.  Perhaps do that in a

              higher-level method.

        """

        from bzrlib.textui import show_status

        # TODO: Re-adding a file that is removed in the working copy

        # should probably put it back with the previous ID.

        if isinstance(files, basestring):

            assert(ids is None or isinstance(ids, basestring))

            files = [files]

            if ids is not None:

                ids = [ids]

        if ids is None:

            ids = [None] * len(files)

        else:

            assert(len(ids) == len(files))

        self.lock_write()

        try:

            inv = self.read_working_inventory()

            for f,file_id in zip(files, ids):

                if is_control_file(f):

                    raise BzrError("cannot add control file %s" % quotefn(f))

                fp = splitpath(f)

                if len(fp) == 0:

                    raise BzrError("cannot add top-level %r" % f)

                fullpath = os.path.normpath(self.abspath(f))

                try:

                    kind = file_kind(fullpath)

                except OSError:

                    # maybe something better?

                    raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))

                if kind != 'file' and kind != 'directory':

                    raise BzrError('cannot add: not a regular file or directory: %s' % quotefn(f))

                if file_id is None:

                    file_id = gen_file_id(f)

                inv.add_path(f, kind=kind, file_id=file_id)

                if verbose:

                    print 'added', quotefn(f)

                mutter("add file %s file_id:{%s} kind=%r" % (f, file_id, kind))

            self._write_inventory(inv)

        finally:

            self.unlock()

    def print_file(self, file, revno):

        """Print `file` to stdout."""

        self.lock_read()

        try:

            tree = self.revision_tree(self.lookup_revision(revno))

            # use inventory as it was in that revision

            file_id = tree.inventory.path2id(file)

            if not file_id:

                raise BzrError("%r is not present in revision %s" % (file, revno))

            tree.print_file(file_id)

        finally:

            self.unlock()

    def remove(self, files, verbose=False):

        """Mark nominated files for removal from the inventory.

        This does not remove their text.  This does not run on 

        TODO: Refuse to remove modified files unless --force is given?

        TODO: Do something useful with directories.

        TODO: Should this remove the text or not?  Tough call; not

        removing may be useful and the user can just use use rm, and

        is the opposite of add.  Removing it is consistent with most

        other tools.  Maybe an option.

        """

        from bzrlib.textui import show_status

        ## TODO: Normalize names

        ## TODO: Remove nested loops; better scalability

        if isinstance(files, basestring):

            files = [files]

        self.lock_write()

        try:

            tree = self.working_tree()

            inv = tree.inventory

            # do this before any modifications

            for f in files:

                fid = inv.path2id(f)

                if not fid:

                    raise BzrError("cannot remove unversioned file %s" % quotefn(f))

                mutter("remove inventory entry %s {%s}" % (quotefn(f), fid))

                if verbose:

                    # having remove it, it must be either ignored or unknown

                    if tree.is_ignored(f):

                        new_status = 'I'

                    else:

                        new_status = '?'

                    show_status(new_status, inv[fid].kind, quotefn(f))

                del inv[fid]

            self._write_inventory(inv)

        finally:

            self.unlock()

    # FIXME: this doesn't need to be a branch method

    def set_inventory(self, new_inventory_list):

        from bzrlib.inventory import Inventory, InventoryEntry

        inv = Inventory()

        for path, file_id, parent, kind in new_inventory_list:

            name = os.path.basename(path)

            if name == "":

                continue

            inv.add(InventoryEntry(file_id, name, kind, parent))

        self._write_inventory(inv)

    def unknowns(self):

        """Return all unknown files.

        These are files in the working directory that are not versioned or

        control files or ignored.

        >>> b = ScratchBranch(files=['foo', 'foo~'])

        >>> list(b.unknowns())

        ['foo']

        >>> b.add('foo')

        >>> list(b.unknowns())

        []

        >>> b.remove('foo')

        >>> list(b.unknowns())

        ['foo']

        """

        return self.working_tree().unknowns()

    def append_revision(self, *revision_ids):

        for revision_id in revision_ids:

            mutter("add {%s} to revision-history" % revision_id)

        rev_history = self.revision_history()

        rev_history.extend(revision_ids)

        self.lock_write()

        try:

            self._transport.put(self._rel_controlfilename('revision-history'),

                    '\n'.join(rev_history))

        finally:

            self.unlock()

    def get_revision(self, revision_id):

        """Return the Revision object for a named revision"""

        from bzrlib.revision import Revision

        from bzrlib.xml import unpack_xml

        self.lock_read()

        try:

            if not revision_id or not isinstance(revision_id, basestring):

                raise ValueError('invalid revision-id: %r' % revision_id)

            r = unpack_xml(Revision, self.revision_store[revision_id])

        finally:

            self.unlock()

        assert r.revision_id == revision_id

        return r

    def get_revision_sha1(self, revision_id):

        """Hash the stored value of a revision, and return it."""

        # In the future, revision entries will be signed. At that

        # point, it is probably best *not* to include the signature

        # in the revision hash. Because that lets you re-sign

        # the revision, (add signatures/remove signatures) and still

        # have all hash pointers stay consistent.

        # But for now, just hash the contents.

        return sha_file(self.revision_store[revision_id])

    def get_inventory(self, inventory_id):

        """Get Inventory object by hash.

        TODO: Perhaps for this and similar methods, take a revision

               parameter which can be either an integer revno or a

               string hash."""

        from bzrlib.inventory import Inventory

        from bzrlib.xml import unpack_xml

        return unpack_xml(Inventory, self.inventory_store[inventory_id])

    def get_inventory_sha1(self, inventory_id):

        """Return the sha1 hash of the inventory entry

        """

        return sha_file(self.inventory_store[inventory_id])

    def get_revision_inventory(self, revision_id):

        """Return inventory of a past revision."""

        # bzr 0.0.6 imposes the constraint that the inventory_id

        # must be the same as its revision, so this is trivial.

        if revision_id == None:

            from bzrlib.inventory import Inventory

            return Inventory()

        else:

            return self.get_inventory(revision_id)

    def revision_history(self):

        """Return sequence of revision hashes on to this branch.

        >>> ScratchBranch().revision_history()

        []

        """

        self.lock_read()

        try:

            return [l.rstrip('\r\n') for l in

                    self.controlfile('revision-history', 'r').readlines()]

        finally:

            self.unlock()

    def common_ancestor(self, other, self_revno=None, other_revno=None):

        """

        >>> import commit

        >>> sb = ScratchBranch(files=['foo', 'foo~'])

        >>> sb.common_ancestor(sb) == (None, None)

        True

        >>> commit.commit(sb, "Committing first revision", verbose=False)

        >>> sb.common_ancestor(sb)[0]

        1

        >>> clone = sb.clone()

        >>> commit.commit(sb, "Committing second revision", verbose=False)

        >>> sb.common_ancestor(sb)[0]

        2

        >>> sb.common_ancestor(clone)[0]

        1

        >>> commit.commit(clone, "Committing divergent second revision", 

        ...               verbose=False)

        >>> sb.common_ancestor(clone)[0]

        1

        >>> sb.common_ancestor(clone) == clone.common_ancestor(sb)

        True

        >>> sb.common_ancestor(sb) != clone.common_ancestor(clone)

        True

        >>> clone2 = sb.clone()

        >>> sb.common_ancestor(clone2)[0]

        2

        >>> sb.common_ancestor(clone2, self_revno=1)[0]

        1

        >>> sb.common_ancestor(clone2, other_revno=1)[0]