* Upd Tokens to use vector, with bounds checking.

This commit is contained in:
Matthew Honnibal 2014-09-15 03:22:40 +02:00
parent 0f6bf2a2ee
commit f77b7098c0
5 changed files with 125 additions and 146 deletions

View File

@ -18,6 +18,8 @@ from .util import read_lang_data
from spacy.tokens import Tokens
from spacy.lexeme cimport LexemeC, lexeme_init
from murmurhash.mrmr cimport hash64
from cpython.ref cimport Py_INCREF
from spacy._hashing cimport PointerHash
from spacy import orth
@ -194,11 +196,11 @@ cdef class Language:
if lexemes != NULL:
i = 0
while lexemes[i] != NULL:
tokens.push_back(lexemes[i])
tokens.v.push_back(lexemes[i])
i += 1
return 0
cdef uint64_t key = string.key
cdef size_t first_token = tokens.length
cdef size_t first_token = len(tokens)
cdef int split
cdef int remaining = string.n
cdef String prefix
@ -210,14 +212,14 @@ cdef class Language:
if lexemes != NULL:
i = 0
while lexemes[i] != NULL:
tokens.push_back(lexemes[i])
tokens.v.push_back(lexemes[i])
i += 1
else:
tokens.push_back(<LexemeC*>self.lexicon.get(&prefix))
lexemes = <LexemeC**>calloc(tokens.length - first_token, sizeof(LexemeC*))
tokens.v.push_back(<LexemeC*>self.lexicon.get(&prefix))
lexemes = <LexemeC**>calloc(len(tokens) - first_token, sizeof(LexemeC*))
cdef size_t j
for i, j in enumerate(range(first_token, tokens.length)):
lexemes[i] = tokens.lexemes[j]
for i, j in enumerate(range(first_token, tokens.v.size())):
lexemes[i] = tokens.v[j]
self.cache.set(key, lexemes)
cdef int _split_one(self, Py_UNICODE* characters, size_t length):
@ -307,10 +309,8 @@ cdef class Lexicon:
return Lexeme(<size_t>lexeme)
_unicodes = set()
cdef void string_from_unicode(String* s, unicode uni):
global _unicodes
_unicodes.add(uni)
Py_INCREF(uni)
cdef Py_UNICODE* c_uni = <Py_UNICODE*>uni
string_from_slice(s, c_uni, 0, len(uni))

View File

@ -9,7 +9,6 @@ cdef struct LexemeC:
char* string
char** views
flag_t flags

View File

@ -1,4 +1,6 @@
from libc.stdlib cimport calloc, free
from cpython.ref cimport Py_INCREF
cdef LexemeC* lexeme_init(unicode string, double prob, size_t cluster,
list views, set flags):
@ -21,13 +23,11 @@ cdef int lexeme_free(LexemeC* lexeme) except -1:
free(lexeme)
cdef set _strings = set()
cdef char* intern_and_encode(unicode string, size_t* length):
global _strings
cdef bytes decoded = string.encode('utf8')
cdef bytes utf8_string = intern(decoded)
cdef bytes utf8_string = intern(string.encode('utf8'))
Py_INCREF(utf8_string)
length[0] = len(utf8_string)
_strings.add(utf8_string)
return <char*>utf8_string

View File

@ -1,50 +1,46 @@
from spacy.lexeme cimport LexemeC
from libcpp.vector cimport vector
cdef class Tokens:
cdef size_t length
cdef size_t size
cdef vector[LexemeC*] v
cdef LexemeC** lexemes
cdef int push_back(self, LexemeC* lexeme) except 01
cpdef size_t id(self, size_t i)
cpdef size_t id(self, size_t i) except 0
cpdef unicode string(self, size_t i)
cpdef double prob(self, size_t i)
cpdef size_t cluster(self, size_t i)
cpdef bint check_flag(self, size_t i, size_t flag_id)
cpdef double prob(self, size_t i) except 1
cpdef size_t cluster(self, size_t i) except *
cpdef bint check_flag(self, size_t i, size_t flag_id) except *
cpdef unicode string_view(self, size_t i, size_t view_id)
cpdef size_t canon(self, size_t i)
cpdef size_t shape(self, size_t i)
cpdef size_t non_sparse(self, size_t i)
cpdef size_t asciied(self, size_t i)
cpdef size_t canon(self, size_t i) except 0
cpdef size_t shape(self, size_t i) except 0
cpdef size_t non_sparse(self, size_t i) except 0
cpdef size_t asciied(self, size_t i) except 0
cpdef unicode canon_string(self, size_t i)
cpdef unicode shape_string(self, size_t i)
cpdef unicode non_sparse_string(self, size_t i)
cpdef unicode asciied_string(self, size_t i)
cpdef bint is_alpha(self, size_t i)
cpdef bint is_ascii(self, size_t i)
cpdef bint is_digit(self, size_t i)
cpdef bint is_lower(self, size_t i)
cpdef bint is_punct(self, size_t i)
cpdef bint is_space(self, size_t i)
cpdef bint is_title(self, size_t i)
cpdef bint is_upper(self, size_t i)
cpdef bint can_adj(self, size_t i)
cpdef bint can_adp(self, size_t i)
cpdef bint can_adv(self, size_t i)
cpdef bint can_conj(self, size_t i)
cpdef bint can_det(self, size_t i)
cpdef bint can_noun(self, size_t i)
cpdef bint can_num(self, size_t i)
cpdef bint can_pdt(self, size_t i)
cpdef bint can_pos(self, size_t i)
cpdef bint can_pron(self, size_t i)
cpdef bint can_prt(self, size_t i)
cpdef bint can_punct(self, size_t i)
cpdef bint can_verb(self, size_t i)
cpdef bint oft_lower(self, size_t i)
cpdef bint oft_title(self, size_t i)
cpdef bint oft_upper(self, size_t i)
cpdef bint is_alpha(self, size_t i) except *
cpdef bint is_ascii(self, size_t i) except *
cpdef bint is_digit(self, size_t i) except *
cpdef bint is_lower(self, size_t i) except *
cpdef bint is_punct(self, size_t i) except *
cpdef bint is_space(self, size_t i) except *
cpdef bint is_title(self, size_t i) except *
cpdef bint is_upper(self, size_t i) except *
cpdef bint can_adj(self, size_t i) except *
cpdef bint can_adp(self, size_t i) except *
cpdef bint can_adv(self, size_t i) except *
cpdef bint can_conj(self, size_t i) except *
cpdef bint can_det(self, size_t i) except *
cpdef bint can_noun(self, size_t i) except *
cpdef bint can_num(self, size_t i) except *
cpdef bint can_pdt(self, size_t i) except *
cpdef bint can_pos(self, size_t i) except *
cpdef bint can_pron(self, size_t i) except *
cpdef bint can_prt(self, size_t i) except *
cpdef bint can_punct(self, size_t i) except *
cpdef bint can_verb(self, size_t i) except *
cpdef bint oft_lower(self, size_t i) except *
cpdef bint oft_title(self, size_t i) except *
cpdef bint oft_upper(self, size_t i) except *

