|
|
c085bd |
diff --git a/tools/genfilter.py b/tools/genfilter.py
|
|
|
c085bd |
index 5ffab32..0c5f75a 100644
|
|
|
c085bd |
--- a/tools/genfilter.py
|
|
|
c085bd |
+++ b/tools/genfilter.py
|
|
|
c085bd |
@@ -84,24 +84,24 @@ class FilterGenerator(object):
|
|
|
c085bd |
|
|
|
c085bd |
def generate(self):
|
|
|
c085bd |
size = os.fstat(self.infile.fileno()).st_size
|
|
|
c085bd |
- n = size / self.record_size
|
|
|
c085bd |
+ n = size // self.record_size
|
|
|
c085bd |
m = int(math.ceil(-n*math.log10(ERROR_RATE) /
|
|
|
c085bd |
math.pow(math.log10(2), 2)))
|
|
|
c085bd |
- m = (m/8 + 1)*8
|
|
|
c085bd |
+ m = (m//8 + 1)*8
|
|
|
c085bd |
inmem = mmap.mmap(self.infile.fileno(),
|
|
|
c085bd |
size,
|
|
|
c085bd |
access=mmap.ACCESS_READ)
|
|
|
c085bd |
- outmem = bytearray(m/8)
|
|
|
c085bd |
- for i in xrange(0, n):
|
|
|
c085bd |
+ outmem = bytearray(m//8)
|
|
|
c085bd |
+ for i in range(0, n):
|
|
|
c085bd |
offset = i*self.record_size
|
|
|
c085bd |
b0, b1 = struct.unpack("=LL", inmem[offset:offset+8])
|
|
|
c085bd |
- for k in xrange(0, 4):
|
|
|
c085bd |
+ for k in range(0, 4):
|
|
|
c085bd |
h = murmur_hash3_32(b0, b1, k)
|
|
|
c085bd |
h = int(h * (m / float(0xFFFFFFFF)))
|
|
|
c085bd |
- outmem[h/8] |= (1 << (h%8))
|
|
|
c085bd |
+ outmem[h//8] |= (1 << (h%8))
|
|
|
c085bd |
inmem.close()
|
|
|
c085bd |
- # Convert bytearray to str, for Python 2.6 compatibility.
|
|
|
c085bd |
- self.outfile.write(str(outmem))
|
|
|
c085bd |
+ # Convert bytearray to bytes, for Python 3 compatibility.
|
|
|
c085bd |
+ self.outfile.write(bytes(outmem))
|
|
|
c085bd |
|
|
|
c085bd |
if __name__ == '__main__':
|
|
|
c085bd |
import sys
|
|
|
c085bd |
@@ -110,7 +110,7 @@ if __name__ == '__main__':
|
|
|
c085bd |
parser = argparse.ArgumentParser(description='filter')
|
|
|
c085bd |
parser.add_argument('infile', type=argparse.FileType('r'),
|
|
|
c085bd |
help='input file')
|
|
|
c085bd |
- parser.add_argument('outfile', type=argparse.FileType('w'),
|
|
|
c085bd |
+ parser.add_argument('outfile', type=argparse.FileType('wb'),
|
|
|
c085bd |
help='output file')
|
|
|
c085bd |
parser.add_argument('record_size', type=int,
|
|
|
c085bd |
help='record size')
|
|
|
c085bd |
diff --git a/tools/sortlm.py b/tools/sortlm.py
|
|
|
c085bd |
index a0dd8fe..40f0837 100644
|
|
|
c085bd |
--- a/tools/sortlm.py
|
|
|
c085bd |
+++ b/tools/sortlm.py
|
|
|
c085bd |
@@ -40,10 +40,10 @@ class SortedGenerator(object):
|
|
|
c085bd |
self.__min_cost = 0.0
|
|
|
c085bd |
|
|
|
c085bd |
def read(self):
|
|
|
c085bd |
- print "reading N-grams"
|
|
|
c085bd |
+ print("reading N-grams")
|
|
|
c085bd |
self.__read_tries()
|
|
|
c085bd |
self.__read_ngrams()
|
|
|
c085bd |
- print "min cost = %lf" % self.__min_cost
|
|
|
c085bd |
+ print("min cost = %lf" % self.__min_cost)
|
|
|
c085bd |
|
|
|
c085bd |
def __read_tries(self):
|
|
|
c085bd |
while True:
|
|
|
c085bd |
@@ -58,7 +58,7 @@ class SortedGenerator(object):
|
|
|
c085bd |
line = self.__infile.readline()
|
|
|
