Limnoria/src/dbi.py

434 lines
13 KiB
Python
Raw Normal View History

2005-01-19 14:14:38 +01:00
###
2005-01-19 14:33:05 +01:00
# Copyright (c) 2002-2005, Jeremiah Fincher
2005-01-19 14:14:38 +01:00
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice,
# this list of conditions, and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions, and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of the author of this software nor the name of
# contributors to this software may be used to endorse or promote products
# derived from this software without specific prior written consent.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
# POSSIBILITY OF SUCH DAMAGE.
###
"""
Module for some slight database-independence for simple databases.
"""
import csv
import math
import supybot.cdb as cdb
import supybot.utils as utils
from supybot.utils.iter import ilen
2005-01-19 14:14:38 +01:00
class Error(Exception):
"""General error for this module."""
class NoRecordError(KeyError):
pass
class InvalidDBError(Exception):
pass
class MappingInterface(object):
"""This is a class to represent the underlying representation of a map
from integer keys to strings."""
def __init__(self, filename, **kwargs):
"""Feel free to ignore the filename."""
raise NotImplementedError
def get(id):
"""Gets the record matching id. Raises NoRecordError otherwise."""
raise NotImplementedError
def set(id, s):
"""Sets the record matching id to s."""
raise NotImplementedError
def add(self, s):
"""Adds a new record, returning a new id for it."""
raise NotImplementedError
def remove(self, id):
"Returns and removes the record with the given id from the database."
raise NotImplementedError
def __iter__(self):
"Return an iterator over (id, s) pairs. Not required to be ordered."
raise NotImplementedError
def flush(self):
"""Flushes current state to disk."""
raise NotImplementedError
def close(self):
"""Flushes current state to disk and invalidates the Mapping."""
raise NotImplementedError
def vacuum(self):
"Cleans up in the database, if possible. Not required to do anything."
pass
class DirMapping(MappingInterface):
def __init__(self, filename, **kwargs):
self.dirname = filename
if not os.path.exists(self.dirname):
os.mkdir(self.dirname)
if not os.path.exists(os.path.join(self.dirname, 'max')):
self._setMax(1)
def _setMax(self, id):
fd = file(os.path.join(self.dirname, 'max'), 'w')
try:
fd.write(str(id))
finally:
fd.close()
def _getMax(self):
fd = file(os.path.join(self.dirname, 'max'))
try:
i = int(fd.read())
return i
finally:
fd.close()
def _makeFilename(self, id):
return os.path.join(self.dirname, str(id))
def get(id):
try:
fd = file(self._makeFilename(id))
return fd.read()
except EnvironmentError, e:
exn = NoRecordError(id)
exn.realException = e
raise exn
def set(id, s):
fd = file(self._makeFilename(id), 'w')
fd.write(s)
fd.close()
def add(self, s):
id = self._getMax()
fd = file(self._makeFilename(id), 'w')
try:
fd.write(s)
return id
finally:
fd.close()
def remove(self, id):
try:
os.remove(self._makeFilename(id))
except EnvironmentError, e:
raise NoRecordError, id
class FlatfileMapping(MappingInterface):
def __init__(self, filename, maxSize=10**6):
self.filename = filename
try:
fd = file(self.filename)
strId = fd.readline().rstrip()
self.maxSize = len(strId)
try:
self.currentId = int(strId)
except ValueError:
raise Error, 'Invalid file for FlatfileMapping: %s' % filename
except EnvironmentError, e:
# File couldn't be opened.
self.maxSize = int(math.log10(maxSize))
self.currentId = 0
self._incrementCurrentId()
def _canonicalId(self, id):
if id is not None:
return str(id).zfill(self.maxSize)
else:
return '-'*self.maxSize
def _incrementCurrentId(self, fd=None):
fdWasNone = fd is None
if fdWasNone:
fd = file(self.filename, 'a')
fd.seek(0)
self.currentId += 1
fd.write(self._canonicalId(self.currentId))
fd.write('\n')
if fdWasNone:
fd.close()
def _splitLine(self, line):
line = line.rstrip('\r\n')
(id, s) = line.split(':', 1)
return (id, s)
def _joinLine(self, id, s):
return '%s:%s\n' % (self._canonicalId(id), s)
def add(self, s):
line = self._joinLine(self.currentId, s)
fd = file(self.filename, 'r+')
try:
fd.seek(0, 2) # End.
fd.write(line)
return self.currentId
finally:
self._incrementCurrentId(fd)
fd.close()
def get(self, id):
strId = self._canonicalId(id)
try:
fd = file(self.filename)
fd.readline() # First line, nextId.
for line in fd:
(lineId, s) = self._splitLine(line)
if lineId == strId:
return s
raise NoRecordError, id
finally:
fd.close()
# XXX This assumes it's not been given out. We should make sure that our
# maximum id remains accurate if this is some value we've never given
# out -- i.e., self.maxid = max(self.maxid, id) or something.
def set(self, id, s):
strLine = self._joinLine(id, s)
try:
fd = file(self.filename, 'r+')
self.remove(id, fd)
fd.seek(0, 2) # End.
