# $Id: back_metakit.py,v 1.57 2004-01-27 18:10:48 wc2so1 Exp $
'''
   Metakit backend for Roundup, originally by Gordon McMillan.

   Notes by Richard:

   This backend has some behaviour specific to metakit:

    - there's no concept of an explicit "unset" in metakit, so all types
      have some "unset" value:

      ========= ===== ====================================================
      Type      Value Action when fetching from mk
      ========= ===== ====================================================
      Strings   ''    convert to None
      Date      0     (seconds since 1970-01-01.00:00:00) convert to None
      Interval  ''    convert to None
      Number    0     ambiguious :( - do nothing (see BACKWARDS_COMPATIBLE)
      Boolean   0     ambiguious :( - do nothing (see BACKWARDS_COMPATABILE)
      Link      0     convert to None
      Multilink []    actually, mk can handle this one ;)
      Password  ''    convert to None
      ========= ===== ====================================================

      The get/set routines handle these values accordingly by converting
      to/from None where they can. The Number/Boolean types are not able
      to handle an "unset" at all, so they default the "unset" to 0.

    Metakit relies in reference counting to close the database, there is
    no explicit close call.  This can cause issues if a metakit
    database is referenced multiple times, one might not actually be
    closing the db.                                    
    
    - probably a bunch of stuff that I'm not aware of yet because I haven't
      fully read through the source. One of these days....
'''
# Enable this flag to break backwards compatibility (i.e. can't read old databases)
#  but comply with more roundup features, like adding NULL support.
BACKWARDS_COMPATIBLE = True

# Changes to version 1.55:
# 1 Added an explicit close to the Indexer class.  This was handled
#   by garbage collection before.
#
# 2  Added an MKBackendError exception that gets thrown on some metakit errors.
#   Should there be a general rdbms exception?
#      
# 3 Added a sanity check when creating metakit tables in __getview
#   There are some test cases that create columns of the same name
#   with different metakit types.  This can cause crashing issues with
#   older version of metakit.  We catch these before hand and raise
#   an exception.
#
# 4 metakit db's cannot be weakref'd so this was removed
#
# 5 fixed a metakit.append, metakit must append lists, objects or
#   dictionaries, it can't handle scalars.
#   sv.append(int(entry))
#    became
#   sv.append((int(entry),))
#
# 6 To make it easier to compare to the other backends
#  Class.keyname is changed to Class.key
#
# 7 Fixed Class.lookup, sometimes it would claim that a valid
#  row was not valid (an _isdel row _property of 0 would
#  be reported as 1) This is because metakit's view.find
#  operation was returning bad results.  This should be
#  view.select or view.find on a single property.
#
# 8 calling create with no parameters raises a value error
#  I'm not sure if this is appropriate, but it fixes
#  a regression test :)
#
# 9 The get method was only converting results for commited
#   values.  uncommited values were not being converted
#   using the metakit conversion table
#
# 10 Added a check to the Class.__init__ to raise a ValueError
#    if the database already has a class of the same name.
#
# 11 Boolean and Number types can now have null values.  This
#    is a backwards incompatible fix in that old databases
#    won't work correctly.
#
#    The fix is simple.  For a boolean column, 0 is now None
#                                              1 is False (returns 0)
#                                              2 is True (returns 1)
#                        For a numeric column, 0 is now None
#                                              values 0 get returned as value-1
#                                              values < 0 get returned as value
#   Set the BACKWARDS_COMPATIBLE flag to False to enable this fix.
#
# 12 Enumerated READ and READWRITE for the getview and getindexedview
#    These will probably be removed because they are not used
#
# Changed to 1.56
#
# 13 worked-around a current metakit bug, so retiring properties now
#    works correctly.
#    metakit 2.9.2 has a bug when using "find" on ordered views,
#     using multiple arguments for find doesn't work.
#
# Known Current Bugs:
#    You can't change a class' key properly.
#    This shouldn't be too hard to fix.
#

from roundup import hyperdb, date, password, roundupdb, security
import metakit
from sessions import Sessions, OneTimeKeys
import re, marshal, os, sys, time, calendar
from roundup import indexer
import locking
from roundup.date import Range

# view modes for opening
# XXX FIXME BPK -> these don't do anything, they are ignored
#  should we just get rid of them for simplicities sake?
READ = 0
READWRITE = 1

# general metakit error
class MKBackendError(Exception):
    pass

_dbs = {}

def Database(config, journaltag=None):
    ''' Only have a single instance of the Database class for each instance
    '''
    db = _dbs.get(config.DATABASE, None)
    if db is None or db._db is None:
        db = _Database(config, journaltag)
        _dbs[config.DATABASE] = db
    else:
        db.journaltag = journaltag
    return db

class _Database(hyperdb.Database, roundupdb.Database):
    def __init__(self, config, journaltag=None):
        self.config = config
        self.journaltag = journaltag
        self.classes = {}
        self.dirty = 0
        self.lockfile = None
        self._db = self.__open()
        self.indexer = Indexer(self.config.DATABASE, self._db)
        self.sessions = Sessions(self.config)
        self.otks = OneTimeKeys(self.config)
        self.security = security.Security(self)

        os.umask(0002)

    def post_init(self):
        if self.indexer.should_reindex():
            self.reindex()

    def refresh_database(self):
        # XXX handle refresh
        self.reindex()

    def reindex(self):
        for klass in self.classes.values():
            for nodeid in klass.list():
                klass.index(nodeid)
        self.indexer.save_index()

    # --- defined in ping's spec
    def __getattr__(self, classname):
        if classname == 'transactions':
            return self.dirty
        # fall back on the classes
        try:
            return self.getclass(classname)
        except KeyError, msg:
            # KeyError's not appropriate here
            raise AttributeError, str(msg)
    def getclass(self, classname):
        try:
            return self.classes[classname]
        except KeyError:
            raise KeyError, 'There is no class called "%s"'%classname
    def getclasses(self):
        return self.classes.keys()
    # --- end of ping's spec 

