/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, os, os.path, random, time, sha, sets, types, re, shutil, tempfile

import traceback, socket, fnmatch, difflib, time

from binascii import hexlify

import bzrlib

from inventory import Inventory

from trace import mutter, note

from tree import Tree, EmptyTree, RevisionTree

from inventory import InventoryEntry, Inventory

from osutils import isdir, quotefn, isfile, uuid, sha_file, username, \

     format_date, compact_date, pumpfile, user_email, rand_bytes, splitpath, \

     joinpath, sha_string, file_kind, local_time_offset, appendpath

from store import ImmutableStore

from revision import Revision

from errors import BzrError

from textui import show_status

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):

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

        import remotebranch 

        return remotebranch.RemoteBranch(f, **args)

    else:

        return Branch(f, **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()

    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 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.")

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

# 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

    def __init__(self, base, init=False, find_root=True):

        """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.

        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('.')):

                from errors import NotBranchError

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

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

                                      'current bzr can only operate from top-of-tree'])

        self._check_format()

        self.text_store = ImmutableStore(self.controlfilename('text-store'))

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

        self.inventory_store = ImmutableStore(self.controlfilename('inventory-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, types.StringTypes):

            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.")

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

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

            os.mkdir(self.controlfilename(d))

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

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

                  'branch-lock'):

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

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

        Inventory().write_xml(self.controlfile('inventory','w'))

    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'])

    def read_working_inventory(self):

        """Read the working inventory."""

        before = time.time()

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

        # binary.

        self.lock_read()

        try:

            inv = Inventory.read_xml(self.controlfile('inventory', 'rb'))

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

                   % (len(inv), time.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.

        """

        ## TODO: factor out to atomicfile?  is rename safe on windows?

        ## TODO: Maybe some kind of clean/dirty marker on inventory?

        tmpfname = self.controlfilename('inventory.tmp')

        tmpf = file(tmpfname, 'wb')

        inv.write_xml(tmpf)

        tmpf.close()

        inv_fname = self.controlfilename('inventory')

        if sys.platform == 'win32':

            os.remove(inv_fname)

        os.rename(tmpfname, inv_fname)

        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.

        """

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

        # should probably put it back with the previous ID.

        if isinstance(files, types.StringTypes):

            assert(ids is None or isinstance(ids, types.StringTypes))

            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:

                    show_status('A', kind, 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 %d" % (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, types.StringTypes):

            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):

        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_id):

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

        rev_history = self.revision_history()

        tmprhname = self.controlfilename('revision-history.tmp')

        rhname = self.controlfilename('revision-history')

        f = file(tmprhname, 'wt')

        rev_history.append(revision_id)

        f.write('\n'.join(rev_history))

        f.write('\n')

        f.close()

        if sys.platform == 'win32':

            os.remove(rhname)

        os.rename(tmprhname, rhname)

    def get_revision(self, revision_id):

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

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

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

        r = Revision.read_xml(self.revision_store[revision_id])

        assert r.revision_id == revision_id

        return r

    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."""

        i = Inventory.read_xml(self.inventory_store[inventory_id])

        return i

    def get_revision_inventory(self, revision_id):

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

        if revision_id == None:

            return Inventory()

        else:

            return self.get_inventory(self.get_revision(revision_id).inventory_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]

        1

        """

        my_history = self.revision_history()

        other_history = other.revision_history()

        if self_revno is None:

            self_revno = len(my_history)

        if other_revno is None:

            other_revno = len(other_history)

        indices = range(min((self_revno, other_revno)))

        indices.reverse()

        for r in indices:

            if my_history[r] == other_history[r]:

                return r+1, my_history[r]

        return None, None

    def enum_history(self, direction):

        """Return (revno, revision_id) for history of branch.

        direction

            'forward' is from earliest to latest

            'reverse' is from latest to earliest

        """

        rh = self.revision_history()

        if direction == 'forward':

            i = 1

            for rid in rh:

                yield i, rid

                i += 1

        elif direction == 'reverse':

            i = len(rh)

            while i > 0:

                yield i, rh[i-1]

                i -= 1

        else:

            raise ValueError('invalid history direction', direction)

    def revno(self):

        """Return current revision number for this branch.

        That is equivalent to the number of revisions committed to

        this branch.

        """

        return len(self.revision_history())

    def last_patch(self):

        """Return last patch hash, or None if no history.

        """

        ph = self.revision_history()

        if ph:

            return ph[-1]

        else:

            return None

    def missing_revisions(self, other):

        """

        If self and other have not diverged, return a list of the revisions

        present in other, but missing from self.

        >>> from bzrlib.commit import commit

        >>> bzrlib.trace.silent = True

        >>> br1 = ScratchBranch()

        >>> br2 = ScratchBranch()

        >>> br1.missing_revisions(br2)

        []

        >>> commit(br2, "lala!", rev_id="REVISION-ID-1")

        >>> br1.missing_revisions(br2)

        [u'REVISION-ID-1']

        >>> br2.missing_revisions(br1)

        []

        >>> commit(br1, "lala!", rev_id="REVISION-ID-1")

        >>> br1.missing_revisions(br2)

        []

        >>> commit(br2, "lala!", rev_id="REVISION-ID-2A")

        >>> br1.missing_revisions(br2)

        [u'REVISION-ID-2A']

        >>> commit(br1, "lala!", rev_id="REVISION-ID-2B")

        >>> br1.missing_revisions(br2)

        Traceback (most recent call last):

        DivergedBranches: These branches have diverged.

        """

        self_history = self.revision_history()

        self_len = len(self_history)

        other_history = other.revision_history()

        other_len = len(other_history)

        common_index = min(self_len, other_len) -1

        if common_index >= 0 and \

            self_history[common_index] != other_history[common_index]:

            raise DivergedBranches(self, other)

        if self_len < other_len:

            return other_history[self_len:]

        return []

    def update_revisions(self, other):

        """Pull in all new revisions from other branch.

        >>> from bzrlib.commit import commit

        >>> bzrlib.trace.silent = True

        >>> br1 = ScratchBranch(files=['foo', 'bar'])

        >>> br1.add('foo')

        >>> br1.add('bar')

        >>> commit(br1, "lala!", rev_id="REVISION-ID-1", verbose=False)

        >>> br2 = ScratchBranch()

        >>> br2.update_revisions(br1)

        Added 2 texts.

        Added 1 inventories.

        Added 1 revisions.

        >>> br2.revision_history()

        [u'REVISION-ID-1']

        >>> br2.update_revisions(br1)

        Added 0 texts.

        Added 0 inventories.

        Added 0 revisions.

        >>> br1.text_store.total_size() == br2.text_store.total_size()

        True

        """

        revision_ids = self.missing_revisions(other)

        revisions = [other.get_revision(f) for f in revision_ids]

        needed_texts = sets.Set()

        for rev in revisions:

            inv = other.get_inventory(str(rev.inventory_id))

            for key, entry in inv.iter_entries():

                if entry.text_id is None:

                    continue

                if entry.text_id not in self.text_store:

                    needed_texts.add(entry.text_id)

        count = self.text_store.copy_multi(other.text_store, needed_texts)

        print "Added %d texts." % count 

        inventory_ids = [ f.inventory_id for f in revisions ]

        count = self.inventory_store.copy_multi(other.inventory_store, 

                                                inventory_ids)