fd.write(strLine)
finally:
fd.close()
def remove(self, id, fd=None):
fdWasNone = fd is None
strId = self._canonicalId(id)
try:
if fdWasNone:
fd = file(self.filename, 'r+')
fd.seek(0)
fd.readline() # First line, nextId
pos = fd.tell()
line = fd.readline()
while line:
(lineId, _) = self._splitLine(line)
if lineId == strId:
fd.seek(pos)
fd.write(self._canonicalId(None))
fd.seek(pos)
fd.readline() # Same line we just rewrote the id for.
pos = fd.tell()
line = fd.readline()
# We should be at the end.
finally:
if fdWasNone:
fd.close()
def __iter__(self):
fd = file(self.filename)
fd.readline() # First line, nextId.
for line in fd:
(id, s) = self._splitLine(line)
if not id.startswith('-'):
yield (int(id), s)
fd.close()
def vacuum(self):
infd = file(self.filename)
2005-02-01 11:13:23 +01:00
outfd = utils.file.AtomicFile(self.filename,makeBackupIfSmaller=False)
2005-01-19 14:14:38 +01:00
outfd.write(infd.readline()) # First line, nextId.
for line in infd:
if not line.startswith('-'):
outfd.write(line)
infd.close()
outfd.close()
def flush(self):
pass # No-op, we maintain no open files.
def close(self):
self.vacuum() # Should we do this? It should be fine.
class CdbMapping(MappingInterface):
def __init__(self, filename, **kwargs):
self.filename = filename
self._openCdb() # So it can be overridden later.
if 'nextId' not in self.db:
self.db['nextId'] = '1'
def _openCdb(self, *args, **kwargs):
self.db = cdb.open(self.filename, 'c', **kwargs)
2005-01-19 14:14:38 +01:00
def _getNextId(self):
i = int(self.db['nextId'])
self.db['nextId'] = str(i+1)
return i
def get(self, id):
try:
return self.db[str(id)]
except KeyError:
raise NoRecordError, id
# XXX Same as above.
def set(self, id, s):
self.db[str(id)] = s
def add(self, s):
id = self._getNextId()
self.set(id, s)
return id
def remove(self, id):
del self.db[str(id)]
def __iter__(self):
for (id, s) in self.db.iteritems():
if id != 'nextId':
yield (int(id), s)
def flush(self):
self.db.flush()
def close(self):
self.db.close()
class DB(object):
Mapping = 'flat' # This is a good, sane default.
Record = None
def __init__(self, filename, Mapping=None, Record=None):
if Record is not None:
self.Record = Record
if Mapping is not None:
self.Mapping = Mapping
if isinstance(self.Mapping, basestring):
self.Mapping = Mappings[self.Mapping]
self.map = self.Mapping(filename)
def _newRecord(self, id, s):
record = self.Record(id=id)
record.deserialize(s)
return record
def get(self, id):
s = self.map.get(id)
return self._newRecord(id, s)
def set(self, id, record):
s = record.serialize()
self.map.set(id, s)
def add(self, record):
s = record.serialize()
id = self.map.add(s)
record.id = id
return id
def remove(self, id):
self.map.remove(id)
def __iter__(self):
for (id, s) in self.map:
# We don't need to yield the id because it's in the record.
yield self._newRecord(id, s)
def select(self, p):
for record in self:
if p(record):
yield record
def random(self):
try:
2005-01-31 16:22:48 +01:00
return self._newRecord(*utils.iter.choice(self.map))
2005-01-19 14:14:38 +01:00
except IndexError:
return None
def size(self):
return ilen(self.map)
def flush(self):
self.map.flush()
def vacuum(self):
self.map.vacuum()
def close(self):
self.map.close()
Mappings = {
'cdb': CdbMapping,
'flat': FlatfileMapping,
}
class Record(object):
def __init__(self, id=None, **kwargs):
if id is not None:
assert isinstance(id, int), 'id must be an integer.'
self.id = id
self.fields = []
self.defaults = {}
self.converters = {}
for name in self.__fields__:
if isinstance(name, tuple):
(name, spec) = name
else:
spec = utils.safeEval
assert name != 'id'
self.fields.append(name)
if isinstance(spec, tuple):
(converter, default) = spec
else:
converter = spec
default = None
self.defaults[name] = default
self.converters[name] = converter
seen = set()
2005-01-19 14:14:38 +01:00
for (name, value) in kwargs.iteritems():
assert name in self.fields, 'name must be a record value.'
seen.add(name)
setattr(self, name, value)
for name in self.fields:
if name not in seen:
default = self.defaults[name]
if callable(default):
default = default()
setattr(self, name, default)
def serialize(self):
return csv.join([repr(getattr(self, name)) for name in self.fields])
def deserialize(self, s):
unseenRecords = set(self.fields)
2005-01-19 14:14:38 +01:00
for (name, strValue) in zip(self.fields, csv.split(s)):
setattr(self, name, self.converters[name](strValue))
unseenRecords.remove(name)
for name in unseenRecords:
setattr(self, name, self.defaults[name])
# vim:set shiftwidth=4 softtabstop=4 expandtab textwidth=79: