move init to C++ and debug stuff

torchaudio/_extension/utils.py

@@ -67,7 +67,6 @@ def _init_sox():
     _load_lib("libtorchaudio_sox")
     import torchaudio.lib._torchaudio_sox  # noqa
 
-    torchaudio.lib._torchaudio_sox.init_libsox()
     torchaudio.lib._torchaudio_sox.set_verbosity(0)
 
     import atexit

torchaudio/csrc/sox/effects.cpp

@@ -130,7 +130,6 @@ auto apply_effects_file(
       dtype,
       normalize.value_or(true),
       channels_first_);
-
   return std::tuple<torch::Tensor, int64_t>(
       tensor, chain.getOutputSampleRate());
 }
@@ -139,14 +138,8 @@ TORCH_LIBRARY_FRAGMENT(torchaudio, m) {
   m.def(
       "torchaudio::sox_effects_initialize_sox_effects",
       &initialize_sox_effects);
-  m.def(
-      "torchaudio::sox_effects_shutdown_sox_effects",
-      &shutdown_sox_effects);
-  m.def(
-      "torchaudio::sox_effects_apply_effects_tensor",
-      &apply_effects_tensor);
-  m.def(
-      "torchaudio::sox_effects_apply_effects_file",
-      &apply_effects_file);
+  m.def("torchaudio::sox_effects_shutdown_sox_effects", &shutdown_sox_effects);
+  m.def("torchaudio::sox_effects_apply_effects_tensor", &apply_effects_tensor);
+  m.def("torchaudio::sox_effects_apply_effects_file", &apply_effects_file);
 }
 } // namespace torchaudio::sox

torchaudio/csrc/sox/effects_chain.cpp

@@ -271,16 +271,16 @@ void SoxEffectsChain::addEffect(const std::vector<std::string> effect) {
   auto returned_effect = lsx().sox_find_effect(name.c_str());
   TORCH_CHECK(returned_effect, "Unsupported effect: ", name)
 
-    SoxEffect e(lsx().sox_create_effect(returned_effect));
+  SoxEffect e(lsx().sox_create_effect(returned_effect));
   const auto num_options = num_args - 1;
 
   std::vector<char*> opts;
   for (size_t i = 1; i < num_args; ++i) {
     opts.push_back((char*)effect[i].c_str());
   }
   TORCH_CHECK(
-      lsx().sox_effect_options(e, num_options, num_options ? opts.data() : nullptr) ==
-          SOX_SUCCESS,
+      lsx().sox_effect_options(
+          e, num_options, num_options ? opts.data() : nullptr) == SOX_SUCCESS,
       "Invalid effect option: ",
       c10::Join(" ", effect))
   TORCH_CHECK(

torchaudio/csrc/sox/io.cpp

@@ -67,8 +67,7 @@ c10::optional<std::tuple<torch::Tensor, int64_t>> load_audio_file(
     c10::optional<bool> channels_first,
     const c10::optional<std::string>& format) {
   auto effects = get_effects(frame_offset, num_frames);
-  return apply_effects_file(
-      path, effects, normalize, channels_first, format);
+  return apply_effects_file(path, effects, normalize, channels_first, format);
 }
 
 void save_audio_file(
@@ -132,12 +131,8 @@ void save_audio_file(
 
 TORCH_LIBRARY_FRAGMENT(torchaudio, m) {
   m.def("torchaudio::sox_io_get_info", &get_info_file);
-  m.def(
-      "torchaudio::sox_io_load_audio_file",
-      &load_audio_file);
-  m.def(
-      "torchaudio::sox_io_save_audio_file",
-      &save_audio_file);
+  m.def("torchaudio::sox_io_load_audio_file", &load_audio_file);
+  m.def("torchaudio::sox_io_save_audio_file", &save_audio_file);
 }
 