    # --- exposed methods
    def commit(self):
        if self.dirty:
            self._db.commit()
            for cl in self.classes.values():
                cl._commit()
            self.indexer.save_index()
        self.dirty = 0
    def rollback(self):
        if self.dirty:
            for cl in self.classes.values():
                cl._rollback()
            self._db.rollback()
            self._db = None
            self._db = metakit.storage(self.dbnm, 1)
            self.hist = self._db.view('history')
            self.tables = self._db.view('tables')
            self.indexer.rollback()
            self.indexer.datadb = self._db
        self.dirty = 0
    def clearCache(self):
        for cl in self.classes.values():
            cl._commit()
    def clear(self):
        for cl in self.classes.values():
            cl._clear()
    def hasnode(self, classname, nodeid):
        return self.getclass(classname).hasnode(nodeid)
    def pack(self, pack_before):
        mindate = int(calendar.timegm(pack_before.get_tuple()))
        i = 0
        while i < len(self.hist):
            if self.hist[i].date < mindate and self.hist[i].action != _CREATE:
                self.hist.delete(i)
            else:
                i = i + 1
    def addclass(self, cl):
        self.classes[cl.classname] = cl
        if self.tables.find(name=cl.classname) < 0:
            self.tables.append(name=cl.classname)
    def addjournal(self, tablenm, nodeid, action, params, creator=None,
                creation=None):
        tblid = self.tables.find(name=tablenm)
        if tblid == -1:
            tblid = self.tables.append(name=tablenm)
        if creator is None:
            creator = int(self.getuid())
        else:
            try:
                creator = int(creator)
            except TypeError:
                creator = int(self.getclass('user').lookup(creator))
        if creation is None:
            creation = int(time.time())
        elif isinstance(creation, date.Date):
            creation = int(calendar.timegm(creation.get_tuple()))
        # tableid:I,nodeid:I,date:I,user:I,action:I,params:B
        self.hist.append(tableid=tblid,
                         nodeid=int(nodeid),
                         date=creation,
                         action=action,
                         user = creator,
                         params = marshal.dumps(params))
    def getjournal(self, tablenm, nodeid):
        rslt = []
        tblid = self.tables.find(name=tablenm)
        if tblid == -1:
            return rslt
        q = self.hist.select(tableid=tblid, nodeid=int(nodeid))
        if len(q) == 0:
            raise IndexError, "no history for id %s in %s" % (nodeid, tablenm)
        i = 0
        #userclass = self.getclass('user')
        for row in q:
            try:
                params = marshal.loads(row.params)
            except ValueError:
                print "history couldn't unmarshal %r" % row.params
                params = {}
            #usernm = userclass.get(str(row.user), 'username')
            dt = date.Date(time.gmtime(row.date))
            #rslt.append((nodeid, dt, usernm, _actionnames[row.action], params))
            rslt.append((nodeid, dt, str(row.user), _actionnames[row.action],
                params))
        return rslt

    def destroyjournal(self, tablenm, nodeid):
        nodeid = int(nodeid)
        tblid = self.tables.find(name=tablenm)
        if tblid == -1:
            return 
        i = 0
        hist = self.hist
        while i < len(hist):
            if hist[i].tableid == tblid and hist[i].nodeid == nodeid:
                hist.delete(i)
            else:
                i = i + 1
        self.dirty = 1
        
    def close(self):
        for cl in self.classes.values():
            cl.db = None
        self._db = None
        if self.lockfile is not None:
            locking.release_lock(self.lockfile)
        if _dbs.has_key(self.config.DATABASE):
            del _dbs[self.config.DATABASE]
        if self.lockfile is not None:
            self.lockfile.close()
            self.lockfile = None
        self.classes = {}

        # force the indexer to close
        self.indexer.close()
        self.indexer = None

    # --- internal
    def __open(self):
        ''' Open the metakit database
        '''
        # make the database dir if it doesn't exist
        if not os.path.exists(self.config.DATABASE):
            os.makedirs(self.config.DATABASE)

        # figure the file names
        self.dbnm = db = os.path.join(self.config.DATABASE, 'tracker.mk4')
        lockfilenm = db[:-3]+'lck'

        # get the database lock
        self.lockfile = locking.acquire_lock(lockfilenm)
        self.lockfile.write(str(os.getpid()))
        self.lockfile.flush()

        # see if the schema has changed since last db access
        self.fastopen = 0
        if os.path.exists(db):
            dbtm = os.path.getmtime(db)
            pkgnm = self.config.__name__.split('.')[0]
            schemamod = sys.modules.get(pkgnm+'.dbinit', None)
            if schemamod:
                if os.path.exists(schemamod.__file__):
                    schematm = os.path.getmtime(schemamod.__file__)
                    if schematm < dbtm:
                        # found schema mod - it's older than the db
                        self.fastopen = 1
                else:
                     # can't find schemamod - must be frozen
                    self.fastopen = 1

        # open the db
        db = metakit.storage(db, 1)
        hist = db.view('history')
        tables = db.view('tables')
        if not self.fastopen:
            # create the database if it's brand new
            if not hist.structure():
                hist = db.getas('history[tableid:I,nodeid:I,date:I,user:I,action:I,params:B]')
            if not tables.structure():
                tables = db.getas('tables[name:S]')
            db.commit()

        # we now have an open, initialised database
        self.tables = tables
        self.hist = hist
        return db

    def setid(self, classname, maxid):
        ''' No-op in metakit
        '''
        pass
        
_STRINGTYPE = type('')
_LISTTYPE = type([])
_CREATE, _SET, _RETIRE, _LINK, _UNLINK, _RESTORE = range(6)

_actionnames = {
    _CREATE : 'create',
    _SET : 'set',
    _RETIRE : 'retire',
    _RESTORE : 'restore',
    _LINK : 'link',
    _UNLINK : 'unlink',
}

_marker = []

_ALLOWSETTINGPRIVATEPROPS = 0

class Class(hyperdb.Class):
    privateprops = None
    def __init__(self, db, classname, **properties):
        if (properties.has_key('creation') or properties.has_key('activity')
            or properties.has_key('creator')):
            raise ValueError, '"creation", "activity" and "creator" are '\
                  'reserved'
        if hasattr(db, classname):
            raise ValueError, "Class %s already exists"%classname

        self.db = db
        self.classname = classname
        self.key = None
        self.ruprops = properties
        self.privateprops = { 'id' : hyperdb.String(),
                              'activity' : hyperdb.Date(),
                              'creation' : hyperdb.Date(),
                              'creator'  : hyperdb.Link('user') }

        # event -> list of callables
        self.auditors = {'create': [], 'set': [], 'retire': [], 'restore': []}
        self.reactors = {'create': [], 'set': [], 'retire': [], 'restore': []}

        view = self.__getview()
        self.maxid = 1
        if view:
            self.maxid = view[-1].id + 1
        self.uncommitted = {}
        self.rbactions = []

        # people reach inside!!
        self.properties = self.ruprops
        self.db.addclass(self)
        self.idcache = {}

        # default is to journal changes
        self.do_journal = 1

    def enableJournalling(self):
        '''Turn journalling on for this class
        '''
        self.do_journal = 1

    def disableJournalling(self):
        '''Turn journalling off for this class
        '''
        self.do_journal = 0
        
    # --- the roundup.Class methods
    def audit(self, event, detector):
        l = self.auditors[event]
        if detector not in l:
            self.auditors[event].append(detector)
    def fireAuditors(self, action, nodeid, newvalues):
        for audit in self.auditors[action]:
            audit(self.db, self, nodeid, newvalues)
    def fireReactors(self, action, nodeid, oldvalues):
        for react in self.reactors[action]:
            react(self.db, self, nodeid, oldvalues)
    def react(self, event, detector):
        l = self.reactors[event]
        if detector not in l:
            self.reactors[event].append(detector)

