]> arthur.barton.de Git - bup.git/blobdiff - lib/bup/vfs.py
tag-cmd: Some fixups
[bup.git] / lib / bup / vfs.py
index c8f688a9c33f096311dabbcbda1bee2a595328ec..16a8d33b858999d05d37e74eeac947592d89fea0 100644 (file)
@@ -1,3 +1,8 @@
+"""Virtual File System representing bup's repository contents.
+
+The vfs.py library makes it possible to expose contents from bup's repository
+and abstracts internal name mangling and storage from the exposition layer.
+"""
 import os, re, stat, time
 from bup import git
 from helpers import *
@@ -6,20 +11,30 @@ EMPTY_SHA='\0'*20
 
 _cp = None
 def cp():
+    """Create a git.CatPipe object or reuse the already existing one."""
     global _cp
     if not _cp:
         _cp = git.CatPipe()
     return _cp
 
 class NodeError(Exception):
+    """VFS base exception."""
     pass
+
 class NoSuchFile(NodeError):
+    """Request of a file that does not exist."""
     pass
+
 class NotDir(NodeError):
+    """Attempt to do a directory action on a file that is not one."""
     pass
+
 class NotFile(NodeError):
+    """Access to a node that does not represent a file."""
     pass
+
 class TooManySymlinks(NodeError):
+    """Symlink dereferencing level is too deep."""
     pass
 
 
@@ -27,7 +42,7 @@ def _treeget(hash):
     it = cp().get(hash.encode('hex'))
     type = it.next()
     assert(type == 'tree')
-    return git._treeparse(''.join(it))
+    return git.treeparse(''.join(it))
 
 
 def _tree_decode(hash):
@@ -88,7 +103,8 @@ class _ChunkReader:
             self.blob = None
         else:
             self.it = None
-            self.blob = cp().get(hash.encode('hex'))
+            self.blob = ''.join(cp().join(hash.encode('hex')))[startofs:]
+        self.ofs = startofs
 
     def next(self, size):
         out = ''
@@ -104,11 +120,12 @@ class _ChunkReader:
                 self.blob = self.blob[want:]
             if not self.it:
                 break
-        log('next(%d) returned %d\n' % (size, len(out)))
+        debug2('next(%d) returned %d\n' % (size, len(out)))
+        self.ofs += len(out)
         return out
 
 
-class _FileReader:
+class _FileReader(object):
     def __init__(self, hash, size, isdir):
         self.hash = hash
         self.ofs = 0
@@ -117,9 +134,6 @@ class _FileReader:
         self.reader = None
 
     def seek(self, ofs):
-        if self.ofs == ofs:
-            return
-        self.reader = None
         if ofs > self.size:
             self.ofs = self.size
         elif ofs < 0:
@@ -133,7 +147,7 @@ class _FileReader:
     def read(self, count = -1):
         if count < 0:
             count = self.size - self.ofs
-        if not self.reader:
+        if not self.reader or self.reader.ofs != self.ofs:
             self.reader = _ChunkReader(self.hash, self.isdir, self.ofs)
         try:
             buf = self.reader.next(count)
@@ -143,8 +157,12 @@ class _FileReader:
         self.ofs += len(buf)
         return buf
 
+    def close(self):
+        pass
+
 
 class Node:
+    """Base class for file representation."""
     def __init__(self, parent, name, mode, hash):
         self.parent = parent
         self.name = name
@@ -152,28 +170,33 @@ class Node:
         self.hash = hash
         self.ctime = self.mtime = self.atime = 0
         self._subs = None
-        
+
     def __cmp__(a, b):
-        return cmp(a.name or None, b.name or None)
-    
+        return cmp(a and a.name or None, b and b.name or None)
+
     def __iter__(self):
         return iter(self.subs())
-    
-    def fullname(self):
-        if self.parent:
-            return os.path.join(self.parent.fullname(), self.name)
+
+    def fullname(self, stop_at=None):
+        """Get this file's full path."""
+        assert(self != stop_at)  # would be the empty string; too weird
+        if self.parent and self.parent != stop_at:
+            return os.path.join(self.parent.fullname(stop_at=stop_at),
+                                self.name)
         else:
             return self.name
-    
+
     def _mksubs(self):
         self._subs = {}
-        
+
     def subs(self):
+        """Get a list of nodes that are contained in this node."""
         if self._subs == None:
             self._mksubs()
         return sorted(self._subs.values())
-        
+
     def sub(self, name):
+        """Get node named 'name' that is contained in this node."""
         if self._subs == None:
             self._mksubs()
         ret = self._subs.get(name)
@@ -182,13 +205,24 @@ class Node:
         return ret
 
