From a17266aa4415c8fab6f6ce02264e71ab4f74eccf Mon Sep 17 00:00:00 2001
From: synodriver <diguohuangjiajinweijun@gmail.com>
Date: Thu, 11 Aug 2022 23:06:28 +0800
Subject: [PATCH 01/18] use nogil, prange and typedview to speedup

---
 pyopenjtalk/htsengine.pyx              |  86 ++++++++-----
 pyopenjtalk/htsengine/__init__.pxd     |   2 +-
 pyopenjtalk/openjtalk.pyx              | 162 ++++++++++++++-----------
 pyopenjtalk/openjtalk/jpcommon.pxd     |  24 ++--
 pyopenjtalk/openjtalk/mecab.pxd        |   2 +-
 pyopenjtalk/openjtalk/mecab2njd.pxd    |   2 +-
 pyopenjtalk/openjtalk/njd.pxd          |  14 +--
 pyopenjtalk/openjtalk/njd2jpcommon.pxd |   2 +-
 pyopenjtalk/openjtalk/text2mecab.pxd   |   2 +-
 9 files changed, 167 insertions(+), 129 deletions(-)

diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index a69400c..5a2aa9e 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -8,7 +8,9 @@ cimport numpy as np
 np.import_array()
 
 cimport cython
-from libc.stdlib cimport malloc, free
+from cython.parallel cimport prange
+from cpython.mem cimport PyMem_Malloc, PyMem_Free
+from libc.stdint cimport uint8_t
 
 from htsengine cimport HTS_Engine
 from htsengine cimport (
@@ -19,7 +21,10 @@ from htsengine cimport (
     HTS_Engine_get_generated_speech, HTS_Engine_get_nsamples
 )
 
-cdef class HTSEngine(object):
+@cython.final
+@cython.no_gc
+@cython.freelist(4)
+cdef class HTSEngine:
     """HTSEngine
 
     Args:
@@ -36,38 +41,46 @@ cdef class HTSEngine(object):
           self.clear()
           raise RuntimeError("Failed to initalize HTS_Engine")
 
-    def load(self, bytes voice):
-        cdef char* voices = voice
+    cpdef inline char load(self, const uint8_t[::1] voice):
         cdef char ret
-        ret = HTS_Engine_load(self.engine, &voices, 1)
+        with nogil:
+            ret = HTS_Engine_load(self.engine, &(<char*>&voice[0]), 1)
         return ret
 
-    def get_sampling_frequency(self):
+    cpdef inline size_t get_sampling_frequency(self):
         """Get sampling frequency
         """
-        return HTS_Engine_get_sampling_frequency(self.engine)
+        cdef size_t ret
+        with nogil:
+            ret = HTS_Engine_get_sampling_frequency(self.engine)
+        return ret
 
-    def get_fperiod(self):
+    cpdef inline size_t get_fperiod(self):
         """Get frame period"""
-        return HTS_Engine_get_fperiod(self.engine)
+        cdef size_t ret
+        with nogil:
+            ret = HTS_Engine_get_fperiod(self.engine)
+        return ret
 
-    def set_speed(self, speed=1.0):
+    cpdef inline void set_speed(self, double speed=1.0):
         """Set speed
 
         Args:
             speed (float): speed
         """
-        HTS_Engine_set_speed(self.engine, speed)
+        with nogil:
+            HTS_Engine_set_speed(self.engine, speed)
 
-    def add_half_tone(self, half_tone=0.0):
+    cpdef inline void add_half_tone(self, double half_tone=0.0):
         """Additional half tone in log-f0
 
         Args:
             half_tone (float): additional half tone
         """
-        HTS_Engine_add_half_tone(self.engine, half_tone)
+        with nogil:
+            HTS_Engine_add_half_tone(self.engine, half_tone)
 
-    def synthesize(self, list labels):
+    cpdef inline np.ndarray[np.float64_t, ndim=1] synthesize(self, list labels):
         """Synthesize waveform from list of full-context labels
 
         Args:
@@ -77,40 +90,49 @@ cdef class HTSEngine(object):
             np.ndarray: speech waveform
         """
         self.synthesize_from_strings(labels)
-        x = self.get_generated_speech()
+        cdef np.ndarray[np.float64_t, ndim=1] x = self.get_generated_speech()
         self.refresh()
         return x
 
-    def synthesize_from_strings(self, list labels):
+    cpdef inline char synthesize_from_strings(self, list labels) except? 0:
         """Synthesize from strings"""
         cdef size_t num_lines = len(labels)
-        cdef char **lines = <char**> malloc((num_lines + 1) * sizeof(char*))
+        cdef char **lines = <char**> PyMem_Malloc((num_lines + 1) * sizeof(char*))
+        cdef int n
         for n in range(len(labels)):
             lines[n] = <char*>labels[n]
-
-        cdef char ret = HTS_Engine_synthesize_from_strings(self.engine, lines, num_lines)
-        free(lines)
+        cdef char ret
+        with nogil:
+            ret = HTS_Engine_synthesize_from_strings(self.engine, lines, num_lines)
+        PyMem_Free(lines) # todo: use finally
         if ret != 1:
             raise RuntimeError("Failed to run synthesize_from_strings")
+        return ret
 
-    def get_generated_speech(self):
+    cpdef inline np.ndarray[np.float64_t, ndim=1] get_generated_speech(self):
         """Get generated speech"""
         cdef size_t nsamples = HTS_Engine_get_nsamples(self.engine)
-        cdef np.ndarray speech = np.zeros([nsamples], dtype=np.float64)
+        cdef np.ndarray[np.float64_t, ndim=1] speech = np.zeros([nsamples], dtype=np.float64)
+        cdef double[::1] speech_view = speech
         cdef size_t index
-        for index in range(nsamples):
-            speech[index] = HTS_Engine_get_generated_speech(self.engine, index)
+        for index in prange(nsamples, nogil=True):
+            speech_view[index] = HTS_Engine_get_generated_speech(self.engine, index)
         return speech
 
-    def get_fullcontext_label_format(self):
+    cpdef inline str get_fullcontext_label_format(self):
         """Get full-context label format"""
-        return (<bytes>HTS_Engine_get_fullcontext_label_format(self.engine)).decode("utf-8")
-
-    def refresh(self):
-        HTS_Engine_refresh(self.engine)
-
-    def clear(self):
-        HTS_Engine_clear(self.engine)
+        cdef const char* f
+        with nogil:
+            f = HTS_Engine_get_fullcontext_label_format(self.engine)
+        return (<bytes>f).decode("utf-8")
+
+    cpdef inline void refresh(self):
+        with nogil:
+            HTS_Engine_refresh(self.engine)
+
+    cpdef inline void clear(self):
+        with nogil:
+            HTS_Engine_clear(self.engine)
 
     def __dealloc__(self):
         self.clear()
diff --git a/pyopenjtalk/htsengine/__init__.pxd b/pyopenjtalk/htsengine/__init__.pxd
index c24d959..fc8f70c 100644
--- a/pyopenjtalk/htsengine/__init__.pxd
+++ b/pyopenjtalk/htsengine/__init__.pxd
@@ -1,7 +1,7 @@
 # distutils: language = c++
 
 
-cdef extern from "HTS_engine.h":
+cdef extern from "HTS_engine.h" nogil:
     cdef cppclass _HTS_Engine:
         pass
     ctypedef _HTS_Engine HTS_Engine
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 46e6004..4dd3124 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -2,12 +2,14 @@
 # cython: boundscheck=True, wraparound=True
 # cython: c_string_type=unicode, c_string_encoding=ascii
 
+from libc.stdint cimport uint8_t
 import numpy as np
 
 cimport numpy as np
 np.import_array()
 
 cimport cython
+from cpython.bytes cimport PyBytes_AS_STRING
 
 from openjtalk.mecab cimport Mecab, Mecab_initialize, Mecab_load, Mecab_analysis
 from openjtalk.mecab cimport Mecab_get_feature, Mecab_get_size, Mecab_refresh, Mecab_clear
@@ -21,77 +23,80 @@ from openjtalk.text2mecab cimport text2mecab
 from openjtalk.mecab2njd cimport mecab2njd
 from openjtalk.njd2jpcommon cimport njd2jpcommon
 
-cdef njd_node_get_string(_njd.NJDNode* node):
+cdef inline str njd_node_get_string(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_string(node))).decode("utf-8")
 
-cdef njd_node_get_pos(_njd.NJDNode* node):
+cdef inline str njd_node_get_pos(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pos(node))).decode("utf-8")
 
-cdef njd_node_get_pos_group1(_njd.NJDNode* node):
+cdef inline str njd_node_get_pos_group1(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pos_group1(node))).decode("utf-8")
 
-cdef njd_node_get_pos_group2(_njd.NJDNode* node):
+cdef inline str njd_node_get_pos_group2(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pos_group2(node))).decode("utf-8")
 
-cdef njd_node_get_pos_group3(_njd.NJDNode* node):
+cdef inline str njd_node_get_pos_group3(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pos_group3(node))).decode("utf-8")
 
-cdef njd_node_get_ctype(_njd.NJDNode* node):
+cdef inline str njd_node_get_ctype(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_ctype(node))).decode("utf-8")
 