    # --- the hyperdb.Class methods
    def create(self, **propvalues):
        if not propvalues:
            raise ValueError, "Need something to create!"
        self.fireAuditors('create', None, propvalues)
        newid = self.create_inner(**propvalues)
        # self.set() (called in self.create_inner()) does reactors)
        return newid

    def create_inner(self, **propvalues):
        rowdict = {}
        rowdict['id'] = newid = self.maxid
        self.maxid += 1
        ndx = self.getview(READWRITE).append(rowdict)
        propvalues['#ISNEW'] = 1
        try:
            self.set(str(newid), **propvalues)
        except Exception:
            self.maxid -= 1
            raise
        return str(newid)
    
    def get(self, nodeid, propname, default=_marker, cache=1):
        '''
            'cache' exists for backwards compatibility, and is not used.
        '''
        view = self.getview()
        id = int(nodeid)
        if cache == 0:
            oldnode = self.uncommitted.get(id, None)
            if oldnode and oldnode.has_key(propname):
                raw = oldnode[propname]
                converter = _converters.get(rutyp.__class__, None)
                if converter:
                    return converter(raw)
                return raw
        ndx = self.idcache.get(id, None)

        if ndx is None:
            ndx = view.find(id=id)
            if ndx < 0:
                raise IndexError, "%s has no node %s" % (self.classname, nodeid)
            self.idcache[id] = ndx
        try:
            raw = getattr(view[ndx], propname)
        except AttributeError:
            raise KeyError, propname
        rutyp = self.ruprops.get(propname, None)

        if rutyp is None:
            rutyp = self.privateprops[propname]

        converter = _converters.get(rutyp.__class__, None)
        if converter:
            raw = converter(raw)
        return raw
        
    def set(self, nodeid, **propvalues):
        isnew = 0
        if propvalues.has_key('#ISNEW'):
            isnew = 1
            del propvalues['#ISNEW']
        if not isnew:
            self.fireAuditors('set', nodeid, propvalues)
        if not propvalues:
            return propvalues
        if propvalues.has_key('id'):
            raise KeyError, '"id" is reserved'
        if self.db.journaltag is None:
            raise hyperdb.DatabaseError, 'Database open read-only'
        view = self.getview(READWRITE)

        # node must exist & not be retired
        id = int(nodeid)
        ndx = view.find(id=id)
        if ndx < 0:
            raise IndexError, "%s has no node %s" % (self.classname, nodeid)
        row = view[ndx]
        if row._isdel:
            raise IndexError, "%s has no node %s" % (self.classname, nodeid)
        oldnode = self.uncommitted.setdefault(id, {})
        changes = {}

        for key, value in propvalues.items():
            # this will raise the KeyError if the property isn't valid
            # ... we don't use getprops() here because we only care about
            # the writeable properties.
            if _ALLOWSETTINGPRIVATEPROPS:
                prop = self.ruprops.get(key, None)
                if not prop:
                    prop = self.privateprops[key]
            else:
                prop = self.ruprops[key]
            converter = _converters.get(prop.__class__, lambda v: v)
            # if the value's the same as the existing value, no sense in
            # doing anything
            oldvalue = converter(getattr(row, key))
            if  value == oldvalue:
                del propvalues[key]
                continue
            
            # check to make sure we're not duplicating an existing key
            if key == self.key:
                iv = self.getindexview(READWRITE)
                ndx = iv.find(k=value)
                if ndx == -1:
                    iv.append(k=value, i=row.id)
                    if not isnew:
                        ndx = iv.find(k=oldvalue)
                        if ndx > -1:
                            iv.delete(ndx)
                else:
                    raise ValueError, 'node with key "%s" exists'%value

            # do stuff based on the prop type
            if isinstance(prop, hyperdb.Link):
                link_class = prop.classname
                # must be a string or None
                if value is not None and not isinstance(value, type('')):
                    raise ValueError, 'property "%s" link value be a string'%(
                        key)
                # Roundup sets to "unselected" by passing None
                if value is None:
                    value = 0   
                # if it isn't a number, it's a key
                try:
                    int(value)
                except ValueError:
                    try:
                        value = self.db.getclass(link_class).lookup(value)
                    except (TypeError, KeyError):
                        raise IndexError, 'new property "%s": %s not a %s'%(
                            key, value, prop.classname)

                if (value is not None and
                        not self.db.getclass(link_class).hasnode(value)):
                    raise IndexError, '%s has no node %s'%(link_class, value)

                setattr(row, key, int(value))
                changes[key] = oldvalue
                
                if self.do_journal and prop.do_journal:
                    # register the unlink with the old linked node
                    if oldvalue:
                        self.db.addjournal(link_class, oldvalue, _UNLINK,
                            (self.classname, str(row.id), key))

                    # register the link with the newly linked node
                    if value:
                        self.db.addjournal(link_class, value, _LINK,
                            (self.classname, str(row.id), key))

            elif isinstance(prop, hyperdb.Multilink):
                if value is not None and type(value) != _LISTTYPE:
                    raise TypeError, 'new property "%s" not a list of ids'%key
                link_class = prop.classname
                l = []
                if value is None:
                    value = []
                for entry in value:
                    if type(entry) != _STRINGTYPE:
                        raise ValueError, 'new property "%s" link value ' \
                            'must be a string'%key
                    # if it isn't a number, it's a key
                    try:
                        int(entry)
                    except ValueError:
                        try:
                            entry = self.db.getclass(link_class).lookup(entry)
                        except (TypeError, KeyError):
                            raise IndexError, 'new property "%s": %s not a %s'%(
                                key, entry, prop.classname)
                    l.append(entry)
                propvalues[key] = value = l

                # handle removals
                rmvd = []
                for id in oldvalue:
                    if id not in value:
                        rmvd.append(id)
                        # register the unlink with the old linked node
                        if self.do_journal and prop.do_journal:
                            self.db.addjournal(link_class, id, _UNLINK,
                                (self.classname, str(row.id), key))

                # handle additions
                adds = []
                for id in value:
                    if id not in oldvalue:
                        if not self.db.getclass(link_class).hasnode(id):
                            raise IndexError, '%s has no node %s'%(
                                link_class, id)
                        adds.append(id)
                        # register the link with the newly linked node
                        if self.do_journal and prop.do_journal:
                            self.db.addjournal(link_class, id, _LINK,
                                (self.classname, str(row.id), key))

                # perform the modifications on the actual property value
                sv = getattr(row, key)
                i = 0
                while i < len(sv):
                    if str(sv[i].fid) in rmvd:
                        sv.delete(i)
                    else:
                        i += 1
                for id in adds:
                    sv.append(fid=int(id))