     def top(self):
+        """Return the very top node of the tree."""
         if self.parent:
             return self.parent.top()
         else:
             return self
 
+    def fs_top(self):
+        """Return the top node of the particular backup set.
+
+        If this node isn't inside a backup set, return the root level.
+        """
+        if self.parent and not isinstance(self.parent, CommitList):
+            return self.parent.fs_top()
+        else:
+            return self
+
     def _lresolve(self, parts):
-        #log('_lresolve %r in %r\n' % (parts, self.name))
+        #debug2('_lresolve %r in %r\n' % (parts, self.name))
         if not parts:
             return self
         (first, rest) = (parts[0], parts[1:])
@@ -203,40 +237,68 @@ class Node:
         else:
             return self.sub(first)
 
-    def lresolve(self, path):
+    def lresolve(self, path, stay_inside_fs=False):
+        """Walk into a given sub-path of this node.
+
+        If the last element is a symlink, leave it as a symlink, don't resolve
+        it.  (like lstat())
+        """
         start = self
+        if not path:
+            return start
         if path.startswith('/'):
-            start = self.top()
+            if stay_inside_fs:
+                start = self.fs_top()
+            else:
+                start = self.top()
             path = path[1:]
         parts = re.split(r'/+', path or '.')
         if not parts[-1]:
             parts[-1] = '.'
-        #log('parts: %r %r\n' % (path, parts))
+        #debug2('parts: %r %r\n' % (path, parts))
         return start._lresolve(parts)
 
-    def resolve(self, path):
-        return self.lresolve(path).lresolve('')
-    
+    def resolve(self, path = ''):
+        """Like lresolve(), and dereference it if it was a symlink."""
+        return self.lresolve(path).lresolve('.')
+
+    def try_resolve(self, path = ''):
+        """Like resolve(), but don't worry if a symlink uses an invalid path.
+
+        Returns an error if any intermediate nodes were invalid.
+        """
+        n = self.lresolve(path)
+        try:
+            n = n.lresolve('.')
+        except NoSuchFile:
+            pass
+        return n
+
     def nlinks(self):
+        """Get the number of hard links to the current node."""
         if self._subs == None:
             self._mksubs()
         return 1
 
     def size(self):
+        """Get the size of the current node."""
         return 0
 
     def open(self):
+        """Open the current node. It is an error to open a non-file node."""
         raise NotFile('%s is not a regular file' % self.name)
 
 
 class File(Node):
+    """A normal file from bup's repository."""
     def __init__(self, parent, name, mode, hash, bupmode):
         Node.__init__(self, parent, name, mode, hash)
         self.bupmode = bupmode
         self._cached_size = None
         self._filereader = None
-        
+
     def open(self):
+        """Open the file."""
         # You'd think FUSE might call this only once each time a file is
         # opened, but no; it's really more of a refcount, and it's called
         # once per read().  Thus, it's important to cache the filereader
@@ -244,54 +306,74 @@ class File(Node):
         if not self._filereader:
             self._filereader = _FileReader(self.hash, self.size(),
                                            self.bupmode == git.BUP_CHUNKED)
+        self._filereader.seek(0)
         return self._filereader
-    
+
     def size(self):
+        """Get this file's size."""
         if self._cached_size == None:
-            log('<<<<File.size() is calculating\n')
+            debug1('<<<<File.size() is calculating (for %r)...\n' % self.name)
             if self.bupmode == git.BUP_CHUNKED:
                 self._cached_size = _total_size(self.hash)
             else:
                 self._cached_size = _chunk_len(self.hash)
+            debug1('<<<<File.size() done.\n')
         return self._cached_size
 