View File

@ -65,148 +65,132 @@ cdef class Tokens:
"""
def __cinit__(self, string_length=0):
size = int(string_length / 3) if string_length >= 3 else 1
self.lexemes = <LexemeC**>calloc(size, sizeof(LexemeC*))
self.size = size
self.length = 0
def __dealloc__(self):
free(self.lexemes)
self.v.reserve(size)
def __getitem__(self, i):
if i >= self.length:
raise IndexError
return Lexeme(<size_t>self.lexemes[i])
return Lexeme(<size_t>self.v.at(i))
def __len__(self):
return self.length
return self.v.size()
def append(self, Lexeme lexeme):
self.push_back(lexeme._c)
cdef int push_back(self, LexemeC* lexeme) except -1:
if (self.length + 1) == self.size:
self.size *= 2
self.lexemes = <LexemeC**>realloc(self.lexemes, self.size * sizeof(LexemeC*))
self.lexemes[self.length] = lexeme
self.length += 1
self.v.push_back(lexeme._c)
cpdef unicode string(self, size_t i):
cdef bytes utf8_string = self.lexemes[i].string[:self.lexemes[i].length]
cdef bytes utf8_string = self.v[i].string[:self.v[i].length]
cdef unicode string = utf8_string.decode('utf8')
return string
cpdef size_t id(self, size_t i):
return id(self.lexemes[i].string)
cpdef size_t id(self, size_t i) except 0:
return <size_t>&self.v.at(i).string
cpdef double prob(self, size_t i):
return self.lexemes[i].prob
cpdef double prob(self, size_t i) except 1:
return self.v.at(i).prob
cpdef size_t cluster(self, size_t i):
return self.lexemes[i].cluster
cpdef size_t cluster(self, size_t i) except *:
return self.v.at(i).cluster
cpdef bint check_flag(self, size_t i, size_t flag_id):
return lexeme_check_flag(self.lexemes[i], flag_id)
cpdef bint check_flag(self, size_t i, size_t flag_id) except *:
return lexeme_check_flag(self.v[i], flag_id)
cpdef unicode string_view(self, size_t i, size_t view_id):
return lexeme_string_view(self.lexemes[i], view_id)
return lexeme_string_view(self.v.at(i), view_id)
# Provide accessor methods for the features supported by the language.
# Without these, clients have to use the underlying string_view and check_flag
# methods, which requires them to know the IDs.
cpdef unicode canon_string(self, size_t i):
return lexeme_string_view(self.lexemes[i], View_CanonForm)
return lexeme_string_view(self.v.at(i), View_CanonForm)
cpdef unicode shape_string(self, size_t i):
return lexeme_string_view(self.lexemes[i], View_WordShape)
return lexeme_string_view(self.v.at(i), View_WordShape)
cpdef unicode non_sparse_string(self, size_t i):
return lexeme_string_view(self.lexemes[i], View_NonSparse)
return lexeme_string_view(self.v.at(i), View_NonSparse)
cpdef unicode asciied_string(self, size_t i):
return lexeme_string_view(self.lexemes[i], View_Asciied)
return lexeme_string_view(self.v.at(i), View_Asciied)
cpdef size_t canon(self, size_t i):
return id(self.lexemes[i].views[<size_t>View_CanonForm])
cpdef size_t canon(self, size_t i) except *:
return id(self.v.at(i).views[<size_t>View_CanonForm])
cpdef size_t shape(self, size_t i):
return id(self.lexemes[i].views[<size_t>View_WordShape])
cpdef size_t shape(self, size_t i) except *:
return id(self.v.at(i).views[<size_t>View_WordShape])
cpdef size_t non_sparse(self, size_t i):
return id(self.lexemes[i].views[<size_t>View_NonSparse])
cpdef size_t non_sparse(self, size_t i) except *:
return id(self.v.at(i).views[<size_t>View_NonSparse])
cpdef size_t asciied(self, size_t i):
return id(self.lexemes[i].views[<size_t>View_Asciied])
cpdef size_t asciied(self, size_t i) except *:
return id(self.v.at(i).views[<size_t>View_Asciied])
cpdef bint is_alpha(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsAlpha)
cpdef bint is_alpha(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsAlpha)
cpdef bint is_ascii(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsAscii)
cpdef bint is_ascii(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsAscii)
cpdef bint is_digit(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsDigit)
cpdef bint is_digit(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsDigit)
cpdef bint is_lower(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsLower)
cpdef bint is_lower(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsLower)
cpdef bint is_punct(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsPunct)
cpdef bint is_punct(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsPunct)
cpdef bint is_space(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsSpace)
cpdef bint is_space(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsSpace)
cpdef bint is_title(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsTitle)
cpdef bint is_title(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsTitle)
cpdef bint is_upper(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_IsUpper)
cpdef bint is_upper(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_IsUpper)
cpdef bint can_adj(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanAdj)
cpdef bint can_adj(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanAdj)
cpdef bint can_adp(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanAdp)
cpdef bint can_adp(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanAdp)
cpdef bint can_adv(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanAdv)
cpdef bint can_adv(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanAdv)
cpdef bint can_conj(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanConj)
cpdef bint can_conj(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanConj)
cpdef bint can_det(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanDet)
cpdef bint can_det(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanDet)
cpdef bint can_noun(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanNoun)
cpdef bint can_noun(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanNoun)
cpdef bint can_num(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanNum)
cpdef bint can_num(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanNum)
cpdef bint can_pdt(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanPdt)
cpdef bint can_pdt(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanPdt)
cpdef bint can_pos(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanPos)
cpdef bint can_pos(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanPos)
cpdef bint can_pron(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanPron)
cpdef bint can_pron(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanPron)
cpdef bint can_prt(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanPrt)
cpdef bint can_prt(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanPrt)
cpdef bint can_punct(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanPunct)
cpdef bint can_punct(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanPunct)
cpdef bint can_verb(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_CanVerb)
cpdef bint can_verb(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_CanVerb)
cpdef bint oft_lower(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_OftLower)
cpdef bint oft_title(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_OftTitle)
cpdef bint oft_upper(self, size_t i):
return lexeme_check_flag(self.lexemes[i], Flag_OftUpper)
cpdef bint oft_lower(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_OftLower)
cpdef bint oft_title(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_OftTitle)
cpdef bint oft_upper(self, size_t i) except *:
return lexeme_check_flag(self.v.at(i), Flag_OftUpper)