/brz/remove-bazaar

# 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

import os

from cStringIO import StringIO

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, InvalidRevisionNumber, InvalidRevisionId,

                           NoSuchRevision, HistoryMissing, NotBranchError)

from bzrlib.textui import show_status

from bzrlib.revision import Revision, validate_revision_id

from bzrlib.delta import compare_trees

from bzrlib.tree import EmptyTree, RevisionTree

from bzrlib.inventory import Inventory

from bzrlib.weavestore import WeaveStore

from bzrlib.store import ImmutableStore

import bzrlib.xml5

import bzrlib.ui

INVENTORY_FILEID = '__inventory'

ANCESTRY_FILEID = '__ancestry'

BZR_BRANCH_FORMAT_4 = "Bazaar-NG branch, format 0.0.4\n"

BZR_BRANCH_FORMAT_5 = "Bazaar-NG branch, format 5\n"

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

# TODO: Some operations like log might retrieve the same revisions

# repeatedly to calculate deltas.  We could perhaps have a weakref

# cache in memory to make this faster.  In general anything can be

# cached in memory between lock and unlock operations.

# TODO: please move the revision-string syntax stuff out of the branch

# object; it's clutter

def find_branch(f, **args):

    if f and (f.startswith('http://') or f.startswith('https://')):

        import remotebranch 

        return remotebranch.RemoteBranch(f, **args)

    else:

        return Branch(f, **args)

def find_cached_branch(f, cache_root, **args):

    from remotebranch import RemoteBranch

    br = find_branch(f, **args)

    def cacheify(br, store_name):

        from meta_store import CachedStore

        cache_path = os.path.join(cache_root, store_name)

        os.mkdir(cache_path)

        new_store = CachedStore(getattr(br, store_name), cache_path)

        setattr(br, store_name, new_store)

    if isinstance(br, RemoteBranch):

        cacheify(br, 'inventory_store')

        cacheify(br, 'text_store')

        cacheify(br, 'revision_store')

    return br

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:

        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 there isn't one, raises NotBranchError.

    """

    if f == None:

        f = os.getcwd()

    elif hasattr(os.path, 'realpath'):

        f = os.path.realpath(f)

    else:

        f = os.path.abspath(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 NotBranchError('%s is not in a branch' % orig_f)

        f = head

# XXX: move into bzrlib.errors; subclass BzrError    

class DivergedBranches(Exception):

    def __init__(self, branch1, branch2):

        self.branch1 = branch1

        self.branch2 = branch2

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

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

# 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

    _inventory_weave = 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, base, init=False, find_root=True,

                 relax_version_check=False):

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

        base -- Base directory for the branch.

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

             unversioned directory.  If False, the branch must already

             be versioned.

        find_root -- If true and init is false, find the root of the

             existing branch containing base.

        relax_version_check -- If true, the usual check for the branch

            version is not applied.  This is intended only for

            upgrade/recovery type use; it's not guaranteed that

            all operations will work on old format branches.

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

        `ScratchBranch` class.

        """

        if init:

            self.base = os.path.realpath(base)

            self._make_control()

        elif find_root:

            self.base = find_branch_root(base)

        else:

            self.base = os.path.realpath(base)

            if not isdir(self.controlfilename('.')):

                raise NotBranchError("not a bzr branch: %s" % quotefn(base),

                                     ['use "bzr init" to initialize a new working tree'])

        self._check_format(relax_version_check)

        self.weave_store = WeaveStore(self.controlfilename('weaves'))

        self.revision_store = ImmutableStore(self.controlfilename('revision-store'))

    def __str__(self):

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

    def lock_write(self):

        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:

            from bzrlib.lock import WriteLock

            self._lock = WriteLock(self.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:

            from bzrlib.lock import ReadLock

            self._lock = ReadLock(self.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 os.path.join(self.base, 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 _relpath(self.base, path)

    def controlfilename(self, file_or_path):

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

        if isinstance(file_or_path, basestring):

            file_or_path = [file_or_path]

        return os.path.join(self.base, bzrlib.BZRDIR, *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.

        """

        fn = self.controlfilename(file_or_path)

        if mode == 'rb' or mode == 'wb':

            return file(fn, mode)

        elif mode == 'r' or mode == 'w':

            # open in binary mode anyhow so there's no newline translation;

            # codecs uses line buffering by default; don't want that.

            import codecs

            return codecs.open(fn, mode + 'b', 'utf-8',

                               buffering=60000)

        else:

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

    def _make_control(self):

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

        self.controlfile('README', 'w').write(

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

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

        self.controlfile('branch-format', 'w').write(BZR_BRANCH_FORMAT_5)

        for d in ('text-store', 'revision-store',

                  'weaves'):

            os.mkdir(self.controlfilename(d))

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

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

                  'branch-lock',

                  'pending-merges'):

            self.controlfile(f, 'w').write('')

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

        # if we want per-tree root ids then this is the place to set

        # them; they're not needed for now and so ommitted for

        # simplicity.

        f = self.controlfile('inventory','w')

        bzrlib.xml5.serializer_v5.write_inventory(Inventory(), f)

    def _check_format(self, relax_version_check):

        """Check this branch format is supported.