 
 _symrefs = 0
 class Symlink(File):
+    """A symbolic link from bup's repository."""
     def __init__(self, parent, name, hash, bupmode):
         File.__init__(self, parent, name, 0120000, hash, bupmode)
 
     def size(self):
+        """Get the file size of the file at which this link points."""
         return len(self.readlink())
 
     def readlink(self):
+        """Get the path that this link points at."""
         return ''.join(cp().join(self.hash.encode('hex')))
 
     def dereference(self):
+        """Get the node that this link points at.
+
+        If the path is invalid, raise a NoSuchFile exception. If the level of
+        indirection of symlinks is 100 levels deep, raise a TooManySymlinks
+        exception.
+        """
         global _symrefs
         if _symrefs > 100:
             raise TooManySymlinks('too many levels of symlinks: %r'
                                   % self.fullname())
         _symrefs += 1
         try:
-            return self.parent.lresolve(self.readlink())
+            try:
+                return self.parent.lresolve(self.readlink(),
+                                            stay_inside_fs=True)
+            except NoSuchFile:
+                raise NoSuchFile("%s: broken symlink to %r"
+                                 % (self.fullname(), self.readlink()))
         finally:
             _symrefs -= 1
 
     def _lresolve(self, parts):
         return self.dereference()._lresolve(parts)
-    
+
 
 class FakeSymlink(Symlink):
+    """A symlink that is not stored in the bup repository."""
     def __init__(self, parent, name, toname):
         Symlink.__init__(self, parent, name, EMPTY_SHA, git.BUP_NORMAL)
         self.toname = toname
-        
+
     def readlink(self):
+        """Get the path that this link points at."""
         return self.toname
-    
+
 
 class Dir(Node):
+    """A directory stored inside of bup's repository."""
     def _mksubs(self):
         self._subs = {}
         it = cp().get(self.hash.encode('hex'))
@@ -301,7 +383,7 @@ class Dir(Node):
             it = cp().get(self.hash.encode('hex') + ':')
             type = it.next()
         assert(type == 'tree')
-        for (mode,mangled_name,sha) in git._treeparse(''.join(it)):
+        for (mode,mangled_name,sha) in git.treeparse(''.join(it)):
             mode = int(mode, 8)
             name = mangled_name
             (name,bupmode) = git.demangle_name(mangled_name)
@@ -313,45 +395,140 @@ class Dir(Node):
                 self._subs[name] = Symlink(self, name, sha, bupmode)
             else:
                 self._subs[name] = File(self, name, mode, sha, bupmode)
-                
+
+
+class CommitDir(Node):
+    """A directory that contains all commits that are reachable by a ref.
+
+    Contains a set of subdirectories named after the commits' first byte in
+    hexadecimal. Each of those directories contain all commits with hashes that
+    start the same as the directory name. The name used for those
+    subdirectories is the hash of the commit without the first byte. This
+    separation helps us avoid having too much directories on the same level as
+    the number of commits grows big.
+    """
+    def __init__(self, parent, name):
+        Node.__init__(self, parent, name, 040000, EMPTY_SHA)
+
+    def _mksubs(self):
+        self._subs = {}
+        refs = git.list_refs()
+        for ref in refs:
+            #debug2('ref name: %s\n' % ref[0])
+            revs = git.rev_list(ref[1].encode('hex'))
+            for (date, commit) in revs:
+                #debug2('commit: %s  date: %s\n' % (commit.encode('hex'), date))
+                commithex = commit.encode('hex')
+                containername = commithex[:2]
+                dirname = commithex[2:]
+                n1 = self._subs.get(containername)
+                if not n1:
+                    n1 = CommitList(self, containername)
+                    self._subs[containername] = n1
+
+                if n1.commits.get(dirname):
+                    # Stop work for this ref, the rest should already be present
+                    break
+
+                n1.commits[dirname] = (commit, date)
+
 
