def create_file(self, name):
def onclose_fn(sfile):
sfile.file.close()
f = StructFile(DatastoreFile(key_name="%s%s" % (self.name, name)),
name=name,
onclose=onclose_fn)
return f
def __init__(self,
table_file,
blocksize=16 * 1024,
compressed=0,
prefixcoding=False,
postings=False,
stringids=False,
checksize=True):
self.table_file = table_file
self.blocksize = blocksize
if compressed > 0 and not has_zlib:
raise Exception("zlib is not available: cannot compress table")
self.compressed = compressed
self.prefixcoding = prefixcoding
self.haspostings = postings
if postings:
self.offset = 0
self.postcount = 0
self.lastpostid = None
self.stringids = stringids
self.posting_file = StructFile(tempfile.TemporaryFile())
self.rowbuffer = []
self.lastkey = None
self.blockfilled = 0
self.dir = []
# Remember where we started writing
self.start = table_file.tell()
# Save space for a pointer to the directory
table_file.write_ulong(0)
# Save space for a pointer to the postings
table_file.write_ulong(0)
self.options = {
"haspostings": postings,
"compressed": compressed,
"prefixcoding": prefixcoding,
"stringids": stringids
}
def __init__(self, table_file, blocksize = 16 * 1024,
compressed = 0, prefixcoding = False,
postings = False, stringids = False,
checksize = True):
self.table_file = table_file
self.blocksize = blocksize
if compressed > 0 and not has_zlib:
raise Exception("zlib is not available: cannot compress table")
self.compressed = compressed
self.prefixcoding = prefixcoding
self.haspostings = postings
if postings:
self.offset = 0
self.postcount = 0
self.lastpostid = None
self.stringids = stringids
self.posting_file = StructFile(tempfile.TemporaryFile())
self.rowbuffer = []
self.lastkey = None
self.blockfilled = 0
self.keys = []
self.pointers = array("L")
# Remember where we started writing
self.start = table_file.tell()
# Save space for a pointer to the directory
table_file.write_ulong(0)
# Save space for a pointer to the postings
table_file.write_ulong(0)
self.options = {"haspostings": postings,
"compressed": compressed,
"prefixcoding": prefixcoding,
"stringids": stringids}
def open_file(self, name):
if name not in self.files:
raise NameError
return StructFile(StringIO(self.files[name]))
def create_file(self, name):
def onclose_fn(sfile):
self.files[name] = sfile.file.getvalue()
f = StructFile(StringIO(), name=name, onclose=onclose_fn)
return f
def open_file(self, name, compressed=False):
f = StructFile(open(self._fpath(name), "rb"))
f._name = name
return f
def create_file(self, name):
f = StructFile(open(self._fpath(name), "wb"))
f._name = name
return f
def open_file(self, name, compressed = False):
f = StructFile(open(self._fpath(name), "rb"))
f._name = name
return f
def create_file(self, name):
f = StructFile(open(self._fpath(name), "wb"))
f._name = name
return f
def open_file(self, name):
return StructFile(DatastoreFile.loadfile("%s%s" % (self.name, name)))
class TableWriter(object):
def __init__(self, table_file, blocksize = 16 * 1024,
compressed = 0, prefixcoding = False,
postings = False, stringids = False,
checksize = True):
self.table_file = table_file
self.blocksize = blocksize
if compressed > 0 and not has_zlib:
raise Exception("zlib is not available: cannot compress table")
self.compressed = compressed
self.prefixcoding = prefixcoding
self.haspostings = postings
if postings:
self.offset = 0
self.postcount = 0
self.lastpostid = None
self.stringids = stringids
self.posting_file = StructFile(tempfile.TemporaryFile())
self.rowbuffer = []
self.lastkey = None
self.blockfilled = 0
self.keys = []
self.pointers = array("L")
# Remember where we started writing
self.start = table_file.tell()
# Save space for a pointer to the directory
table_file.write_ulong(0)
# Save space for a pointer to the postings
table_file.write_ulong(0)
self.options = {"haspostings": postings,
"compressed": compressed,
"prefixcoding": prefixcoding,
"stringids": stringids}
def close(self):
# If there is still a block waiting to be written, flush it out
if self.rowbuffer:
self._write_block()
tf = self.table_file
haspostings = self.haspostings
# Remember where we started writing the directory
dirpos = tf.tell()
# Write the directory
tf.write_pickle(self.keys)
tf.write_array(self.pointers)
tf.write_pickle(self.options)
if haspostings:
# Remember where we started the postings
postpos = tf.tell()
# Seek back to the beginning of the postings and
# copy them onto the end of the table file.
self.posting_file.seek(0)
shutil.copyfileobj(self.posting_file, tf)
self.posting_file.close()
# Seek back to where we started writing and write a
# pointer to the directory
tf.seek(self.start)
tf.write_ulong(dirpos)
if haspostings:
# Write a pointer to the postings
tf.write_ulong(postpos)
tf.close()
def _write_block(self):
buf = self.rowbuffer
key = buf[0][0]
compressed = self.compressed
self.keys.append(key)
self.pointers.append(self.table_file.tell())
if compressed:
pck = dumps(buf)
self.table_file.write_string(compress(pck, compressed))
else:
self.table_file.write_pickle(buf)
self.rowbuffer = []
self.blockfilled = 0
def write_posting(self, id, data, writefn):
# IDs must be added in increasing order
if id <= self.lastpostid:
raise IndexError("IDs must increase: %r..%r" % (self.lastpostid, id))
pf = self.posting_file
if self.stringids:
pf.write_string(id.encode("utf8"))
else:
lastpostid = self.lastpostid or 0
pf.write_varint(id - lastpostid)
self.lastpostid = id
self.postcount += 1
return writefn(pf, data)
def add_row(self, key, data):