        The format level is stored, as an integer, in

        self._branch_format for code that needs to check it later.

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

        classes to support downlevel branches.  But not yet.

        """

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

        if fmt == BZR_BRANCH_FORMAT_5:

            self._branch_format = 5

            return

        elif relax_version_check:

            if fmt == BZR_BRANCH_FORMAT_4:

                self._branch_format = 4

                return

        raise BzrError('sorry, branch format "%s" not supported; ' 

                       'use a different bzr version, '

                       'or run "bzr upgrade"'

                       % fmt.rstrip('\n\r'))

    def get_root_id(self):

        """Return the id of this branches root"""

        inv = self.read_working_inventory()

        return inv.root.file_id

    def set_root_id(self, file_id):

        inv = self.read_working_inventory()

        orig_root_id = inv.root.file_id

        del inv._byid[inv.root.file_id]

        inv.root.file_id = file_id

        inv._byid[inv.root.file_id] = inv.root

        for fid in inv:

            entry = inv[fid]

            if entry.parent_id in (None, orig_root_id):

                entry.parent_id = inv.root.file_id

        self._write_inventory(inv)

    def read_working_inventory(self):

        """Read the working inventory."""

        self.lock_read()

        try:

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

            # binary.

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

            return bzrlib.xml5.serializer_v5.read_inventory(f)

        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.atomicfile import AtomicFile

        self.lock_write()

        try:

            f = AtomicFile(self.controlfilename('inventory'), 'wb')

            try:

                bzrlib.xml5.serializer_v5.write_inventory(inv, f)

                f.commit()

            finally:

                f.close()

        finally:

            self.unlock()

        mutter('wrote working inventory')

    inventory = property(read_working_inventory, _write_inventory, None,

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

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

        """Make files versioned.

        Note that the command line normally calls smart_add instead,

        which can automatically recurse.

        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 yield the ids and paths as they're added.

        """

        # 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)

                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.

        """

        ## 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(self.get_root_id())

        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):

        from bzrlib.atomicfile import AtomicFile

        for revision_id in revision_ids:

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

        rev_history = self.revision_history()

        rev_history.extend(revision_ids)

        f = AtomicFile(self.controlfilename('revision-history'))

        try:

            for rev_id in rev_history:

                print >>f, rev_id

            f.commit()

        finally:

            f.close()

    def has_revision(self, revision_id):

        """True if this branch has a copy of the revision.

        This does not necessarily imply the revision is merge

        or on the mainline."""

        return revision_id in self.revision_store

    def get_revision_xml_file(self, revision_id):

        """Return XML file object for revision object."""

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

            raise InvalidRevisionId(revision_id)

        self.lock_read()

        try:

            try:

                return self.revision_store[revision_id]

            except IndexError:

                raise bzrlib.errors.NoSuchRevision(self, revision_id)

        finally:

            self.unlock()

    def get_revision_xml(self, revision_id):

        return self.get_revision_xml_file(revision_id).read()

    def get_revision(self, revision_id):

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

        xml_file = self.get_revision_xml_file(revision_id)

        try:

            r = bzrlib.xml5.serializer_v5.read_revision(xml_file)

        except SyntaxError, e:

            raise bzrlib.errors.BzrError('failed to unpack revision_xml',

                                         [revision_id,

                                          str(e)])

        assert r.revision_id == revision_id

        return r

    def get_revision_delta(self, revno):

        """Return the delta for one revision.

        The delta is relative to its mainline predecessor, or the

        empty tree for revision 1.

        """

        assert isinstance(revno, int)

        rh = self.revision_history()

        if not (1 <= revno <= len(rh)):

            raise InvalidRevisionNumber(revno)

        # revno is 1-based; list is 0-based

        new_tree = self.revision_tree(rh[revno-1])

        if revno == 1:

            old_tree = EmptyTree()

        else:

            old_tree = self.revision_tree(rh[revno-2])

        return compare_trees(old_tree, new_tree)

    def get_revision_sha1(self, revision_id):

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

        return bzrlib.osutils.sha_file(self.get_revision_xml_file(revision_id))

    def get_ancestry(self, revision_id):

        """Return a list of revision-ids integrated by a revision.

        """

        w = self.weave_store.get_weave(ANCESTRY_FILEID)

        # strip newlines

        return [l[:-1] for l in w.get_iter(w.lookup(revision_id))]

    def get_inventory_weave(self):

        return self.weave_store.get_weave(INVENTORY_FILEID)

    def get_inventory(self, revision_id):

        """Get Inventory object by hash."""

        # FIXME: The text gets passed around a lot coming from the weave.

        f = StringIO(self.get_inventory_xml(revision_id))

        return bzrlib.xml5.serializer_v5.read_inventory(f)

    def get_inventory_xml(self, revision_id):

        """Get inventory XML as a file object."""

        try:

            assert isinstance(revision_id, basestring), type(revision_id)

            iw = self.get_inventory_weave()

            return iw.get_text(iw.lookup(revision_id))

        except IndexError:

            raise bzrlib.errors.HistoryMissing(self, 'inventory', revision_id)

    def get_inventory_sha1(self, revision_id):