                # figure the journal entry
                l = []
                if adds:
                    l.append(('+', adds))
                if rmvd:
                    l.append(('-', rmvd))
                if l:
                    changes[key] = tuple(l)
                #changes[key] = oldvalue

                if not rmvd and not adds:
                    del propvalues[key]

            elif isinstance(prop, hyperdb.String):
                if value is not None and type(value) != _STRINGTYPE:
                    raise TypeError, 'new property "%s" not a string'%key
                if value is None:
                    value = ''
                setattr(row, key, value)
                changes[key] = oldvalue
                if hasattr(prop, 'isfilename') and prop.isfilename:
                    propvalues[key] = os.path.basename(value)
                if prop.indexme:
                    self.db.indexer.add_text((self.classname, nodeid, key),
                        value, 'text/plain')

            elif isinstance(prop, hyperdb.Password):
                if value is not None and not isinstance(value, password.Password):
                    raise TypeError, 'new property "%s" not a Password'% key
                if value is None:
                    value = ''
                setattr(row, key, str(value))
                changes[key] = str(oldvalue)
                propvalues[key] = str(value)

            elif isinstance(prop, hyperdb.Date):
                if value is not None and not isinstance(value, date.Date):
                    raise TypeError, 'new property "%s" not a Date'% key
                if value is None:
                    setattr(row, key, 0)
                else:
                    setattr(row, key, int(calendar.timegm(value.get_tuple())))
                changes[key] = str(oldvalue)
                propvalues[key] = str(value)

            elif isinstance(prop, hyperdb.Interval):
                if value is not None and not isinstance(value, date.Interval):
                    raise TypeError, 'new property "%s" not an Interval'% key
                if value is None:
                    setattr(row, key, '')
                else:
                    # kedder: we should store interval values serialized
                    setattr(row, key, value.serialise())
                changes[key] = str(oldvalue)
                propvalues[key] = str(value)
 
            elif isinstance(prop, hyperdb.Number):
                if value is None:
                    v = 0
                else:
                    try:
                        v = int(value)
                    except ValueError:
                        raise TypeError, "%s (%s) is not numeric"%(key, repr(value))
                    if not BACKWARDS_COMPATIBLE:
                        if v >=0:
                            v = v + 1
                setattr(row, key, v)
                changes[key] = oldvalue
                propvalues[key] = value

            elif isinstance(prop, hyperdb.Boolean):
                if value is None:
                    bv = 0
                elif value not in (0,1):
                    raise TypeError, "%s (%s) is not boolean"%(key, repr(value))
                else:
                    bv = value
                    if not BACKWARDS_COMPATIBLE:
                        bv += 1
                setattr(row, key, bv)
                changes[key] = oldvalue
                propvalues[key] = value

            oldnode[key] = oldvalue

        # nothing to do?
        if not propvalues:
            return propvalues
        if not propvalues.has_key('activity'):
            row.activity = int(time.time())
        if isnew:
            if not row.creation:
                row.creation = int(time.time())
            if not row.creator:
                row.creator = int(self.db.getuid())

        self.db.dirty = 1
        if self.do_journal:
            if isnew:
                self.db.addjournal(self.classname, nodeid, _CREATE, {})
                self.fireReactors('create', nodeid, None)
            else:
                self.db.addjournal(self.classname, nodeid, _SET, changes)
                self.fireReactors('set', nodeid, oldnode)

        return propvalues
    
    def retire(self, nodeid):
        if self.db.journaltag is None:
            raise hyperdb.DatabaseError, 'Database open read-only'
        self.fireAuditors('retire', nodeid, None)
        view = self.getview(READWRITE)
        ndx = view.find(id=int(nodeid))
        if ndx < 0:
            raise KeyError, "nodeid %s not found" % nodeid

        row = view[ndx]
        oldvalues = self.uncommitted.setdefault(row.id, {})
        oldval = oldvalues['_isdel'] = row._isdel
        row._isdel = 1

        if self.do_journal:
            self.db.addjournal(self.classname, nodeid, _RETIRE, {})
        if self.key:
            iv = self.getindexview(READWRITE)
            ndx = iv.find(k=getattr(row, self.key))
            # find is broken with multiple attribute lookups
            # on ordered views
            #ndx = iv.find(k=getattr(row, self.key),i=row.id)
            if ndx > -1 and iv[ndx].i == row.id:
                iv.delete(ndx)

        self.db.dirty = 1
        self.fireReactors('retire', nodeid, None)

    def restore(self, nodeid):
        """Restore a retired node.

        Make node available for all operations like it was before retirement.
        """
        if self.db.journaltag is None:
            raise hyperdb.DatabaseError, 'Database open read-only'

        # check if key property was overrided
        key = self.getkey()
        keyvalue = self.get(nodeid, key)
        
        try:
            id = self.lookup(keyvalue)
        except KeyError:
            pass
        else:
            raise KeyError, "Key property (%s) of retired node clashes with \
                existing one (%s)" % (key, keyvalue)
        # Now we can safely restore node
        self.fireAuditors('restore', nodeid, None)
        view = self.getview(READWRITE)
        ndx = view.find(id=int(nodeid))
        if ndx < 0:
            raise KeyError, "nodeid %s not found" % nodeid

        row = view[ndx]
        oldvalues = self.uncommitted.setdefault(row.id, {})
        oldval = oldvalues['_isdel'] = row._isdel
        row._isdel = 0

        if self.do_journal:
            self.db.addjournal(self.classname, nodeid, _RESTORE, {})
        if self.key:
            iv = self.getindexview(READWRITE)
            ndx = iv.find(k=getattr(row, self.key),i=row.id)
            if ndx > -1:
                iv.delete(ndx)
        self.db.dirty = 1
        self.fireReactors('restore', nodeid, None)

    def is_retired(self, nodeid):
        view = self.getview(READWRITE)
        # node must exist & not be retired
        id = int(nodeid)
        ndx = view.find(id=id)
        if ndx < 0:
            raise IndexError, "%s has no node %s" % (self.classname, nodeid)
        row = view[ndx]
        return row._isdel

    def history(self, nodeid):
        if not self.do_journal:
            raise ValueError, 'Journalling is disabled for this class'
        return self.db.getjournal(self.classname, nodeid)

    def setkey(self, propname):
        if self.key:
            if propname == self.key:
                return
            raise ValueError, "%s already indexed on %s"%(self.classname,
                self.key)
        prop = self.properties.get(propname, None)
        if prop is None:
            prop = self.privateprops.get(propname, None)
        if prop is None:
            raise KeyError, "no property %s" % propname
        if not isinstance(prop, hyperdb.String):
            raise TypeError, "%s is not a String" % propname

        # TODO: metakit needs to be able to cope with the key property
        # *changing*, which it can't do at present. At the moment, it
        # creates the key prop index once, with no record of the name of
        # the property for the index.

        # first setkey for this run
        self.key = propname
        iv = self.db._db.view('_%s' % self.classname)
        if self.db.fastopen and iv.structure():
            return

        # very first setkey ever
        self.db.dirty = 1
        iv = self.db._db.getas('_%s[k:S,i:I]' % self.classname)
        iv = iv.ordered(1)
        for row in self.getview():
            iv.append(k=getattr(row, propname), i=row.id)
        self.db.commit()

    def getkey(self):
        return self.key

    def lookup(self, keyvalue):
        """Locate a particular node by its key property and return its id.