# Note: call this AFTER you add any postings!
# Keys must be added in increasing order
if key <= self.lastkey:
raise IndexError("Keys must increase: %r..%r" % (self.lastkey, key))
rb = self.rowbuffer
if isinstance(data, array):
self.blockfilled += len(data) * data.itemsize
else:
# Ugh! We're pickling twice! At least it's fast.
self.blockfilled += len(dumps(data))
self.lastkey = key
if self.haspostings:
# Add the posting info to the stored row data
endoffset = self.posting_file.tell()
length = endoffset - self.offset
rb.append((key, (self.offset, length, self.postcount, data)))
# Reset the posting variables
self.offset = endoffset
self.postcount = 0
self.lastpostid = None
else:
rb.append((key, data))
# If this row filled up a block, flush it out
if self.blockfilled >= self.blocksize:
#print len(rb)
self._write_block()
class TableWriter(object):
def __init__(self,
table_file,
blocksize=16 * 1024,
compressed=0,
prefixcoding=False,
postings=False,
stringids=False,
checksize=True):
self.table_file = table_file
self.blocksize = blocksize
if compressed > 0 and not has_zlib:
raise Exception("zlib is not available: cannot compress table")
self.compressed = compressed
self.prefixcoding = prefixcoding
self.haspostings = postings
if postings:
self.offset = 0
self.postcount = 0
self.lastpostid = None
self.stringids = stringids
self.posting_file = StructFile(tempfile.TemporaryFile())
self.rowbuffer = []
self.lastkey = None
self.blockfilled = 0
self.dir = []
# Remember where we started writing
self.start = table_file.tell()
# Save space for a pointer to the directory
table_file.write_ulong(0)
# Save space for a pointer to the postings
table_file.write_ulong(0)
self.options = {
"haspostings": postings,
"compressed": compressed,
"prefixcoding": prefixcoding,
"stringids": stringids
}
def close(self):
# If there is still a block waiting to be written, flush it out
if self.rowbuffer:
self._write_block()
tf = self.table_file
haspostings = self.haspostings
# Remember where we started writing the directory
dirpos = tf.tell()
# Write the directory
tf.write_pickle((tuple(self.dir), self.options))
if haspostings:
# Remember where we started the postings
postpos = tf.tell()
# Seek back to the beginning of the postings and
# copy them onto the end of the table file.
self.posting_file.seek(0)
shutil.copyfileobj(self.posting_file, tf)
self.posting_file.close()
# Seek back to where we started writing and write a
# pointer to the directory
tf.seek(self.start)
tf.write_ulong(dirpos)
if haspostings:
# Write a pointer to the postings
tf.write_ulong(postpos)
tf.close()
def _write_block(self):
buf = self.rowbuffer
key = buf[0][0]
compressed = self.compressed
self.dir.append((key, self.table_file.tell()))
if compressed:
pck = dump_pickle_str(buf, -1)
self.table_file.write_string(compress(pck, compressed))
else:
self.table_file.write_pickle(buf)
self.rowbuffer = []
self.blockfilled = 0
def write_posting(self, id, data, writefn):
# IDs must be added in increasing order
if id <= self.lastpostid:
raise IndexError("IDs must increase: %r..%r" %
(self.lastpostid, id))
pf = self.posting_file
if self.stringids:
pf.write_string(id.encode("utf8"))
else:
lastpostid = self.lastpostid or 0
pf.write_varint(id - lastpostid)
self.lastpostid = id
self.postcount += 1
return writefn(pf, data)
def add_row(self, key, data):
# Note: call this AFTER you add any postings!
# Keys must be added in increasing order
if key <= self.lastkey:
raise IndexError("Keys must increase: %r..%r" %
(self.lastkey, key))
rb = self.rowbuffer
if isinstance(data, array):
self.blockfilled += len(data) * data.itemsize
else:
# Ugh! We're pickling twice! At least it's fast.
self.blockfilled += len(dump_pickle_str(data, -1))
self.lastkey = key
if self.haspostings:
# Add the posting info to the stored row data
endoffset = self.posting_file.tell()
length = endoffset - self.offset
rb.append((key, (self.offset, length, self.postcount, data)))
# Reset the posting variables
self.offset = endoffset
self.postcount = 0
self.lastpostid = None
else:
rb.append((key, data))
# If this row filled up a block, flush it out
if self.blockfilled >= self.blocksize:
self._write_block()