-cdef njd_node_get_cform(_njd.NJDNode* node):
+cdef inline  str njd_node_get_cform(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_cform(node))).decode("utf-8")
 
-cdef njd_node_get_orig(_njd.NJDNode* node):
+cdef inline str njd_node_get_orig(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_orig(node))).decode("utf-8")
 
-cdef njd_node_get_read(_njd.NJDNode* node):
+cdef inline str njd_node_get_read(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_read(node))).decode("utf-8")
 
-cdef njd_node_get_pron(_njd.NJDNode* node):
+cdef inline str njd_node_get_pron(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pron(node))).decode("utf-8")
 
-cdef njd_node_get_acc(_njd.NJDNode* node):
+cdef inline str njd_node_get_acc(_njd.NJDNode* node):
     return _njd.NJDNode_get_acc(node)
 
-cdef njd_node_get_mora_size(_njd.NJDNode* node):
+cdef inline str njd_node_get_mora_size(_njd.NJDNode* node):
     return _njd.NJDNode_get_mora_size(node)
 
-cdef njd_node_get_chain_rule(_njd.NJDNode* node):
+cdef inline str njd_node_get_chain_rule(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_chain_rule(node))).decode("utf-8")
 
-cdef njd_node_get_chain_flag(_njd.NJDNode* node):
-      return _njd.NJDNode_get_chain_flag(node)
-
-
-cdef njd_node_print(_njd.NJDNode* node):
-  return "{},{},{},{},{},{},{},{},{},{},{}/{},{},{}".format(
-    njd_node_get_string(node),
-    njd_node_get_pos(node),
-    njd_node_get_pos_group1(node),
-    njd_node_get_pos_group2(node),
-    njd_node_get_pos_group3(node),
-    njd_node_get_ctype(node),
-    njd_node_get_cform(node),
-    njd_node_get_orig(node),
-    njd_node_get_read(node),
-    njd_node_get_pron(node),
-    njd_node_get_acc(node),
-    njd_node_get_mora_size(node),
-    njd_node_get_chain_rule(node),
-    njd_node_get_chain_flag(node)
+cdef inline str njd_node_get_chain_flag(_njd.NJDNode* node):
+    return _njd.NJDNode_get_chain_flag(node)
+
+
+cdef str njd_node_print(_njd.NJDNode* node):
+    return "{},{},{},{},{},{},{},{},{},{},{}/{},{},{}".format(
+        njd_node_get_string(node),
+        njd_node_get_pos(node),
+        njd_node_get_pos_group1(node),
+        njd_node_get_pos_group2(node),
+        njd_node_get_pos_group3(node),
+        njd_node_get_ctype(node),
+        njd_node_get_cform(node),
+        njd_node_get_orig(node),
+        njd_node_get_read(node),
+        njd_node_get_pron(node),
+        njd_node_get_acc(node),
+        njd_node_get_mora_size(node),
+        njd_node_get_chain_rule(node),
+        njd_node_get_chain_flag(node)
     )
 
 
-cdef njd_print(_njd.NJD* njd):
+cdef list njd_print(_njd.NJD* njd):
     cdef _njd.NJDNode* node = njd.head
     njd_results = []
     while node is not NULL:
-      njd_results.append(njd_node_print(node))
-      node = node.next
+        njd_results.append(njd_node_print(node))
+        node = node.next
     return njd_results
 
-cdef class OpenJTalk(object):
+@cython.no_gc
+@cython.final
+@cython.freelist(4)
+cdef class OpenJTalk:
     """OpenJTalk
 
     Args:
@@ -112,62 +117,73 @@ cdef class OpenJTalk(object):
 
         r = self._load(dn_mecab)
         if r != 1:
-          self._clear()
-          raise RuntimeError("Failed to initalize Mecab")
+            self._clear()
+            raise RuntimeError("Failed to initalize Mecab")
 
+    cpdef inline void _clear(self):
+        with nogil:
+            Mecab_clear(self.mecab)
+            NJD_clear(self.njd)
+            JPCommon_clear(self.jpcommon)
 
-    def _clear(self):
-      Mecab_clear(self.mecab)
-      NJD_clear(self.njd)
-      JPCommon_clear(self.jpcommon)
+    cpdef inline int _load(self, const uint8_t[::1] dn_mecab):
+        cdef int ret
+        with nogil:
+            ret = Mecab_load(self.mecab, <const char*>&dn_mecab[0])
+        return ret
 
-    def _load(self, bytes dn_mecab):
-        return Mecab_load(self.mecab, dn_mecab)
 
-
-    def run_frontend(self, text, verbose=0):
+    cpdef inline tuple run_frontend(self, object text, int verbose=0):
         """Run OpenJTalk's text processing frontend
         """
         if isinstance(text, str):
-          text = text.encode("utf-8")
-        cdef char buff[8192]
-        text2mecab(buff, text)
-        Mecab_analysis(self.mecab, buff)
-        mecab2njd(self.njd, Mecab_get_feature(self.mecab), Mecab_get_size(self.mecab))
-        _njd.njd_set_pronunciation(self.njd)
-        _njd.njd_set_digit(self.njd)
-        _njd.njd_set_accent_phrase(self.njd)
-        _njd.njd_set_accent_type(self.njd)
-        _njd.njd_set_unvoiced_vowel(self.njd)
-        _njd.njd_set_long_vowel(self.njd)
-        njd2jpcommon(self.jpcommon, self.njd)
-        JPCommon_make_label(self.jpcommon)
-
-        cdef int label_size = JPCommon_get_label_size(self.jpcommon)
-        cdef char** label_feature
-        label_feature = JPCommon_get_label_feature(self.jpcommon)
+            text = text.encode("utf-8")
+        cdef:
+            char buff[8192]
+            const char* text_ptr
+            int label_size
+            char** label_feature
+        text_ptr = PyBytes_AS_STRING(text)
+        with nogil:
+            text2mecab(buff, text_ptr)
+            Mecab_analysis(self.mecab, buff)
+            mecab2njd(self.njd, Mecab_get_feature(self.mecab), Mecab_get_size(self.mecab))
+            _njd.njd_set_pronunciation(self.njd)
+            _njd.njd_set_digit(self.njd)
+            _njd.njd_set_accent_phrase(self.njd)
+            _njd.njd_set_accent_type(self.njd)
+            _njd.njd_set_unvoiced_vowel(self.njd)
+            _njd.njd_set_long_vowel(self.njd)
+            njd2jpcommon(self.jpcommon, self.njd)
+            JPCommon_make_label(self.jpcommon)
+
+            label_size = JPCommon_get_label_size(self.jpcommon)
+            JPCommon_get_label_feature(self.jpcommon)
 
         labels = []
+        cdef int i
         for i in range(label_size):
-          # This will create a copy of c string
-          # http://cython.readthedocs.io/en/latest/src/tutorial/strings.html
-          labels.append(<unicode>label_feature[i])
+            # This will create a copy of c string
+            # http://cython.readthedocs.io/en/latest/src/tutorial/strings.html
+            labels.append(<unicode>label_feature[i])
 
-        njd_results = njd_print(self.njd)
+        cdef list njd_results = njd_print(self.njd)
 
         if verbose > 0:
-          NJD_print(self.njd)
+            NJD_print(self.njd)
 
         # Note that this will release memory for label feature
-        JPCommon_refresh(self.jpcommon)
-        NJD_refresh(self.njd)
-        Mecab_refresh(self.mecab)
+        with nogil:
+            JPCommon_refresh(self.jpcommon)
+            NJD_refresh(self.njd)
+            Mecab_refresh(self.mecab)
 
         return njd_results, labels
 
-    def g2p(self, text, kana=False, join=True):
+    cpdef inline g2p(self, object text, bint kana=False, bint join=True):
         """Grapheme-to-phoeneme (G2P) conversion
         """
+        cdef list njd_results, labels
         njd_results, labels = self.run_frontend(text)
         if not kana:
             prons = list(map(lambda s: s.split("-")[1].split("+")[0], labels[1:-1]))
diff --git a/pyopenjtalk/openjtalk/jpcommon.pxd b/pyopenjtalk/openjtalk/jpcommon.pxd
index 8e86bea..b120fc1 100644
--- a/pyopenjtalk/openjtalk/jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/jpcommon.pxd
@@ -2,21 +2,21 @@
 
 from libc.stdio cimport FILE
 
-cdef extern from "jpcommon.h":
+cdef extern from "jpcommon.h" nogil:
     cdef cppclass JPCommonNode:
-       char *pron
-       char *pos
-       char *ctype
-       char *cform
-       int acc
-       int chain_flag
-       void *prev
-       void *next
+        char *pron
+        char *pos
+        char *ctype
+        char *cform
+        int acc
+        int chain_flag
+        void *prev
+        void *next
 
     cdef cppclass JPCommon:
-       JPCommonNode *head
-       JPCommonNode *tail
-       void *label
+        JPCommonNode *head
+        JPCommonNode *tail
+        void *label
 
     void JPCommon_initialize(JPCommon * jpcommon)
     void JPCommon_push(JPCommon * jpcommon, JPCommonNode * node)
diff --git a/pyopenjtalk/openjtalk/mecab.pxd b/pyopenjtalk/openjtalk/mecab.pxd
index bd367c7..2449db0 100644
--- a/pyopenjtalk/openjtalk/mecab.pxd
+++ b/pyopenjtalk/openjtalk/mecab.pxd
@@ -1,6 +1,6 @@
 # distutils: language = c++
 
-cdef extern from "mecab.h":
+cdef extern from "mecab.h" nogil:
     cdef cppclass Mecab:
         char **feature
         int size
diff --git a/pyopenjtalk/openjtalk/mecab2njd.pxd b/pyopenjtalk/openjtalk/mecab2njd.pxd
index be57ccc..a86d790 100644
--- a/pyopenjtalk/openjtalk/mecab2njd.pxd
+++ b/pyopenjtalk/openjtalk/mecab2njd.pxd
@@ -2,5 +2,5 @@
 
 from .njd cimport NJD
 
-cdef extern from "mecab2njd.h":
+cdef extern from "mecab2njd.h" nogil:
     void mecab2njd(NJD * njd, char **feature, int size);
diff --git a/pyopenjtalk/openjtalk/njd.pxd b/pyopenjtalk/openjtalk/njd.pxd
index 38d3887..2831324 100644
--- a/pyopenjtalk/openjtalk/njd.pxd
+++ b/pyopenjtalk/openjtalk/njd.pxd
@@ -2,7 +2,7 @@
 
 from libc.stdio cimport FILE
 
-cdef extern from "njd.h":
+cdef extern from "njd.h" nogil:
     cdef cppclass NJDNode:
         char *string
         char *pos
@@ -78,20 +78,20 @@ cdef extern from "njd.h":
     void NJD_refresh(NJD * njd)
     void NJD_clear(NJD * wl)
 
-cdef extern from "njd_set_accent_phrase.h":
+cdef extern from "njd_set_accent_phrase.h" nogil:
     void njd_set_accent_phrase(NJD * njd)
 
-cdef extern from "njd_set_accent_type.h":
+cdef extern from "njd_set_accent_type.h" nogil:
     void njd_set_accent_type(NJD * njd)
 
-cdef extern from "njd_set_digit.h":
+cdef extern from "njd_set_digit.h" nogil:
     void njd_set_digit(NJD * njd)
 
-cdef extern from "njd_set_long_vowel.h":
+cdef extern from "njd_set_long_vowel.h" nogil:
     void njd_set_long_vowel(NJD * njd)
 
-cdef extern from "njd_set_pronunciation.h":
+cdef extern from "njd_set_pronunciation.h" nogil:
     void njd_set_pronunciation(NJD * njd)
 
-cdef extern from "njd_set_unvoiced_vowel.h":
+cdef extern from "njd_set_unvoiced_vowel.h" nogil:
     void njd_set_unvoiced_vowel(NJD * njd)
diff --git a/pyopenjtalk/openjtalk/njd2jpcommon.pxd b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
index 8309288..4d6fb62 100644
--- a/pyopenjtalk/openjtalk/njd2jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
@@ -3,5 +3,5 @@
 from .jpcommon cimport JPCommon
 from .njd cimport NJD
 
-cdef extern from "njd2jpcommon.h":
+cdef extern from "njd2jpcommon.h" nogil:
     void njd2jpcommon(JPCommon * jpcommon, NJD * njd)
diff --git a/pyopenjtalk/openjtalk/text2mecab.pxd b/pyopenjtalk/openjtalk/text2mecab.pxd
index 6081757..718f7a5 100644
--- a/pyopenjtalk/openjtalk/text2mecab.pxd
+++ b/pyopenjtalk/openjtalk/text2mecab.pxd
@@ -1,4 +1,4 @@
 # distutils: language = c++
 
-cdef extern from "text2mecab.h":
+cdef extern from "text2mecab.h" nogil:
     void text2mecab(char *output, const char *input)

From 8eed18ceb04b0ef8b6dd8c8d3d271f19fe0dc273 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 03:35:59 +0000
Subject: [PATCH 02/18] speed up and fix omp

---
 pyopenjtalk/htsengine.pyx | 10 ++++++----
 pyopenjtalk/openjtalk.pyx |  8 ++++----
 setup.py                  |  8 ++++----
 3 files changed, 14 insertions(+), 12 deletions(-)

diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index 5a2aa9e..d2117f1 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -1,6 +1,6 @@
 # coding: utf-8
-# cython: boundscheck=True, wraparound=True
-# cython: c_string_type=unicode, c_string_encoding=ascii
+# cython: boundscheck=False, wraparound=True
+# cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 import numpy as np
 
@@ -42,9 +42,11 @@ cdef class HTSEngine:
           raise RuntimeError("Failed to initalize HTS_Engine")
 
     cpdef inline char load(self, const uint8_t[::1] voice):
-        cdef char ret
+        cdef:
+            char ret
+            const uint8_t *voice_ptr = &voice[0]
         with nogil:
-            ret = HTS_Engine_load(self.engine, &(<char*>&voice[0]), 1)
+            ret = HTS_Engine_load(self.engine, <char**>(&voice_ptr), 1)
         return ret
 
     cpdef inline size_t get_sampling_frequency(self):
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 4dd3124..9bb198f 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -1,6 +1,6 @@
 # coding: utf-8
-# cython: boundscheck=True, wraparound=True
-# cython: c_string_type=unicode, c_string_encoding=ascii
+# cython: boundscheck=False, wraparound=True
+# cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 from libc.stdint cimport uint8_t
 import numpy as np
@@ -66,7 +66,7 @@ cdef inline str njd_node_get_chain_flag(_njd.NJDNode* node):
     return _njd.NJDNode_get_chain_flag(node)
 
 
-cdef str njd_node_print(_njd.NJDNode* node):
+cdef inline str njd_node_print(_njd.NJDNode* node):
     return "{},{},{},{},{},{},{},{},{},{},{}/{},{},{}".format(
         njd_node_get_string(node),
         njd_node_get_pos(node),
@@ -180,7 +180,7 @@ cdef class OpenJTalk:
 
         return njd_results, labels
 
-    cpdef inline g2p(self, object text, bint kana=False, bint join=True):
+    def g2p(self, object text, bint kana=False, bint join=True):
         """Grapheme-to-phoeneme (G2P) conversion
         """
         cdef list njd_results, labels
diff --git a/setup.py b/setup.py
index c68db9c..971a08b 100644
--- a/setup.py
+++ b/setup.py
@@ -180,8 +180,8 @@ def escape_macros(macros):
         name="pyopenjtalk.openjtalk",
         sources=[join("pyopenjtalk", "openjtalk" + ext)] + all_src,
         include_dirs=[np.get_include()] + include_dirs,
-        extra_compile_args=[],
-        extra_link_args=[],
+        extra_compile_args=['-fopenmp'],
+        extra_link_args=['-fopenmp'],
         language="c++",
         define_macros=custom_define_macros(
             [
@@ -204,8 +204,8 @@ def escape_macros(macros):
         name="pyopenjtalk.htsengine",
         sources=[join("pyopenjtalk", "htsengine" + ext)] + all_htsengine_src,
         include_dirs=[np.get_include(), join(htsengine_src_top, "include")],
-        extra_compile_args=[],
-        extra_link_args=[],
+        extra_compile_args=['-fopenmp'],
+        extra_link_args=['-fopenmp'],
         libraries=["winmm"] if platform_is_windows else [],
         language="c++",
         define_macros=custom_define_macros(

From 5af6b9d859800f5efe55c468b1967a533bc64d7c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=BA=90=E6=96=87=E9=9B=A8?=
 <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 11:41:04 +0800
Subject: [PATCH 03/18] feat: add dict in setup time

---
 setup.py | 33 ++++++++++++++++++++++++++++++++-
 1 file changed, 32 insertions(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index c68db9c..b7f8feb 100644
--- a/setup.py
+++ b/setup.py
@@ -33,6 +33,10 @@
     "/execution-charset:utf-8",
 ]
 
+_dict_folder_name = "open_jtalk_dic_utf_8-1.11"
+_dict_download_url = "https://github.com/r9y9/open_jtalk/releases/download/v1.11.1"
+_DICT_URL = f"{_dict_download_url}/{_dict_folder_name}.tar.gz"
+
 try:
     if not _CYTHON_INSTALLED:
         raise ImportError("No supported version of Cython installed.")
@@ -138,6 +142,33 @@ def escape_macros(macros):
 # open_jtalk sources
 src_top = join("lib", "open_jtalk", "src")
 
+
+# https://github.com/tqdm/tqdm#hooks-and-callbacks
+class _TqdmUpTo(tqdm):  # type: ignore
+    def update_to(self, b=1, bsize=1, tsize=None):
+        if tsize is not None:
+            self.total = tsize
+        return self.update(b * bsize - self.n)
+
+
+# extract dic
+filename = "dic.tar.gz"
+print('Downloading: "{}"'.format(_DICT_URL))
+with _TqdmUpTo(
+    unit="B",
+    unit_scale=True,
+    unit_divisor=1024,
+    miniters=1,
+    desc="dic.tar.gz",
+) as t:  # all optional kwargs
+    urlretrieve(_DICT_URL, filename, reporthook=t.update_to)
+    t.total = t.n
+print("Extracting tar file {}".format(filename))
+with tarfile.open(filename, mode="r|gz") as f:
+    f.extractall(path="./")
+os.remove(filename)
+
+
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h
 # we could do it on python side but it would be very tricky,
@@ -272,7 +303,7 @@ def run(self):
     url="https://github.com/r9y9/pyopenjtalk",
     license="MIT",
     packages=find_packages(),
-    package_data={"": ["htsvoice/*"]},
+    package_data={"": ["htsvoice/*", f"{_dict_folder_name}/*"]},
     ext_modules=ext_modules,
     cmdclass=cmdclass,
     install_requires=[