        If this class has no key property, a TypeError is raised.  If the
        keyvalue matches one of the values for the key property among
        the nodes in this class, the matching node's id is returned;
        otherwise a KeyError is raised.
        """
        if not self.key:
            raise TypeError, 'No key property set for class %s'%self.classname
        
        if type(keyvalue) is not _STRINGTYPE:
            raise TypeError, '%r is not a string'%keyvalue

        # XXX FIX ME -> this is a bit convoluted
        # First we search the index view to get the id
        # which is a quicker look up.
        # Then we lookup the row with id=id
        # if the _isdel property of the row is 0, return the
        # string version of the id. (Why string version???)
        #
        # Otherwise, just lookup the non-indexed key
        # in the non-index table and check the _isdel property
        iv = self.getindexview()
        if iv:
            # look up the index view for the id,
            # then instead of looking up the keyvalue, lookup the
            # quicker id
            ndx = iv.find(k=keyvalue)
            if ndx > -1:
                view = self.getview()
                ndx = view.find(id=iv[ndx].i)
                if ndx > -1:
                    row = view[ndx]
                    if not row._isdel:
                        return str(row.id)
        else:
            # perform the slower query
            view = self.getview()
            ndx = view.find({self.key:keyvalue})
            if ndx > -1:
                row = view[ndx]
                if not row._isdel:
                    return str(row.id)

        raise KeyError, keyvalue

    def destroy(self, id):
        view = self.getview(READWRITE)
        ndx = view.find(id=int(id))
        if ndx > -1:
            if self.key:
                keyvalue = getattr(view[ndx], self.key)
                iv = self.getindexview(READWRITE)
                if iv:
                    ivndx = iv.find(k=keyvalue)
                    if ivndx > -1:
                        iv.delete(ivndx)
            view.delete(ndx)
            self.db.destroyjournal(self.classname, id)
            self.db.dirty = 1
        
    def find(self, **propspec):
        """Get the ids of nodes in this class which link to the given nodes.

        'propspec' consists of keyword args propname={nodeid:1,}   
        'propname' must be the name of a property in this class, or a
                   KeyError is raised.  That property must be a Link or
                   Multilink property, or a TypeError is raised.

        Any node in this class whose propname property links to any of the
        nodeids will be returned. Used by the full text indexing, which knows
        that "foo" occurs in msg1, msg3 and file7; so we have hits on these
        issues:

            db.issue.find(messages={'1':1,'3':1}, files={'7':1})

        """
        propspec = propspec.items()
        for propname, nodeid in propspec:
            # check the prop is OK
            prop = self.ruprops[propname]
            if (not isinstance(prop, hyperdb.Link) and
                    not isinstance(prop, hyperdb.Multilink)):
                raise TypeError, "'%s' not a Link/Multilink property"%propname

        vws = []
        for propname, ids in propspec:
            if type(ids) is _STRINGTYPE:
                ids = {int(ids):1}
            elif ids is None:
                ids = {0:1}
            else:
                d = {}
                for id in ids.keys():
                    if id is None:
                        d[0] = 1
                    else:
                        d[int(id)] = 1
                ids = d
            prop = self.ruprops[propname]
            view = self.getview()
            if isinstance(prop, hyperdb.Multilink):
                def ff(row, nm=propname, ids=ids):
                    if not row._isdel:
                        sv = getattr(row, nm)
                        for sr in sv:
                            if ids.has_key(sr.fid):
                                return 1
                    return 0
            else:
                def ff(row, nm=propname, ids=ids):
                    return not row._isdel and ids.has_key(getattr(row, nm))
            ndxview = view.filter(ff)
            vws.append(ndxview.unique())

        # handle the empty match case
        if not vws:
            return []

        ndxview = vws[0]
        for v in vws[1:]:
            ndxview = ndxview.union(v)
        view = self.getview().remapwith(ndxview)
        rslt = []
        for row in view:
            rslt.append(str(row.id))
        return rslt
            

    def list(self):
        l = []
        for row in self.getview().select(_isdel=0):
            l.append(str(row.id))
        return l

    def getnodeids(self):
        l = []
        for row in self.getview():
            l.append(str(row.id))
        return l

    def count(self):
        return len(self.getview())

    def getprops(self, protected=1):
        # protected is not in ping's spec
        allprops = self.ruprops.copy()
        if protected and self.privateprops is not None:
            allprops.update(self.privateprops)
        return allprops

    def addprop(self, **properties):
        for key in properties.keys():
            if self.ruprops.has_key(key):
                raise ValueError, "%s is already a property of %s"%(key,
                    self.classname)
        self.ruprops.update(properties)
        # Class structure has changed
        self.db.fastopen = 0
        view = self.__getview()
        self.db.commit()
    # ---- end of ping's spec

    def filter(self, search_matches, filterspec, sort=(None,None),
            group=(None,None)):
        # search_matches is None or a set (dict of {nodeid: {propname:[nodeid,...]}})
        # filterspec is a dict {propname:value}
        # sort and group are (dir, prop) where dir is '+', '-' or None
        #                    and prop is a prop name or None

        timezone = self.db.getUserTimezone()

        where = {'_isdel':0}
        wherehigh = {}
        mlcriteria = {}
        regexes = {}
        orcriteria = {}
        for propname, value in filterspec.items():
            prop = self.ruprops.get(propname, None)
            if prop is None:
                prop = self.privateprops[propname]
            if isinstance(prop, hyperdb.Multilink):
                if value in ('-1', ['-1']):
                    value = []
                elif type(value) is not _LISTTYPE:
                    value = [value]
                # transform keys to ids
                u = []
                for item in value:
                    try:
                        item = int(item)
                    except (TypeError, ValueError):
                        item = int(self.db.getclass(prop.classname).lookup(item))
                    if item == -1:
                        item = 0
                    u.append(item)
                mlcriteria[propname] = u
            elif isinstance(prop, hyperdb.Link):
                if type(value) is not _LISTTYPE:
                    value = [value]
                # transform keys to ids
                u = []
                for item in value:
                    try:
                        item = int(item)
                    except (TypeError, ValueError):
                        item = int(self.db.getclass(prop.classname).lookup(item))
                    if item == -1:
                        item = 0
                    u.append(item)
                if len(u) == 1:
                    where[propname] = u[0]
                else:
                    orcriteria[propname] = u
            elif isinstance(prop, hyperdb.String):
                if type(value) is not type([]):
                    value = [value]
                m = []
                for v in value:
                    # simple glob searching
                    v = re.sub(r'([\|\{\}\\\.\+\[\]\(\)])', r'\\\1', v)
                    v = v.replace('?', '.')
                    v = v.replace('*', '.*?')
                    m.append(v)
                regexes[propname] = re.compile('(%s)'%('|'.join(m)), re.I)
            elif propname == 'id':
                where[propname] = int(value)
            elif isinstance(prop, hyperdb.Boolean):
                if type(value) is _STRINGTYPE:
                    bv = value.lower() in ('yes', 'true', 'on', '1')
                else:
                    bv = value
                where[propname] = bv
            elif isinstance(prop, hyperdb.Date):
                try:
                    # Try to filter on range of dates
                    date_rng = Range(value, date.Date, offset=timezone)
                    if date_rng.from_value:
                        t = date_rng.from_value.get_tuple()
                        where[propname] = int(calendar.timegm(t))
                    else:
                        # use minimum possible value to exclude items without
                        # 'prop' property
                        where[propname] = 0
                    if date_rng.to_value:
                        t = date_rng.to_value.get_tuple()
                        wherehigh[propname] = int(calendar.timegm(t))
                    else:
                        wherehigh[propname] = None
                except ValueError:
                    # If range creation fails - ignore that search parameter
                    pass                        
            elif isinstance(prop, hyperdb.Interval):
                try:
                    # Try to filter on range of intervals
                    date_rng = Range(value, date.Interval)
                    if date_rng.from_value:
                        #t = date_rng.from_value.get_tuple()
                        where[propname] = date_rng.from_value.serialise()
                    else:
                        # use minimum possible value to exclude items without
                        # 'prop' property
                        where[propname] = '-99999999999999'
                    if date_rng.to_value:
                        #t = date_rng.to_value.get_tuple()
                        wherehigh[propname] = date_rng.to_value.serialise()
                    else:
                        wherehigh[propname] = None
                except ValueError:
                    # If range creation fails - ignore that search parameter
                    pass                        
            elif isinstance(prop, hyperdb.Number):
                where[propname] = int(value)
            else:
                where[propname] = str(value)
        v = self.getview()
        #print "filter start at  %s" % time.time() 
        if where:
            where_higherbound = where.copy()
            where_higherbound.update(wherehigh)
            v = v.select(where, where_higherbound)
        #print "filter where at  %s" % time.time() 