 class CommitList(Node):
+    """A list of commits with hashes that start with the current node's name."""
+    def __init__(self, parent, name):
+        Node.__init__(self, parent, name, 040000, EMPTY_SHA)
+        self.commits = {}
+
+    def _mksubs(self):
+        self._subs = {}
+        for (name, (hash, date)) in self.commits.items():
+            n1 = Dir(self, name, 040000, hash)
+            n1.ctime = n1.mtime = date
+            self._subs[name] = n1
+
+
+class TagDir(Node):
+    """A directory that contains all tags in the repository."""
+    def __init__(self, parent, name):
+        Node.__init__(self, parent, name, 040000, EMPTY_SHA)
+
+    def _mksubs(self):
+        self._subs = {}
+        for (name, sha) in git.list_refs():
+            if name.startswith('refs/tags/'):
+                name = name[10:]
+                date = git.rev_get_date(sha.encode('hex'))
+                commithex = sha.encode('hex')
+                target = '../.commit/%s/%s' % (commithex[:2], commithex[2:])
+                tag1 = FakeSymlink(self, name, target)
+                tag1.ctime = tag1.mtime = date
+                self._subs[name] = tag1
+
+
+class BranchList(Node):
+    """A list of links to commits reachable by a branch in bup's repository.
+
+    Represents each commit as a symlink that points to the commit directory in
+    /.commit/??/ . The symlink is named after the commit date.
+    """
     def __init__(self, parent, name, hash):
         Node.__init__(self, parent, name, 040000, hash)
-        
+
     def _mksubs(self):
         self._subs = {}
+
+        tags = git.tags()
+
         revs = list(git.rev_list(self.hash.encode('hex')))
         for (date, commit) in revs:
             l = time.localtime(date)
             ls = time.strftime('%Y-%m-%d-%H%M%S', l)
-            commithex = '.' + commit.encode('hex')
-            n1 = Dir(self, commithex, 040000, commit)
-            n2 = FakeSymlink(self, ls, commithex)
-            n1.ctime = n1.mtime = n2.ctime = n2.mtime = date
-            self._subs[commithex] = n1
-            self._subs[ls] = n2
-            latest = max(revs)
+            commithex = commit.encode('hex')
+            target = '../.commit/%s/%s' % (commithex[:2], commithex[2:])
+            n1 = FakeSymlink(self, ls, target)
+            n1.ctime = n1.mtime = date
+            self._subs[ls] = n1
+
+            for tag in tags.get(commit, []):
+                t1 = FakeSymlink(self, tag, target)
+                t1.ctime = t1.mtime = date
+                self._subs[tag] = t1
+
+        latest = max(revs)
         if latest:
             (date, commit) = latest
-            commithex = '.' + commit.encode('hex')
-            n2 = FakeSymlink(self, 'latest', commithex)
-            n2.ctime = n2.mtime = date
-            self._subs['latest'] = n2
+            commithex = commit.encode('hex')
+            target = '../.commit/%s/%s' % (commithex[:2], commithex[2:])
+            n1 = FakeSymlink(self, 'latest', target)
+            n1.ctime = n1.mtime = date
+            self._subs['latest'] = n1
+
 
-    
 class RefList(Node):
+    """A list of branches in bup's repository.
+
+    The sub-nodes of the ref list are a series of CommitList for each commit
+    hash pointed to by a branch.
+
+    Also, a special sub-node named '.commit' contains all commit directories
+    that are reachable via a ref (e.g. a branch).  See CommitDir for details.
+    """
     def __init__(self, parent):
         Node.__init__(self, parent, '/', 040000, EMPTY_SHA)
-        
+
     def _mksubs(self):
         self._subs = {}
+
+        commit_dir = CommitDir(self, '.commit')
+        self._subs['.commit'] = commit_dir
+
+        tag_dir = TagDir(self, '.tag')
+        self._subs['.tag'] = tag_dir
+
         for (name,sha) in git.list_refs():
             if name.startswith('refs/heads/'):
                 name = name[11:]
                 date = git.rev_get_date(sha.encode('hex'))
-                n1 = CommitList(self, name, sha)
+                n1 = BranchList(self, name, sha)
                 n1.ctime = n1.mtime = date
                 self._subs[name] = n1
-        
-