From 23641047b3321fc0dac2f8e21446291391555f1d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=BA=90=E6=96=87=E9=9B=A8?=
 <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 11:45:38 +0800
Subject: [PATCH 04/18] add imports

---
 setup.py | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/setup.py b/setup.py
index b7f8feb..ba0e767 100644
--- a/setup.py
+++ b/setup.py
@@ -14,6 +14,15 @@
 import setuptools.command.develop
 from setuptools import Extension, find_packages, setup
 
+import six
+from tqdm.auto import tqdm
+if six.PY2:
+    from urllib import urlretrieve
+else:
+    from urllib.request import urlretrieve
+import tarfile
+
+
 platform_is_windows = sys.platform == "win32"
 
 version = "0.3.0"

From 94b8b10e09a16d7b91f7aaad95c764fa9aad92e5 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 08:54:07 +0000
Subject: [PATCH 05/18] fix coredump

---
 pyopenjtalk/htsengine.pyx              |  7 +++---
 pyopenjtalk/htsengine/__init__.pxd     |  2 ++
 pyopenjtalk/openjtalk.pyx              | 34 +++++++++++++-------------
 pyopenjtalk/openjtalk/__init__.pxd     |  1 +
 pyopenjtalk/openjtalk/jpcommon.pxd     |  1 +
 pyopenjtalk/openjtalk/mecab.pxd        |  1 +
 pyopenjtalk/openjtalk/mecab2njd.pxd    |  1 +
 pyopenjtalk/openjtalk/njd.pxd          |  1 +
 pyopenjtalk/openjtalk/njd2jpcommon.pxd |  1 +
 pyopenjtalk/openjtalk/text2mecab.pxd   |  1 +
 10 files changed, 30 insertions(+), 20 deletions(-)

diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index d2117f1..868c07a 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -1,5 +1,6 @@
 # coding: utf-8
-# cython: boundscheck=False, wraparound=True
+# cython: language_level=3
+# cython: boundscheck=False, wraparound=False
 # cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 import numpy as np
@@ -12,8 +13,8 @@ from cython.parallel cimport prange
 from cpython.mem cimport PyMem_Malloc, PyMem_Free
 from libc.stdint cimport uint8_t
 
-from htsengine cimport HTS_Engine
-from htsengine cimport (
+from pyopenjtalk.htsengine cimport HTS_Engine
+from pyopenjtalk.htsengine cimport (
     HTS_Engine_initialize, HTS_Engine_load, HTS_Engine_clear, HTS_Engine_refresh,
     HTS_Engine_get_sampling_frequency, HTS_Engine_get_fperiod,
     HTS_Engine_set_speed, HTS_Engine_add_half_tone,
diff --git a/pyopenjtalk/htsengine/__init__.pxd b/pyopenjtalk/htsengine/__init__.pxd
index fc8f70c..e033ea5 100644
--- a/pyopenjtalk/htsengine/__init__.pxd
+++ b/pyopenjtalk/htsengine/__init__.pxd
@@ -1,6 +1,8 @@
 # distutils: language = c++
 
 
+# cython: language_level=3
+
 cdef extern from "HTS_engine.h" nogil:
     cdef cppclass _HTS_Engine:
         pass
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 9bb198f..1ef7923 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -1,5 +1,6 @@
 # coding: utf-8
-# cython: boundscheck=False, wraparound=True
+# cython: language_level=3
+# cython: boundscheck=False, wraparound=False
 # cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 from libc.stdint cimport uint8_t
@@ -11,17 +12,17 @@ np.import_array()
 cimport cython
 from cpython.bytes cimport PyBytes_AS_STRING
 
-from openjtalk.mecab cimport Mecab, Mecab_initialize, Mecab_load, Mecab_analysis
-from openjtalk.mecab cimport Mecab_get_feature, Mecab_get_size, Mecab_refresh, Mecab_clear
-from openjtalk.njd cimport NJD, NJD_initialize, NJD_refresh, NJD_print, NJD_clear
-from openjtalk cimport njd as _njd
-from openjtalk.jpcommon cimport JPCommon, JPCommon_initialize,JPCommon_make_label
-from openjtalk.jpcommon cimport JPCommon_get_label_size, JPCommon_get_label_feature
-from openjtalk.jpcommon cimport JPCommon_refresh, JPCommon_clear
-from openjtalk cimport njd2jpcommon
-from openjtalk.text2mecab cimport text2mecab
-from openjtalk.mecab2njd cimport mecab2njd
-from openjtalk.njd2jpcommon cimport njd2jpcommon
+from pyopenjtalk.openjtalk.mecab cimport Mecab, Mecab_initialize, Mecab_load, Mecab_analysis
+from pyopenjtalk.openjtalk.mecab cimport Mecab_get_feature, Mecab_get_size, Mecab_refresh, Mecab_clear
+from pyopenjtalk.openjtalk.njd cimport NJD, NJD_initialize, NJD_refresh, NJD_print, NJD_clear
+from pyopenjtalk.openjtalk cimport njd as _njd
+from pyopenjtalk.openjtalk.jpcommon cimport JPCommon, JPCommon_initialize,JPCommon_make_label
+from pyopenjtalk.openjtalk.jpcommon cimport JPCommon_get_label_size, JPCommon_get_label_feature
+from pyopenjtalk.openjtalk.jpcommon cimport JPCommon_refresh, JPCommon_clear
+from pyopenjtalk.openjtalk cimport njd2jpcommon
+from pyopenjtalk.openjtalk.text2mecab cimport text2mecab
+from pyopenjtalk.openjtalk.mecab2njd cimport mecab2njd
+from pyopenjtalk.openjtalk.njd2jpcommon cimport njd2jpcommon
 
 cdef inline str njd_node_get_string(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_string(node))).decode("utf-8")
@@ -53,16 +54,16 @@ cdef inline str njd_node_get_read(_njd.NJDNode* node):
 cdef inline str njd_node_get_pron(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_pron(node))).decode("utf-8")
 
-cdef inline str njd_node_get_acc(_njd.NJDNode* node):
+cdef inline int njd_node_get_acc(_njd.NJDNode* node):
     return _njd.NJDNode_get_acc(node)
 
-cdef inline str njd_node_get_mora_size(_njd.NJDNode* node):
+cdef inline int njd_node_get_mora_size(_njd.NJDNode* node):
     return _njd.NJDNode_get_mora_size(node)
 
 cdef inline str njd_node_get_chain_rule(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_chain_rule(node))).decode("utf-8")
 
-cdef inline str njd_node_get_chain_flag(_njd.NJDNode* node):
+cdef inline int njd_node_get_chain_flag(_njd.NJDNode* node):
     return _njd.NJDNode_get_chain_flag(node)
 
 
@@ -158,7 +159,7 @@ cdef class OpenJTalk:
             JPCommon_make_label(self.jpcommon)
 
             label_size = JPCommon_get_label_size(self.jpcommon)
-            JPCommon_get_label_feature(self.jpcommon)
+            label_feature = JPCommon_get_label_feature(self.jpcommon)
 
         labels = []
         cdef int i
@@ -177,7 +178,6 @@ cdef class OpenJTalk:
             JPCommon_refresh(self.jpcommon)
             NJD_refresh(self.njd)
             Mecab_refresh(self.mecab)
-
         return njd_results, labels
 
     def g2p(self, object text, bint kana=False, bint join=True):
diff --git a/pyopenjtalk/openjtalk/__init__.pxd b/pyopenjtalk/openjtalk/__init__.pxd
index e69de29..019523c 100644
--- a/pyopenjtalk/openjtalk/__init__.pxd
+++ b/pyopenjtalk/openjtalk/__init__.pxd
@@ -0,0 +1 @@
+# cython: language_level=3
\ No newline at end of file
diff --git a/pyopenjtalk/openjtalk/jpcommon.pxd b/pyopenjtalk/openjtalk/jpcommon.pxd
index b120fc1..2dd450a 100644
--- a/pyopenjtalk/openjtalk/jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/jpcommon.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 from libc.stdio cimport FILE
 
diff --git a/pyopenjtalk/openjtalk/mecab.pxd b/pyopenjtalk/openjtalk/mecab.pxd
index 2449db0..2aa39e5 100644
--- a/pyopenjtalk/openjtalk/mecab.pxd
+++ b/pyopenjtalk/openjtalk/mecab.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 cdef extern from "mecab.h" nogil:
     cdef cppclass Mecab:
diff --git a/pyopenjtalk/openjtalk/mecab2njd.pxd b/pyopenjtalk/openjtalk/mecab2njd.pxd
index a86d790..d8fb8c4 100644
--- a/pyopenjtalk/openjtalk/mecab2njd.pxd
+++ b/pyopenjtalk/openjtalk/mecab2njd.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 from .njd cimport NJD
 
diff --git a/pyopenjtalk/openjtalk/njd.pxd b/pyopenjtalk/openjtalk/njd.pxd
index 2831324..abc1f2d 100644
--- a/pyopenjtalk/openjtalk/njd.pxd
+++ b/pyopenjtalk/openjtalk/njd.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 from libc.stdio cimport FILE
 
diff --git a/pyopenjtalk/openjtalk/njd2jpcommon.pxd b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
index 4d6fb62..680aadf 100644
--- a/pyopenjtalk/openjtalk/njd2jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 from .jpcommon cimport JPCommon
 from .njd cimport NJD
diff --git a/pyopenjtalk/openjtalk/text2mecab.pxd b/pyopenjtalk/openjtalk/text2mecab.pxd
index 718f7a5..0190f1a 100644
--- a/pyopenjtalk/openjtalk/text2mecab.pxd
+++ b/pyopenjtalk/openjtalk/text2mecab.pxd
@@ -1,4 +1,5 @@
 # distutils: language = c++
+# cython: language_level=3
 
 cdef extern from "text2mecab.h" nogil:
     void text2mecab(char *output, const char *input)

From 8bde56c0219805b2b0e606e816c7cdcfe9253dcd Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 09:06:55 +0000
Subject: [PATCH 06/18] add build ci

---
 .github/workflows/build_whl.yml | 37 ++++++++++++++++++++++++++++++++
 .github/workflows/upload.yml    | 38 +++++++++++++++++++++++++++++++++
 2 files changed, 75 insertions(+)
 create mode 100644 .github/workflows/build_whl.yml
 create mode 100644 .github/workflows/upload.yml

diff --git a/.github/workflows/build_whl.yml b/.github/workflows/build_whl.yml
new file mode 100644
index 0000000..6799c9a
--- /dev/null
+++ b/.github/workflows/build_whl.yml
@@ -0,0 +1,37 @@
+name: build wheel
+
+on:
+  workflow_dispatch:
+
+jobs:
+  build:
+
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        python-version: ["3.6", "3.7", "3.8", "3.9", "3.10"]
+        os: [ubuntu-latest, macos-latest, windows-latest]
+      fail-fast: false
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Check out recursively
+      run: git submodule update --init --recursive
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        python -m pip install --upgrade setuptools
+        python -m pip install --upgrade wheel
+        pip install flake8 pytest
+        pip install cython numpy tqdm six
+    - name: build_whl
+      run: |
+        python setup.py sdist bdist_wheel
+    - uses: actions/upload-artifact@v2
+      with:
+        name: ${{ matrix.os }}-${{ matrix.python-version }}
+        path: dist/*.whl
\ No newline at end of file
diff --git a/.github/workflows/upload.yml b/.github/workflows/upload.yml
new file mode 100644
index 0000000..748a021
--- /dev/null
+++ b/.github/workflows/upload.yml
@@ -0,0 +1,38 @@
+name: upload
+
+on:
+  workflow_dispatch:
+
+jobs:
+  build:
+
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        python-version: ["3.6", "3.7", "3.8", "3.9", "3.10"]
+        os: [ubuntu-latest, macos-latest, windows-latest]
+      fail-fast: false
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Check out recursively
+      run: git submodule update --init --recursive
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        python -m pip install --upgrade setuptools
+        python -m pip install --upgrade wheel
+        pip install flake8 pytest
+        pip install twine
+        pip install cython numpy tqdm six
+    - name: build_whl
+      run: |
+        python setup.py sdist bdist_wheel
+        python changename.py
+    - name: Publish package
+      run: |
+        twine upload dist/* -u ${{ secrets.PYPI_USER }} -p ${{ secrets.PYPI_PASSWORD }}
\ No newline at end of file

From bcfa63a2b722f6ccd08d1aa725b40669373bdf78 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 09:19:29 +0000
Subject: [PATCH 07/18] fix macox build

---
 pyopenjtalk/htsengine.pyx |  4 ++--
 pyopenjtalk/openjtalk.pyx |  2 +-
 setup.py                  | 27 +++++++++++++++++++++++----
 3 files changed, 26 insertions(+), 7 deletions(-)

diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index 868c07a..cdc6018 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -117,9 +117,9 @@ cdef class HTSEngine:
         cdef size_t nsamples = HTS_Engine_get_nsamples(self.engine)
         cdef np.ndarray[np.float64_t, ndim=1] speech = np.zeros([nsamples], dtype=np.float64)
         cdef double[::1] speech_view = speech
-        cdef size_t index
+        cdef int index
         for index in prange(nsamples, nogil=True):
-            speech_view[index] = HTS_Engine_get_generated_speech(self.engine, index)
+            speech_view[index] = HTS_Engine_get_generated_speech(self.engine, <size_t>index)
         return speech
 
     cpdef inline str get_fullcontext_label_format(self):
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 1ef7923..4f37112 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -1,6 +1,6 @@
 # coding: utf-8
 # cython: language_level=3
-# cython: boundscheck=False, wraparound=False
+# cython: boundscheck=False, wraparound=True
 # cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 from libc.stdint cimport uint8_t
diff --git a/setup.py b/setup.py
index 971a08b..6308e10 100644
--- a/setup.py
+++ b/setup.py
@@ -1,6 +1,7 @@
 import os
 import subprocess
 import sys
+import platform
 from distutils.errors import DistutilsExecError
 from distutils.spawn import spawn
 from distutils.version import LooseVersion
@@ -70,6 +71,24 @@ def build_extensions(self):
     if not os.path.exists(join("pyopenjtalk", "openjtalk" + ext)):
         raise RuntimeError("Cython is required to generate C++ code")
 
+# make openmp available
+system = platform.system()
+if system == "Windows":
+    extra_compile_args = []
+    extra_link_args = ['/openmp']
+elif system == "Linux":
+    extra_compile_args = ['-fopenmp']
+    extra_link_args = ['-fopenmp']
+elif system == "Darwin":
+    os.system("brew install llvm libomp")
+    os.system("brew install clang-omp")
+    # os.environ["CPP"] = "/usr/local/opt/llvm/bin/clang"
+    extra_compile_args = []
+    extra_link_args = []
+else:
+    extra_compile_args = ['-fopenmp']
+    extra_link_args = ['-fopenmp']
+
 
 # Workaround for `distutils.spawn` problem on Windows python < 3.9
 # See details: [bpo-39763: distutils.spawn now uses subprocess (GH-18743)]
@@ -180,8 +199,8 @@ def escape_macros(macros):
         name="pyopenjtalk.openjtalk",
         sources=[join("pyopenjtalk", "openjtalk" + ext)] + all_src,
         include_dirs=[np.get_include()] + include_dirs,
-        extra_compile_args=['-fopenmp'],
-        extra_link_args=['-fopenmp'],
+        extra_compile_args=extra_compile_args,
+        extra_link_args=extra_link_args,
         language="c++",
         define_macros=custom_define_macros(
             [
@@ -204,8 +223,8 @@ def escape_macros(macros):
         name="pyopenjtalk.htsengine",
         sources=[join("pyopenjtalk", "htsengine" + ext)] + all_htsengine_src,
         include_dirs=[np.get_include(), join(htsengine_src_top, "include")],
-        extra_compile_args=['-fopenmp'],
-        extra_link_args=['-fopenmp'],
+        extra_compile_args=extra_compile_args,
+        extra_link_args=extra_link_args,
         libraries=["winmm"] if platform_is_windows else [],
         language="c++",
         define_macros=custom_define_macros(

From d351b356ebb5ed790099ecd53fccb49d76004008 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 09:19:29 +0000
Subject: [PATCH 08/18] fix macox build

---
 pyopenjtalk/htsengine.pyx |  4 ++--
 pyopenjtalk/openjtalk.pyx |  2 +-
 setup.py                  | 27 +++++++++++++++++++++++----
 3 files changed, 26 insertions(+), 7 deletions(-)

diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index 868c07a..cdc6018 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -117,9 +117,9 @@ cdef class HTSEngine:
         cdef size_t nsamples = HTS_Engine_get_nsamples(self.engine)
         cdef np.ndarray[np.float64_t, ndim=1] speech = np.zeros([nsamples], dtype=np.float64)
         cdef double[::1] speech_view = speech
-        cdef size_t index
+        cdef int index
         for index in prange(nsamples, nogil=True):
-            speech_view[index] = HTS_Engine_get_generated_speech(self.engine, index)
+            speech_view[index] = HTS_Engine_get_generated_speech(self.engine, <size_t>index)
         return speech
 
     cpdef inline str get_fullcontext_label_format(self):
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 1ef7923..4f37112 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -1,6 +1,6 @@
 # coding: utf-8
 # cython: language_level=3
-# cython: boundscheck=False, wraparound=False
+# cython: boundscheck=False, wraparound=True
 # cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 from libc.stdint cimport uint8_t
diff --git a/setup.py b/setup.py
index 971a08b..8a6ddcc 100644
--- a/setup.py
+++ b/setup.py
@@ -1,6 +1,7 @@
 import os
 import subprocess
 import sys
+import platform
 from distutils.errors import DistutilsExecError
 from distutils.spawn import spawn
 from distutils.version import LooseVersion
@@ -70,6 +71,24 @@ def build_extensions(self):
     if not os.path.exists(join("pyopenjtalk", "openjtalk" + ext)):
         raise RuntimeError("Cython is required to generate C++ code")
 