        if mlcriteria:
            # multilink - if any of the nodeids required by the
            # filterspec aren't in this node's property, then skip it
            def ff(row, ml=mlcriteria):
                for propname, values in ml.items():
                    sv = getattr(row, propname)
                    if not values and sv:
                        return 0
                    for id in values:
                        if sv.find(fid=id) == -1:
                            return 0
                return 1
            iv = v.filter(ff)
            v = v.remapwith(iv)

        #print "filter mlcrit at %s" % time.time() 
        
        if orcriteria:
            def ff(row, crit=orcriteria):
                for propname, allowed in crit.items():
                    val = getattr(row, propname)
                    if val not in allowed:
                        return 0
                return 1
            
            iv = v.filter(ff)
            v = v.remapwith(iv)
        
        #print "filter orcrit at %s" % time.time() 
        if regexes:
            def ff(row, r=regexes):
                for propname, regex in r.items():
                    val = str(getattr(row, propname))
                    if not regex.search(val):
                        return 0
                return 1
            
            iv = v.filter(ff)
            v = v.remapwith(iv)
        #print "filter regexs at %s" % time.time() 
        
        if sort or group:
            sortspec = []
            rev = []
            for dir, propname in group, sort:
                if propname is None: continue
                isreversed = 0
                if dir == '-':
                    isreversed = 1
                try:
                    prop = getattr(v, propname)
                except AttributeError:
                    print "MK has no property %s" % propname
                    continue
                propclass = self.ruprops.get(propname, None)
                if propclass is None:
                    propclass = self.privateprops.get(propname, None)
                    if propclass is None:
                        print "Schema has no property %s" % propname
                        continue
                if isinstance(propclass, hyperdb.Link):
                    linkclass = self.db.getclass(propclass.classname)
                    lv = linkclass.getview()
                    lv = lv.rename('id', propname)
                    v = v.join(lv, prop, 1)
                    if linkclass.getprops().has_key('order'):
                        propname = 'order'
                    else:
                        propname = linkclass.labelprop()
                    prop = getattr(v, propname)
                if isreversed:
                    rev.append(prop)
                sortspec.append(prop)
            v = v.sortrev(sortspec, rev)[:] #XXX Metakit bug
        #print "filter sort   at %s" % time.time() 
            
        rslt = []
        for row in v:
            id = str(row.id)
            if search_matches is not None:
                if search_matches.has_key(id):
                    rslt.append(id)
            else:
                rslt.append(id)
        return rslt
    
    def hasnode(self, nodeid):
        return int(nodeid) < self.maxid
    
    def labelprop(self, default_to_id=0):
        ''' Return the property name for a label for the given node.

        This method attempts to generate a consistent label for the node.
        It tries the following in order:
            1. key property
            2. "name" property
            3. "title" property
            4. first property from the sorted property name list
        '''
        k = self.getkey()
        if  k:
            return k
        props = self.getprops()
        if props.has_key('name'):
            return 'name'
        elif props.has_key('title'):
            return 'title'
        if default_to_id:
            return 'id'
        props = props.keys()
        props.sort()
        return props[0]

    def stringFind(self, **requirements):
        """Locate a particular node by matching a set of its String
        properties in a caseless search.

        If the property is not a String property, a TypeError is raised.
        
        The return is a list of the id of all nodes that match.
        """
        for propname in requirements.keys():
            prop = self.properties[propname]
            if isinstance(not prop, hyperdb.String):
                raise TypeError, "'%s' not a String property"%propname
            requirements[propname] = requirements[propname].lower()
        requirements['_isdel'] = 0
        
        l = []
        for row in self.getview().select(requirements):
            l.append(str(row.id))
        return l

    def addjournal(self, nodeid, action, params):
        self.db.addjournal(self.classname, nodeid, action, params)

    def index(self, nodeid):
        ''' Add (or refresh) the node to search indexes '''
        # find all the String properties that have indexme
        for prop, propclass in self.getprops().items():
            if isinstance(propclass, hyperdb.String) and propclass.indexme:
                # index them under (classname, nodeid, property)
                self.db.indexer.add_text((self.classname, nodeid, prop),
                                str(self.get(nodeid, prop)))

    def export_list(self, propnames, nodeid):
        ''' Export a node - generate a list of CSV-able data in the order
            specified by propnames for the given node.
        '''
        properties = self.getprops()
        l = []
        for prop in propnames:
            proptype = properties[prop]
            value = self.get(nodeid, prop)
            # "marshal" data where needed
            if value is None:
                pass
            elif isinstance(proptype, hyperdb.Date):
                value = value.get_tuple()
            elif isinstance(proptype, hyperdb.Interval):
                value = value.get_tuple()
            elif isinstance(proptype, hyperdb.Password):
                value = str(value)
            l.append(repr(value))

        # append retired flag
        l.append(repr(self.is_retired(nodeid)))

        return l
        
    def import_list(self, propnames, proplist):
        ''' Import a node - all information including "id" is present and
            should not be sanity checked. Triggers are not triggered. The
            journal should be initialised using the "creator" and "creation"
            information.

            Return the nodeid of the node imported.
        '''
        if self.db.journaltag is None:
            raise hyperdb.DatabaseError, 'Database open read-only'
        properties = self.getprops()

        d = {}
        view = self.getview(READWRITE)
        for i in range(len(propnames)):
            value = eval(proplist[i])
            if not value:
                continue

            propname = propnames[i]
            if propname == 'id':
                newid = value = int(value)
            elif propname == 'is retired':
                # is the item retired?
                if int(value):
                    d['_isdel'] = 1
                continue
            elif value is None:
                d[propname] = None
                continue

            prop = properties[propname]
            if isinstance(prop, hyperdb.Date):
                value = int(calendar.timegm(value))
            elif isinstance(prop, hyperdb.Interval):
                value = date.Interval(value).serialise()
            elif isinstance(prop, hyperdb.Number):
                value = int(value)
            elif isinstance(prop, hyperdb.Boolean):
                value = int(value)
            elif isinstance(prop, hyperdb.Link) and value:
                value = int(value)
            elif isinstance(prop, hyperdb.Multilink):
                # we handle multilinks separately
                continue
            d[propname] = value

        # possibly make a new node
        if not d.has_key('id'):
            d['id'] = newid = self.maxid
            self.maxid += 1

        # save off the node
        view.append(d)

        # fix up multilinks
        ndx = view.find(id=newid)
        row = view[ndx]
        for i in range(len(propnames)):
            value = eval(proplist[i])
            propname = propnames[i]
            if propname == 'is retired':
                continue
            prop = properties[propname]
            if not isinstance(prop, hyperdb.Multilink):
                continue
            sv = getattr(row, propname)
            for entry in value:
                sv.append((int(entry),))

        self.db.dirty = 1
        creator = d.get('creator', 0)
        creation = d.get('creation', 0)
        self.db.addjournal(self.classname, str(newid), _CREATE, {}, creator,
            creation)
        return newid

    # --- used by Database
    def _commit(self):
        """ called post commit of the DB.
            interested subclasses may override """
        self.uncommitted = {}
        self.rbactions = []
        self.idcache = {}
    def _rollback(self):  
        """ called pre rollback of the DB.
            interested subclasses may override """
        for action in self.rbactions:
            action()
        self.rbactions = []
        self.uncommitted = {}
        self.idcache = {}
    def _clear(self):
        view = self.getview(READWRITE)
        if len(view):
            view[:] = []
            self.db.dirty = 1
        iv = self.getindexview(READWRITE)
        if iv:
            iv[:] = []
    def rollbackaction(self, action):
        """ call this to register a callback called on rollback
            callback is removed on end of transaction """
        self.rbactions.append(action)
    # --- internal
    def __getview(self):
        ''' Find the interface for a specific Class in the hyperdb.

            This method checks to see whether the schema has changed and
            re-works the underlying metakit structure if it has.
        '''
        db = self.db._db
        view = db.view(self.classname)
        mkprops = view.structure()

        # if we have structure in the database, and the structure hasn't
        # changed
        # note on view.ordered ->        
        # return a metakit view ordered on the id column
        # id is always the first column.  This speeds up
        # look-ups on the id column.
        
        if mkprops and self.db.fastopen:
            return view.ordered(1)

        # is the definition the same?
        for nm, rutyp in self.ruprops.items():
            for mkprop in mkprops:
                if mkprop.name == nm:
                    break
            else:
                mkprop = None
            if mkprop is None:
                break
            if _typmap[rutyp.__class__] != mkprop.type:
                break
        else:
            
            return view.ordered(1)
        # The schema has changed.  We need to create or restructure the mk view
        # id comes first, so we can use view.ordered(1) so that
        # MK will order it for us to allow binary-search quick lookups on
        # the id column
        self.db.dirty = 1
        s = ["%s[id:I" % self.classname]

        # these columns will always be added, we can't trample them :)
        _columns = {"id":"I", "_isdel":"I", "activity":"I", "creation":"I", "creator":"I"}

        for nm, rutyp in self.ruprops.items():
            mktyp = _typmap[rutyp.__class__].upper()
            if nm in _columns and _columns[nm] != mktyp:
                # oops, two columns with the same name and different properties
               raise MKBackendError("column %s for table %sis defined with multiple types"%(nm, self.classname))
            _columns[nm] = mktyp
            s.append('%s:%s' % (nm, mktyp))
            if mktyp == 'V':
                s[-1] += ('[fid:I]')

        # XXX FIX ME -> in some tests, creation:I becomes creation:S is this
        # okay?  Does this need to be supported?
        s.append('_isdel:I,activity:I,creation:I,creator:I]')
        view = self.db._db.getas(','.join(s))
        self.db.commit()
        return view.ordered(1)
    def getview(self, RW=0):
        # XXX FIX ME -> The RW flag doesn't do anything.
        return self.db._db.view(self.classname).ordered(1)
    def getindexview(self, RW=0):
        # XXX FIX ME -> The RW flag doesn't do anything.
        return self.db._db.view("_%s" % self.classname).ordered(1)

def _fetchML(sv):
    l = []
    for row in sv:
        if row.fid:
            l.append(str(row.fid))
    return l

def _fetchPW(s):
    ''' Convert to a password.Password unless the password is '' which is
        our sentinel for "unset".
    '''
    if s == '':
        return None
    p = password.Password()
    p.unpack(s)
    return p

def _fetchLink(n):
    ''' Return None if the link is 0 - otherwise strify it.
    '''
    return n and str(n) or None

def _fetchDate(n):
    ''' Convert the timestamp to a date.Date instance - unless it's 0 which
        is our sentinel for "unset".
    '''
    if n == 0:
        return None
    return date.Date(time.gmtime(n))

def _fetchInterval(n):
    ''' Convert to a date.Interval unless the interval is '' which is our
        sentinel for "unset".
    '''
    if n == '':
        return None
    return date.Interval(n)

# Converters for boolean and numbers to properly
# return None values.
# These are in conjunction with the setters above
#  look for hyperdb.Boolean and hyperdb.Number
if BACKWARDS_COMPATIBLE:
    def getBoolean(bool): return bool
    def getNumber(number): return number
else:
    def getBoolean(bool):
        if not bool: res = None
        else: res = bool - 1
        return res
    
    def getNumber(number):
        if number == 0: res = None
        elif number < 0: res = number
        else: res = number - 1
        return res

_converters = {
    hyperdb.Date   : _fetchDate,
    hyperdb.Link   : _fetchLink,
    hyperdb.Multilink : _fetchML,
    hyperdb.Interval  : _fetchInterval,
    hyperdb.Password  : _fetchPW,
    hyperdb.Boolean   : getBoolean,
    hyperdb.Number    : getNumber,
    hyperdb.String    : lambda s: s and str(s) or None,
}                

class FileName(hyperdb.String):
    isfilename = 1            

_typmap = {
    FileName : 'S',
    hyperdb.String : 'S',
    hyperdb.Date   : 'I',
    hyperdb.Link   : 'I',
    hyperdb.Multilink : 'V',
    hyperdb.Interval  : 'S',
    hyperdb.Password  : 'S',
    hyperdb.Boolean   : 'I',
    hyperdb.Number    : 'I',
}
class FileClass(Class, hyperdb.FileClass):
    ''' like Class but with a content property
    '''
    default_mime_type = 'text/plain'
    def __init__(self, db, classname, **properties):
        properties['content'] = FileName()
        if not properties.has_key('type'):
            properties['type'] = hyperdb.String()
        Class.__init__(self, db, classname, **properties)