c085bd |
if line == "":
|
|
|
c085bd |
break
|
|
|
c085bd |
- line = line.strip()
|
|
|
c085bd |
+ line = line.strip('\n')
|
|
|
c085bd |
if line == "":
|
|
|
c085bd |
break
|
|
|
c085bd |
match = self.__ngram_line_regex.match(line)
|
|
|
c085bd |
@@ -89,7 +89,7 @@ class SortedGenerator(object):
|
|
|
c085bd |
line = self.__infile.readline()
|
|
|
c085bd |
if line == "":
|
|
|
c085bd |
break
|
|
|
c085bd |
- line = line.strip()
|
|
|
c085bd |
+ line = line.strip('\n')
|
|
|
c085bd |
if line == "":
|
|
|
c085bd |
break
|
|
|
c085bd |
match = self.__ngram_line_regex.match(line)
|
|
|
c085bd |
@@ -125,14 +125,11 @@ class SortedGenerator(object):
|
|
|
c085bd |
def quantize(cost, min_cost):
|
|
|
c085bd |
return max(0, min(65535, int(cost * 65535 / min_cost)))
|
|
|
c085bd |
|
|
|
c085bd |
- def cmp_header(a, b):
|
|
|
c085bd |
- return cmp(a[0], b[0])
|
|
|
c085bd |
-
|
|
|
c085bd |
- print "writing 1-gram file"
|
|
|
c085bd |
+ print("writing 1-gram file")
|
|
|
c085bd |
unigram_offsets = {}
|
|
|
c085bd |
unigram_file = open("%s.1gram" % self.__output_prefix, "wb")
|
|
|
c085bd |
offset = 0
|
|
|
c085bd |
- for ids, value in sorted(self.__ngram_entries[0].iteritems()):
|
|
|
c085bd |
+ for ids, value in sorted(self.__ngram_entries[0].items()):
|
|
|
c085bd |
unigram_offsets[ids[0]] = offset
|
|
|
c085bd |
s = struct.pack("=HHH",
|
|
|
c085bd |
quantize(value[0], self.__min_cost),
|
|
|
c085bd |
@@ -143,13 +140,13 @@ class SortedGenerator(object):
|
|
|
c085bd |
offset += 1
|
|
|
c085bd |
unigram_file.close()
|
|
|
c085bd |
|
|
|
c085bd |
- print "writing 2-gram file"
|
|
|
c085bd |
+ print("writing 2-gram file")
|
|
|
c085bd |
bigram_offsets = {}
|
|
|
c085bd |
bigram_file = open("%s.2gram" % self.__output_prefix, "wb")
|
|
|
c085bd |
keys = self.__ngram_entries[1].keys()
|
|
|
c085bd |
items = [(struct.pack("=LL", ids[1], unigram_offsets[ids[0]]), ids) for ids in keys]
|
|
|
c085bd |
offset = 0
|
|
|
c085bd |
- for header, ids in sorted(items, cmp=cmp_header):
|
|
|
c085bd |
+ for header, ids in sorted(items, key=lambda x: x[0]):
|
|
|
c085bd |
value = self.__ngram_entries[1][ids]
|
|
|
c085bd |
bigram_offsets[ids] = offset
|
|
|
c085bd |
s = struct.pack("=HH",
|
|
|
c085bd |
@@ -160,11 +157,11 @@ class SortedGenerator(object):
|
|
|
c085bd |
bigram_file.close()
|
|
|
c085bd |
|
|
|
c085bd |
if len(self.__ngram_entries[2]) > 0:
|
|
|
c085bd |
- print "writing 3-gram file"
|
|
|
c085bd |
+ print("writing 3-gram file")
|
|
|
c085bd |
trigram_file = open("%s.3gram" % self.__output_prefix, "wb")
|
|
|
c085bd |
keys = self.__ngram_entries[2].keys()
|
|
|
c085bd |
items = [(struct.pack("=LL", ids[2], bigram_offsets[(ids[0], ids[1])]), ids) for ids in keys]
|
|
|
c085bd |
- for header, ids in sorted(items, cmp=cmp_header):
|
|
|
c085bd |
+ for header, ids in sorted(items, key=lambda x: x[0]):
|
|
|
c085bd |
value = self.__ngram_entries[2][ids]
|
|
|
c085bd |
s = struct.pack("=H",
|
|
|
c085bd |
quantize(value[0], self.__min_cost))
|