+# make openmp available
+system = platform.system()
+if system == "Windows":
+    extra_compile_args = []
+    extra_link_args = ['/openmp']
+elif system == "Linux":
+    extra_compile_args = ['-fopenmp']
+    extra_link_args = ['-fopenmp']
+elif system == "Darwin":
+    os.system("brew install llvm libomp")
+    os.system("brew install clang-omp")
+    # os.environ["CPP"] = "/usr/local/opt/llvm/bin/clang"
+    extra_compile_args = ["-Xpreprocessor", "-fopenmp"]
+    extra_link_args = ["-Xpreprocessor", "-fopenmp"]
+else:
+    extra_compile_args = ['-fopenmp']
+    extra_link_args = ['-fopenmp']
+
 
 # Workaround for `distutils.spawn` problem on Windows python < 3.9
 # See details: [bpo-39763: distutils.spawn now uses subprocess (GH-18743)]
@@ -180,8 +199,8 @@ def escape_macros(macros):
         name="pyopenjtalk.openjtalk",
         sources=[join("pyopenjtalk", "openjtalk" + ext)] + all_src,
         include_dirs=[np.get_include()] + include_dirs,
-        extra_compile_args=['-fopenmp'],
-        extra_link_args=['-fopenmp'],
+        extra_compile_args=extra_compile_args,
+        extra_link_args=extra_link_args,
         language="c++",
         define_macros=custom_define_macros(
             [
@@ -204,8 +223,8 @@ def escape_macros(macros):
         name="pyopenjtalk.htsengine",
         sources=[join("pyopenjtalk", "htsengine" + ext)] + all_htsengine_src,
         include_dirs=[np.get_include(), join(htsengine_src_top, "include")],
-        extra_compile_args=['-fopenmp'],
-        extra_link_args=['-fopenmp'],
+        extra_compile_args=extra_compile_args,
+        extra_link_args=extra_link_args,
         libraries=["winmm"] if platform_is_windows else [],
         language="c++",
         define_macros=custom_define_macros(

From b3de8752bc13593b1a2c634c023c117cae6e6301 Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:02:58 +0800
Subject: [PATCH 09/18] fix args

---
 setup.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/setup.py b/setup.py
index 3eaab0a..36fae85 100644
--- a/setup.py
+++ b/setup.py
@@ -93,8 +93,8 @@ def build_extensions(self):
     extra_compile_args = ['-fopenmp']
     extra_link_args = ['-fopenmp']
 elif system == "Darwin":
-    os.system("brew install libomp")
-    extra_compile_args = ['-Xpreprocessor -fopenmp']
+    os.system('brew install libomp')
+    extra_compile_args = ['-Xpreprocessor', '-fopenmp']
     extra_link_args = ['-lomp']
 else:
     extra_compile_args = ['-fopenmp']

From acbebe5bb78107a0ded15a21a5933e7bf7974bf4 Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:03:48 +0800
Subject: [PATCH 10/18] add ignore

---
 .gitignore | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.gitignore b/.gitignore
index 4d3b0cb..8b2bc54 100644
--- a/.gitignore
+++ b/.gitignore
@@ -194,3 +194,4 @@ Temporary Items
 
 # Linux trash folder which might appear on any partition or disk
 .Trash-*
+dic.tar.gz

From a342129e6e4c65718c0a210a04e479b05b012efb Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:07:22 +0800
Subject: [PATCH 11/18] add link arg

---
 setup.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/setup.py b/setup.py
index 36fae85..ecb4117 100644
--- a/setup.py
+++ b/setup.py
@@ -95,7 +95,7 @@ def build_extensions(self):
 elif system == "Darwin":
     os.system('brew install libomp')
     extra_compile_args = ['-Xpreprocessor', '-fopenmp']
-    extra_link_args = ['-lomp']
+    extra_link_args = ['-L/usr/local/lib', '-lomp']
 else:
     extra_compile_args = ['-fopenmp']
     extra_link_args = ['-fopenmp']
@@ -176,7 +176,7 @@ def update_to(self, b=1, bsize=1, tsize=None):
             self.total = tsize
         return self.update(b * bsize - self.n)
 
-
+"""
 # extract dic
 filename = "dic.tar.gz"
 print('Downloading: "{}"'.format(_DICT_URL))
@@ -193,7 +193,7 @@ def update_to(self, b=1, bsize=1, tsize=None):
 with tarfile.open(filename, mode="r|gz") as f:
     f.extractall(path="./")
 os.remove(filename)
-
+"""
 
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h
@@ -329,7 +329,7 @@ def run(self):
     url="https://github.com/r9y9/pyopenjtalk",
     license="MIT",
     packages=find_packages(),
-    package_data={"": ["htsvoice/*", f"{_dict_folder_name}/*"]},
+    package_data={"": ["htsvoice/*"]},
     ext_modules=ext_modules,
     cmdclass=cmdclass,
     install_requires=[

From 13d1d4bbb5f5a7e34606bcebdb47ac349e377a07 Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:25:14 +0800
Subject: [PATCH 12/18] r

---
 setup.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/setup.py b/setup.py
index ecb4117..ec27e14 100644
--- a/setup.py
+++ b/setup.py
@@ -176,7 +176,6 @@ def update_to(self, b=1, bsize=1, tsize=None):
             self.total = tsize
         return self.update(b * bsize - self.n)
 
-"""
 # extract dic
 filename = "dic.tar.gz"
 print('Downloading: "{}"'.format(_DICT_URL))
@@ -193,7 +192,6 @@ def update_to(self, b=1, bsize=1, tsize=None):
 with tarfile.open(filename, mode="r|gz") as f:
     f.extractall(path="./")
 os.remove(filename)
-"""
 
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h

From 1fc6f63a40cb06ccf32ae6ee58ca5a43ad2f5fdd Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:26:08 +0800
Subject: [PATCH 13/18] eee

---
 setup.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index ec27e14..3f26a8a 100644
--- a/setup.py
+++ b/setup.py
@@ -327,7 +327,7 @@ def run(self):
     url="https://github.com/r9y9/pyopenjtalk",
     license="MIT",
     packages=find_packages(),
-    package_data={"": ["htsvoice/*"]},
+    package_data={"": ["htsvoice/*", f"{_dict_folder_name}/*"]},
     ext_modules=ext_modules,
     cmdclass=cmdclass,
     install_requires=[

From 1c493f5ef4c56c7b93086f4a3153c72a143fdd91 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E6=BA=90=E6=96=87=E9=9B=A8?=
 <41315874+fumiama@users.noreply.github.com>
Date: Fri, 12 Aug 2022 21:27:26 +0800
Subject: [PATCH 14/18] Update setup.py

---
 setup.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/setup.py b/setup.py
index 3f26a8a..b7d109d 100644
--- a/setup.py
+++ b/setup.py
@@ -176,6 +176,7 @@ def update_to(self, b=1, bsize=1, tsize=None):
             self.total = tsize
         return self.update(b * bsize - self.n)
 
+
 # extract dic
 filename = "dic.tar.gz"
 print('Downloading: "{}"'.format(_DICT_URL))
@@ -193,6 +194,7 @@ def update_to(self, b=1, bsize=1, tsize=None):
     f.extractall(path="./")
 os.remove(filename)
 
+
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h
 # we could do it on python side but it would be very tricky,

From a9873b82e8a71bc11a2ee3f837198905aa3407f2 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 13:16:48 +0000
Subject: [PATCH 15/18] add download flag

---
 setup.py | 27 ++++++++-------------------
 1 file changed, 8 insertions(+), 19 deletions(-)

diff --git a/setup.py b/setup.py
index 5093dd9..f6c51ff 100644
--- a/setup.py
+++ b/setup.py
@@ -2,6 +2,7 @@
 import subprocess
 import sys
 import platform
+import shutil
 from distutils.errors import DistutilsExecError
 from distutils.spawn import spawn
 from distutils.version import LooseVersion
@@ -16,7 +17,6 @@
 from setuptools import Extension, find_packages, setup
 
 import six
-from tqdm.auto import tqdm
 if six.PY2:
     from urllib import urlretrieve
 else:
@@ -171,31 +171,20 @@ def escape_macros(macros):
 src_top = join("lib", "open_jtalk", "src")
 
 
-# https://github.com/tqdm/tqdm#hooks-and-callbacks
-class _TqdmUpTo(tqdm):  # type: ignore
-    def update_to(self, b=1, bsize=1, tsize=None):
-        if tsize is not None:
-            self.total = tsize
-        return self.update(b * bsize - self.n)
-
 