    def get(self, nodeid, propname, default=_marker, cache=1):
        x = Class.get(self, nodeid, propname, default)
        poss_msg = 'Possibly an access right configuration problem.'
        if propname == 'content':
            if x.startswith('file:'):
                fnm = x[5:]
                try:
                    f = open(fnm, 'rb')
                except IOError, (strerror):
                    # XXX by catching this we donot see an error in the log.
                    return 'ERROR reading file: %s%s\n%s\n%s'%(
                            self.classname, nodeid, poss_msg, strerror)
                x = f.read()
                f.close()
        return x

    def create(self, **propvalues):
        if not propvalues:
            raise ValueError, "Need something to create!"
        self.fireAuditors('create', None, propvalues)
        content = propvalues['content']
        del propvalues['content']
        newid = Class.create_inner(self, **propvalues)
        if not content:
            return newid
        nm = bnm = '%s%s' % (self.classname, newid)
        sd = str(int(int(newid) / 1000))
        d = os.path.join(self.db.config.DATABASE, 'files', self.classname, sd)
        if not os.path.exists(d):
            os.makedirs(d)
        nm = os.path.join(d, nm)
        open(nm, 'wb').write(content)
        self.set(newid, content = 'file:'+nm)
        mimetype = propvalues.get('type', self.default_mime_type)
        self.db.indexer.add_text((self.classname, newid, 'content'), content,
            mimetype)
        def undo(fnm=nm, action1=os.remove, indexer=self.db.indexer):
            action1(fnm)
        self.rollbackaction(undo)
        return newid

    def index(self, nodeid):
        Class.index(self, nodeid)
        mimetype = self.get(nodeid, 'type')
        if not mimetype:
            mimetype = self.default_mime_type
        self.db.indexer.add_text((self.classname, nodeid, 'content'),
                    self.get(nodeid, 'content'), mimetype)
 
class IssueClass(Class, roundupdb.IssueClass):
    ''' The newly-created class automatically includes the "messages",
        "files", "nosy", and "superseder" properties.  If the 'properties'
        dictionary attempts to specify any of these properties or a
        "creation" or "activity" property, a ValueError is raised.
    '''
    def __init__(self, db, classname, **properties):
        if not properties.has_key('title'):
            properties['title'] = hyperdb.String(indexme='yes')
        if not properties.has_key('messages'):
            properties['messages'] = hyperdb.Multilink("msg")
        if not properties.has_key('files'):
            properties['files'] = hyperdb.Multilink("file")
        if not properties.has_key('nosy'):
            # note: journalling is turned off as it really just wastes
            # space. this behaviour may be overridden in an instance
            properties['nosy'] = hyperdb.Multilink("user", do_journal="no")
        if not properties.has_key('superseder'):
            properties['superseder'] = hyperdb.Multilink(classname)
        Class.__init__(self, db, classname, **properties)
        
CURVERSION = 2

class Indexer(indexer.Indexer):
    disallows = {'THE':1, 'THIS':1, 'ZZZ':1, 'THAT':1, 'WITH':1}
    def __init__(self, path, datadb):
        self.path = os.path.join(path, 'index.mk4')
        self.db = metakit.storage(self.path, 1)
        self.datadb = datadb
        self.reindex = 0
        v = self.db.view('version')
        if not v.structure():
            v = self.db.getas('version[vers:I]')
            self.db.commit()
            v.append(vers=CURVERSION)
            self.reindex = 1
        elif v[0].vers != CURVERSION:
            v[0].vers = CURVERSION
            self.reindex = 1
        if self.reindex:
            self.db.getas('ids[tblid:I,nodeid:I,propid:I,ignore:I]')
            self.db.getas('index[word:S,hits[pos:I]]')
            self.db.commit()
            self.reindex = 1
        self.changed = 0
        self.propcache = {}

    def close(self):
        """close the indexing database"""
        del self.db
        self.db = None
  
    def force_reindex(self):
        """Force a reindexing of the database.  This essentially
        empties the tables ids and index and sets a flag so
        that the databases are reindexed"""
        v = self.db.view('ids')
        v[:] = []
        v = self.db.view('index')
        v[:] = []
        self.db.commit()
        self.reindex = 1

    def should_reindex(self):
        """returns True if the indexes need to be rebuilt"""
        return self.reindex

    def _getprops(self, classname):
        props = self.propcache.get(classname, None)
        if props is None:
            props = self.datadb.view(classname).structure()
            props = [prop.name for prop in props]
            self.propcache[classname] = props
        return props

    def _getpropid(self, classname, propname):
        return self._getprops(classname).index(propname)

    def _getpropname(self, classname, propid):
        return self._getprops(classname)[propid]

    def add_text(self, identifier, text, mime_type='text/plain'):
        if mime_type != 'text/plain':
            return
        classname, nodeid, property = identifier
        tbls = self.datadb.view('tables')
        tblid = tbls.find(name=classname)
        if tblid < 0:
            raise KeyError, "unknown class %r"%classname
        nodeid = int(nodeid)
        propid = self._getpropid(classname, property)
        ids = self.db.view('ids')
        oldpos = ids.find(tblid=tblid,nodeid=nodeid,propid=propid,ignore=0)
        if oldpos > -1:
            ids[oldpos].ignore = 1
            self.changed = 1
        pos = ids.append(tblid=tblid,nodeid=nodeid,propid=propid)
        
        wordlist = re.findall(r'\b\w{2,25}\b', text.upper())
        words = {}
        for word in wordlist:
	    if not self.disallows.has_key(word):
            	words[word] = 1
        words = words.keys()
        
        index = self.db.view('index').ordered(1)
        for word in words:
            ndx = index.find(word=word)
            if ndx < 0:
                index.append(word=word)
                ndx = index.find(word=word)
            index[ndx].hits.append(pos=pos)
            self.changed = 1

    def find(self, wordlist):
        """look up all the words in the wordlist.
        If none are found return an empty dictionary
        * more rules here
        """        
        hits = None
        index = self.db.view('index').ordered(1)
        for word in wordlist:
            word = word.upper()
            if not 2 < len(word) < 26:
                continue
            ndx = index.find(word=word)
            if ndx < 0:
                return {}
            if hits is None:
                hits = index[ndx].hits
            else:
                hits = hits.intersect(index[ndx].hits)
            if len(hits) == 0:
                return {}
        if hits is None:
            return {}
        rslt = {}
        ids = self.db.view('ids').remapwith(hits)
        tbls = self.datadb.view('tables')
        for i in range(len(ids)):
            hit = ids[i]
            if not hit.ignore:
                classname = tbls[hit.tblid].name
                nodeid = str(hit.nodeid)
                property = self._getpropname(classname, hit.propid)
                rslt[i] = (classname, nodeid, property)
        return rslt

    def save_index(self):
        if self.changed:
            self.db.commit()
        self.changed = 0

    def rollback(self):
        if self.changed:
            self.db.rollback()
            self.db = metakit.storage(self.path, 1)
        self.changed = 0