explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-26 09:56:28 +03:00
Code Issues Packages Projects Releases Wiki Activity

* Roll back to using unicode, and never Py_UNICODE. No dependence on murmurhash either.

Browse Source
This commit is contained in:
Matthew Honnibal 2014-08-18 20:48:48 +02:00
parent bbf9a2c944
commit 1b71cbfe28
2 changed files with 14 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
spacy/spacy.pxd
View File

@ -33,7 +33,7 @@ cdef class Language:
cdef Tokens tokenize(self, unicode text) cdef Tokens tokenize(self, unicode text)
cdef Lexeme* lookup(self, unicode string) except NULL cdef Lexeme* lookup(self, unicode string) except NULL
cdef Lexeme** lookup_chunk(self, Py_UNICODE* chunk, size_t length) except NULL cdef Lexeme** lookup_chunk(self, unicode chunk) except NULL
cdef Lexeme** new_chunk(self, unicode string, list substrings) except NULL cdef Lexeme** new_chunk(self, unicode string, list substrings) except NULL
cdef Lexeme* new_lexeme(self, unicode lex) except NULL cdef Lexeme* new_lexeme(self, unicode lex) except NULL

42
spacy/spacy.pyx
View File

@ -7,20 +7,11 @@ from cython.operator cimport dereference as deref
from spacy.lexeme cimport Lexeme from spacy.lexeme cimport Lexeme
from spacy.lexeme cimport BLANK_WORD from spacy.lexeme cimport BLANK_WORD
from murmurhash cimport mrmr
from spacy.string_tools cimport substr from spacy.string_tools cimport substr
from . import util from . import util
from os import path from os import path
cimport cython
#cdef inline StringHash hash_string(unicode string, size_t length):
# '''Hash unicode with MurmurHash64A'''
# return hash(string)
# #cdef bytes byte_string = string.encode('utf8')
# #return mrmr.hash32(<char*>byte_string, len(byte_string) * sizeof(char), 0)
def get_normalized(unicode lex, size_t length): def get_normalized(unicode lex, size_t length):
@ -74,19 +65,17 @@ cdef class Language:
cdef Lexeme** chunk cdef Lexeme** chunk
cdef Tokens tokens = Tokens(self) cdef Tokens tokens = Tokens(self)
cdef size_t length = len(string) cdef size_t length = len(string)
cdef Py_UNICODE* characters = <Py_UNICODE*>string
cdef Py_UNICODE c
cdef size_t start = 0 cdef size_t start = 0
cdef size_t i cdef size_t i = 0
for i in range(length): for c in string:
c = characters[i]
if _is_whitespace(c): if _is_whitespace(c):
if start < i: if start < i:
chunk = self.lookup_chunk(&characters[start], i - start) chunk = self.lookup_chunk(string[start:i])
_extend(tokens, chunk) _extend(tokens, chunk)
start = i + 1 start = i + 1
i += 1
if start < i: if start < i:
chunk = self.lookup_chunk(&characters[start], length - start) chunk = self.lookup_chunk(string[start:])
_extend(tokens, chunk) _extend(tokens, chunk)
return tokens return tokens
@ -98,13 +87,12 @@ cdef class Language:
word = self.new_lexeme(string) word = self.new_lexeme(string)
return word return word
cdef Lexeme** lookup_chunk(self, Py_UNICODE* c_string, size_t length) except NULL: cdef Lexeme** lookup_chunk(self, unicode string) except NULL:
cdef StringHash h = mrmr.hash32(c_string, length * sizeof(Py_UNICODE), 0) cdef StringHash h = hash(string)
cdef Lexeme** chunk = <Lexeme**>self.chunks[h] cdef Lexeme** chunk = <Lexeme**>self.chunks[h]
cdef int split cdef int split
if chunk == NULL: if chunk == NULL:
chunk = self.new_chunk(c_string[:length], self.find_substrings(c_string[:length])) chunk = self.new_chunk(string, self.find_substrings(string))
self.chunks[h] = <size_t>chunk
return chunk return chunk
cdef Lexeme** new_chunk(self, unicode string, list substrings) except NULL: cdef Lexeme** new_chunk(self, unicode string, list substrings) except NULL:
@ -112,6 +100,7 @@ cdef class Language:
for i, substring in enumerate(substrings): for i, substring in enumerate(substrings):
chunk[i] = self.lookup(substring) chunk[i] = self.lookup(substring)
chunk[i + 1] = NULL chunk[i + 1] = NULL
self.chunks[hash(string)] = <size_t>chunk
return chunk return chunk
cdef Lexeme* new_lexeme(self, unicode string) except NULL: cdef Lexeme* new_lexeme(self, unicode string) except NULL:
@ -172,14 +161,8 @@ cdef class Language:
return len(word) return len(word)
def load_tokenization(self, token_rules=None): def load_tokenization(self, token_rules=None):
cdef StringHash h
cdef Py_UNICODE* c_string
cdef bytes byte_string
for chunk, tokens in token_rules: for chunk, tokens in token_rules:
length = len(chunk) self.new_chunk(chunk, tokens)
c_string = <Py_UNICODE*>chunk
h = mrmr.hash32(c_string, length * sizeof(Py_UNICODE), 0)
self.chunks[h] = <size_t>self.new_chunk(chunk, tokens)
def load_clusters(self): def load_clusters(self):
cdef Lexeme* w cdef Lexeme* w
@ -209,10 +192,11 @@ cdef inline bint _is_whitespace(Py_UNICODE c) nogil:
return False return False
cdef int _extend(Tokens tokens, Lexeme** chunk) except -1: cdef inline int _extend(Tokens tokens, Lexeme** chunk) nogil:
cdef size_t i = 0 cdef size_t i = 0
while chunk[i] != NULL: while chunk[i] != NULL:
tokens.append(<Lexeme_addr>chunk[i]) tokens.vctr[0].push_back(<Lexeme_addr>chunk[i])
tokens.length += 1
i += 1 i += 1