 # extract dic
 filename = "dic.tar.gz"
-print('Downloading: "{}"'.format(_DICT_URL))
-with _TqdmUpTo(
-    unit="B",
-    unit_scale=True,
-    unit_divisor=1024,
-    miniters=1,
-    desc="dic.tar.gz",
-) as t:  # all optional kwargs
-    urlretrieve(_DICT_URL, filename, reporthook=t.update_to)
-    t.total = t.n
+print(f"Downloading: {_DICT_URL}")
+urlretrieve(_DICT_URL, filename)
+print("Download complete")
+
 print("Extracting tar file {}".format(filename))
 with tarfile.open(filename, mode="r|gz") as f:
     f.extractall(path="./")
 os.remove(filename)
-
+print("Extract complete")
+shutil.copytree(f"./{_dict_folder_name}", f"./pyopenjtalk/{_dict_folder_name}")
+sys.stdout.flush()
 
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h

From 89843b39754e368c79be118698b07ca8afc1d002 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Fri, 12 Aug 2022 14:35:58 +0000
Subject: [PATCH 16/18] reformat with isort and black

---
 pyopenjtalk/__init__.py                |  6 +++---
 pyopenjtalk/htsengine.pyx              | 20 +++++++++++---------
 pyopenjtalk/openjtalk.pyx              | 23 ++++++++++++++++-------
 pyopenjtalk/openjtalk/jpcommon.pxd     |  1 +
 pyopenjtalk/openjtalk/mecab2njd.pxd    |  3 ++-
 pyopenjtalk/openjtalk/njd.pxd          |  1 +
 pyopenjtalk/openjtalk/njd2jpcommon.pxd |  5 +++--
 setup.py                               | 19 +++++++++----------
 8 files changed, 46 insertions(+), 32 deletions(-)

diff --git a/pyopenjtalk/__init__.py b/pyopenjtalk/__init__.py
index a266104..7b83da1 100644
--- a/pyopenjtalk/__init__.py
+++ b/pyopenjtalk/__init__.py
@@ -13,12 +13,12 @@
 import tarfile
 
 try:
-    from .version import __version__  # NOQA
+    from pyopenjtalk.version import __version__  # NOQA
 except ImportError:
     raise ImportError("BUG: version.py doesn't exist. Please file a bug report.")
 
-from .htsengine import HTSEngine
-from .openjtalk import OpenJTalk
+from pyopenjtalk.htsengine import HTSEngine
+from pyopenjtalk.openjtalk import OpenJTalk
 
 # Dictionary directory
 # defaults to the package directory where the dictionary will be automatically downloaded
diff --git a/pyopenjtalk/htsengine.pyx b/pyopenjtalk/htsengine.pyx
index cdc6018..b30e813 100644
--- a/pyopenjtalk/htsengine.pyx
+++ b/pyopenjtalk/htsengine.pyx
@@ -6,21 +6,23 @@
 import numpy as np
 
 cimport numpy as np
+
 np.import_array()
 
 cimport cython
+from cpython.mem cimport PyMem_Free, PyMem_Malloc
 from cython.parallel cimport prange
-from cpython.mem cimport PyMem_Malloc, PyMem_Free
 from libc.stdint cimport uint8_t
 
-from pyopenjtalk.htsengine cimport HTS_Engine
-from pyopenjtalk.htsengine cimport (
-    HTS_Engine_initialize, HTS_Engine_load, HTS_Engine_clear, HTS_Engine_refresh,
-    HTS_Engine_get_sampling_frequency, HTS_Engine_get_fperiod,
-    HTS_Engine_set_speed, HTS_Engine_add_half_tone,
-    HTS_Engine_synthesize_from_strings,
-    HTS_Engine_get_generated_speech, HTS_Engine_get_nsamples
-)
+from pyopenjtalk.htsengine cimport (HTS_Engine, HTS_Engine_add_half_tone,
+                                    HTS_Engine_clear, HTS_Engine_get_fperiod,
+                                    HTS_Engine_get_generated_speech,
+                                    HTS_Engine_get_nsamples,
+                                    HTS_Engine_get_sampling_frequency,
+                                    HTS_Engine_initialize, HTS_Engine_load,
+                                    HTS_Engine_refresh, HTS_Engine_set_speed,
+                                    HTS_Engine_synthesize_from_strings)
+
 
 @cython.final
 @cython.no_gc
diff --git a/pyopenjtalk/openjtalk.pyx b/pyopenjtalk/openjtalk.pyx
index 4f37112..eda7ae1 100644
--- a/pyopenjtalk/openjtalk.pyx
+++ b/pyopenjtalk/openjtalk.pyx
@@ -4,25 +4,34 @@
 # cython: c_string_type=unicode, c_string_encoding=ascii, cdivision=True
 
 from libc.stdint cimport uint8_t
+
 import numpy as np
 
 cimport numpy as np
+
 np.import_array()
 
 cimport cython
 from cpython.bytes cimport PyBytes_AS_STRING
 
-from pyopenjtalk.openjtalk.mecab cimport Mecab, Mecab_initialize, Mecab_load, Mecab_analysis
-from pyopenjtalk.openjtalk.mecab cimport Mecab_get_feature, Mecab_get_size, Mecab_refresh, Mecab_clear
-from pyopenjtalk.openjtalk.njd cimport NJD, NJD_initialize, NJD_refresh, NJD_print, NJD_clear
 from pyopenjtalk.openjtalk cimport njd as _njd
-from pyopenjtalk.openjtalk.jpcommon cimport JPCommon, JPCommon_initialize,JPCommon_make_label
-from pyopenjtalk.openjtalk.jpcommon cimport JPCommon_get_label_size, JPCommon_get_label_feature
-from pyopenjtalk.openjtalk.jpcommon cimport JPCommon_refresh, JPCommon_clear
 from pyopenjtalk.openjtalk cimport njd2jpcommon
-from pyopenjtalk.openjtalk.text2mecab cimport text2mecab
+from pyopenjtalk.openjtalk.jpcommon cimport (JPCommon, JPCommon_clear,
+                                             JPCommon_get_label_feature,
+                                             JPCommon_get_label_size,
+                                             JPCommon_initialize,
+                                             JPCommon_make_label,
+                                             JPCommon_refresh)
+from pyopenjtalk.openjtalk.mecab cimport (Mecab, Mecab_analysis, Mecab_clear,
+                                          Mecab_get_feature, Mecab_get_size,
+                                          Mecab_initialize, Mecab_load,
+                                          Mecab_refresh)
 from pyopenjtalk.openjtalk.mecab2njd cimport mecab2njd
+from pyopenjtalk.openjtalk.njd cimport (NJD, NJD_clear, NJD_initialize,
+                                        NJD_print, NJD_refresh)
 from pyopenjtalk.openjtalk.njd2jpcommon cimport njd2jpcommon
+from pyopenjtalk.openjtalk.text2mecab cimport text2mecab
+
 
 cdef inline str njd_node_get_string(_njd.NJDNode* node):
     return (<bytes>(_njd.NJDNode_get_string(node))).decode("utf-8")
diff --git a/pyopenjtalk/openjtalk/jpcommon.pxd b/pyopenjtalk/openjtalk/jpcommon.pxd
index 2dd450a..3667d1e 100644
--- a/pyopenjtalk/openjtalk/jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/jpcommon.pxd
@@ -3,6 +3,7 @@
 
 from libc.stdio cimport FILE
 
+
 cdef extern from "jpcommon.h" nogil:
     cdef cppclass JPCommonNode:
         char *pron
diff --git a/pyopenjtalk/openjtalk/mecab2njd.pxd b/pyopenjtalk/openjtalk/mecab2njd.pxd
index d8fb8c4..f42d3ec 100644
--- a/pyopenjtalk/openjtalk/mecab2njd.pxd
+++ b/pyopenjtalk/openjtalk/mecab2njd.pxd
@@ -1,7 +1,8 @@
 # distutils: language = c++
 # cython: language_level=3
 
-from .njd cimport NJD
+from pyopenjtalk.openjtalk.njd cimport NJD
+
 
 cdef extern from "mecab2njd.h" nogil:
     void mecab2njd(NJD * njd, char **feature, int size);
diff --git a/pyopenjtalk/openjtalk/njd.pxd b/pyopenjtalk/openjtalk/njd.pxd
index abc1f2d..ef2b8b3 100644
--- a/pyopenjtalk/openjtalk/njd.pxd
+++ b/pyopenjtalk/openjtalk/njd.pxd
@@ -3,6 +3,7 @@
 
 from libc.stdio cimport FILE
 
+
 cdef extern from "njd.h" nogil:
     cdef cppclass NJDNode:
         char *string
diff --git a/pyopenjtalk/openjtalk/njd2jpcommon.pxd b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
index 680aadf..a803032 100644
--- a/pyopenjtalk/openjtalk/njd2jpcommon.pxd
+++ b/pyopenjtalk/openjtalk/njd2jpcommon.pxd
@@ -1,8 +1,9 @@
 # distutils: language = c++
 # cython: language_level=3
 
-from .jpcommon cimport JPCommon
-from .njd cimport NJD
+from pyopenjtalk.openjtalk.jpcommon cimport JPCommon
+from pyopenjtalk.openjtalk.njd cimport NJD
+
 
 cdef extern from "njd2jpcommon.h" nogil:
     void njd2jpcommon(JPCommon * jpcommon, NJD * njd)
diff --git a/setup.py b/setup.py
index f6c51ff..700e4b8 100644
--- a/setup.py
+++ b/setup.py
@@ -1,8 +1,8 @@
 import os
-import subprocess
-import sys
 import platform
 import shutil
+import subprocess
+import sys
 from distutils.errors import DistutilsExecError
 from distutils.spawn import spawn
 from distutils.version import LooseVersion
@@ -14,15 +14,15 @@
 import numpy as np
 import setuptools.command.build_py
 import setuptools.command.develop
+import six
 from setuptools import Extension, find_packages, setup
 
-import six
 if six.PY2:
     from urllib import urlretrieve
 else:
     from urllib.request import urlretrieve
-import tarfile
 
+import tarfile
 
 platform_is_windows = sys.platform == "win32"
 
@@ -88,10 +88,10 @@ def build_extensions(self):
 system = platform.system()
 if system == "Windows":
     extra_compile_args = []
-    extra_link_args = ['/openmp']
+    extra_link_args = ["/openmp"]
 elif system == "Linux":
-    extra_compile_args = ['-fopenmp']
-    extra_link_args = ['-fopenmp']
+    extra_compile_args = ["-fopenmp"]
+    extra_link_args = ["-fopenmp"]
 elif system == "Darwin":
     os.system("brew install llvm libomp")
     os.system("brew install clang-omp")
@@ -99,8 +99,8 @@ def build_extensions(self):
     extra_compile_args = ["-Xpreprocessor", "-fopenmp"]
     extra_link_args = ["-Xpreprocessor", "-fopenmp"]
 else:
-    extra_compile_args = ['-fopenmp']
-    extra_link_args = ['-fopenmp']
+    extra_compile_args = ["-fopenmp"]
+    extra_link_args = ["-fopenmp"]
 
 
 # Workaround for `distutils.spawn` problem on Windows python < 3.9
@@ -171,7 +171,6 @@ def escape_macros(macros):
 src_top = join("lib", "open_jtalk", "src")
 
 
-
 # extract dic
 filename = "dic.tar.gz"
 print(f"Downloading: {_DICT_URL}")

From 70f37d2ab8b6576ea421771b53500fa83c0bb824 Mon Sep 17 00:00:00 2001
From: fumiama <41315874+fumiama@users.noreply.github.com>
Date: Mon, 15 Aug 2022 10:11:55 +0800
Subject: [PATCH 17/18] feat: remove six and lazy init and tqdm

---
 .github/workflows/build_whl.yml |  2 +-
 pyopenjtalk/__init__.py         | 48 ---------------------------------
 setup.py                        |  8 +-----
 3 files changed, 2 insertions(+), 56 deletions(-)

diff --git a/.github/workflows/build_whl.yml b/.github/workflows/build_whl.yml
index 6799c9a..14c450a 100644
--- a/.github/workflows/build_whl.yml
+++ b/.github/workflows/build_whl.yml
@@ -27,7 +27,7 @@ jobs:
         python -m pip install --upgrade setuptools
         python -m pip install --upgrade wheel
         pip install flake8 pytest
-        pip install cython numpy tqdm six
+        pip install cython numpy tqdm
     - name: build_whl
       run: |
         python setup.py sdist bdist_wheel
diff --git a/pyopenjtalk/__init__.py b/pyopenjtalk/__init__.py
index 7b83da1..3aa2d8e 100644
--- a/pyopenjtalk/__init__.py
+++ b/pyopenjtalk/__init__.py
@@ -1,16 +1,7 @@
 import os
-from os.path import exists
 
 import pkg_resources
-import six
-from tqdm.auto import tqdm
 
-if six.PY2:
-    from urllib import urlretrieve
-else:
-    from urllib.request import urlretrieve
-
-import tarfile
 
 try:
     from pyopenjtalk.version import __version__  # NOQA
@@ -26,8 +17,6 @@
     "OPEN_JTALK_DICT_DIR",
     pkg_resources.resource_filename(__name__, "open_jtalk_dic_utf_8-1.11"),
 ).encode("utf-8")
-_dict_download_url = "https://github.com/r9y9/open_jtalk/releases/download/v1.11.1"
-_DICT_URL = f"{_dict_download_url}/open_jtalk_dic_utf_8-1.11.tar.gz"
 
 # Default mei_normal.voice for HMM-based TTS
 DEFAULT_HTS_VOICE = pkg_resources.resource_filename(
@@ -41,41 +30,6 @@
 _global_htsengine = None
 
 
-# https://github.com/tqdm/tqdm#hooks-and-callbacks
-class _TqdmUpTo(tqdm):  # type: ignore
-    def update_to(self, b=1, bsize=1, tsize=None):
-        if tsize is not None:
-            self.total = tsize
-        return self.update(b * bsize - self.n)
-
-
-def _extract_dic():
-    global OPEN_JTALK_DICT_DIR
-    filename = pkg_resources.resource_filename(__name__, "dic.tar.gz")
-    print('Downloading: "{}"'.format(_DICT_URL))
-    with _TqdmUpTo(
-        unit="B",
-        unit_scale=True,
-        unit_divisor=1024,
-        miniters=1,
-        desc="dic.tar.gz",
-    ) as t:  # all optional kwargs
-        urlretrieve(_DICT_URL, filename, reporthook=t.update_to)
-        t.total = t.n
-    print("Extracting tar file {}".format(filename))
-    with tarfile.open(filename, mode="r|gz") as f:
-        f.extractall(path=pkg_resources.resource_filename(__name__, ""))
-    OPEN_JTALK_DICT_DIR = pkg_resources.resource_filename(
-        __name__, "open_jtalk_dic_utf_8-1.11"
-    ).encode("utf-8")
-    os.remove(filename)
-
-
-def _lazy_init():
-    if not exists(OPEN_JTALK_DICT_DIR):
-        _extract_dic()
-
-
 def g2p(*args, **kwargs):
     """Grapheme-to-phoeneme (G2P) conversion
 
@@ -93,7 +47,6 @@ def g2p(*args, **kwargs):
     """
     global _global_jtalk
     if _global_jtalk is None:
-        _lazy_init()
         _global_jtalk = OpenJTalk(dn_mecab=OPEN_JTALK_DICT_DIR)
     return _global_jtalk.g2p(*args, **kwargs)
 
@@ -164,6 +117,5 @@ def run_frontend(text, verbose=0):
     """
     global _global_jtalk
     if _global_jtalk is None:
-        _lazy_init()
         _global_jtalk = OpenJTalk(dn_mecab=OPEN_JTALK_DICT_DIR)
     return _global_jtalk.run_frontend(text, verbose)
diff --git a/setup.py b/setup.py
index 6115e32..a32205c 100644
--- a/setup.py
+++ b/setup.py
@@ -14,13 +14,9 @@
 import numpy as np
 import setuptools.command.build_py
 import setuptools.command.develop
-import six
 from setuptools import Extension, find_packages, setup
 
-if six.PY2:
-    from urllib import urlretrieve
-else:
-    from urllib.request import urlretrieve
+from urllib.request import urlretrieve
 
 import tarfile
 
@@ -323,8 +319,6 @@ def run(self):
     install_requires=[
         "numpy >= 1.20.0",
         "cython >= " + min_cython_ver,
-        "six",
-        "tqdm",
     ],
     tests_require=["nose", "coverage"],
     extras_require={

From bc8699a5375036c4d6d63194eed57a6a7e35a1f0 Mon Sep 17 00:00:00 2001
From: synodriver <624805065@qq.com>
Date: Mon, 15 Aug 2022 09:06:54 +0000
Subject: [PATCH 18/18] ignore exc when datafile exists

---
 pyopenjtalk/__init__.py |  1 -
 setup.py                | 13 +++++++------
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/pyopenjtalk/__init__.py b/pyopenjtalk/__init__.py
index 3aa2d8e..2e0e3f5 100644
--- a/pyopenjtalk/__init__.py
+++ b/pyopenjtalk/__init__.py
@@ -2,7 +2,6 @@
 
 import pkg_resources
 
-
 try:
     from pyopenjtalk.version import __version__  # NOQA
 except ImportError:
diff --git a/setup.py b/setup.py
index a32205c..ce6e3fe 100644
--- a/setup.py
+++ b/setup.py
@@ -3,6 +3,7 @@
 import shutil
 import subprocess
 import sys
+import tarfile
 from distutils.errors import DistutilsExecError
 from distutils.spawn import spawn
 from distutils.version import LooseVersion
@@ -10,16 +11,13 @@
 from itertools import chain
 from os.path import exists, join
 from subprocess import run
+from urllib.request import urlretrieve
 
 import numpy as np
 import setuptools.command.build_py
 import setuptools.command.develop
 from setuptools import Extension, find_packages, setup
 
-from urllib.request import urlretrieve
-
-import tarfile
-
 platform_is_windows = sys.platform == "win32"
 
 version = "0.3.0"
@@ -176,8 +174,11 @@ def escape_macros(macros):
     f.extractall(path="./")
 os.remove(filename)
 print("Extract complete")
-shutil.copytree(f"./{_dict_folder_name}", f"./pyopenjtalk/{_dict_folder_name}")
-sys.stdout.flush()
+try:
+    shutil.copytree(f"./{_dict_folder_name}", f"./pyopenjtalk/{_dict_folder_name}")
+    sys.stdout.flush()
+except FileExistsError:
+    pass
 
 # generate config.h for mecab
 # NOTE: need to run cmake to generate config.h