 } // namespace torchaudio::sox

torchaudio/csrc/sox/libsox.cpp

@@ -1,39 +1,59 @@
 #include <torchaudio/csrc/sox/libsox.h>
-
-#include <iostream>
+#include <c10/util/CallOnce.h>
 
 namespace torchaudio::sox {
-
-using Library = std::unique_ptr<at::DynamicLibrary>;
 namespace {
-static Library libsox;
+
+// Handle to the dlopen-ed libsox
+static std::unique_ptr<at::DynamicLibrary> libsox;
+// LSX class which torchaudio will be using
 static LSX _lsx;
+
+// dlopen libsox and populate mehotds on _lsx.
+void _init_lsx();
+
 } // namespace
-void init_libsox() {
+
+// Fetch lsx
+LSX& lsx() {
+  static c10::once_flag init_flag;
+  c10::call_once(init_flag, _init_lsx);
+  return _lsx;
+}
+
+namespace {
+
+// dlopen libsox and populate mehotds on _lsx.
+void _init_lsx() {
   libsox = []() {
 #if defined(_WIN32)
 #error Windows is not supported.
 #elif defined(__APPLE__)
-    auto lsx_ = std::make_unique<at::DynamicLibrary>("libsox.3.dylib", "libsox.dylib");
+    auto lsx_ =
+        std::make_unique<at::DynamicLibrary>("libsox.3.dylib", "libsox.dylib");
 #else
-    auto lsx_ = std::make_unique<at::DynamicLibrary>("libsox.3.so", "libsox.so");
+    auto lsx_ =
+        std::make_unique<at::DynamicLibrary>("libsox.3.so", "libsox.so");
 #endif
-
     // check version: we only support 14.4.2
     auto fn = (const char* (*)(void))lsx_->sym("sox_version");
-    std::cerr << fn() << std::endl;
-    TORCH_CHECK(strcmp(fn(), "14.4.2") == 0, "libsox library was found but not 14.4.2.");
+    TORCH_CHECK(
+        strcmp(fn(), "14.4.2") == 0,
+        "Need libsox 14.4.2, but found", fn());
     return lsx_;
   }();
 
-#define set_func(NAME) \
-  _lsx.NAME = (decltype(LSX::NAME))libsox->sym(#NAME)
+#define set_func(NAME) _lsx.NAME = (decltype(LSX::NAME))libsox->sym(#NAME)
 
+  // Note
+  // If any of the following fails, it will leave _lsx in invalid state.
+  // But _lsx cannot be accessed without this fuction succeful, so it'okay.
   set_func(sox_add_effect);
   set_func(sox_close);
   set_func(sox_create_effect);
   set_func(sox_create_effects_chain);
   set_func(sox_delete_effect);
+  set_func(sox_delete_effects_chain);
   set_func(sox_effect_options);
   set_func(sox_find_effect);
   set_func(sox_flow_effects);
@@ -46,10 +66,8 @@ void init_libsox() {
   set_func(sox_quit);
   set_func(sox_strerror);
   set_func(sox_write);
-
-  #undef set_func
+#undef set_func
 }
 
-LSX& lsx() { return _lsx; }
-
-}  // namespace torchaudio::sox
+} // namespace
+} // namespace torchaudio::sox

torchaudio/csrc/sox/libsox.h

@@ -1,7 +1,7 @@
 #pragma once
-#include <sox.h>
 #include <ATen/DynamicLibrary.h>
 #include <memory.h>
+#include <sox.h>
 
 namespace torchaudio::sox {
 
@@ -11,7 +11,7 @@ struct LSX {
       sox_effect_t* effp,
       sox_signalinfo_t* in,
       sox_signalinfo_t const* out) = nullptr;
-  int (*sox_close) (sox_format_t* ft) = nullptr;
+  int (*sox_close)(sox_format_t* ft) = nullptr;
 
   sox_effect_t* (*sox_create_effect)(sox_effect_handler_t const* eh) = nullptr;
 
@@ -20,8 +20,8 @@ struct LSX {
       sox_encodinginfo_t const* out_enc) = nullptr;
 
   void (*sox_delete_effect)(sox_effect_t* effp);
-  void (*sox_delete_effects_chain)(sox_effects_chain_t *ecp);
-  
+  void (*sox_delete_effects_chain)(sox_effects_chain_t* ecp);
+
   int (*sox_effect_options)(sox_effect_t* effp, int argc, char* const argv[]);
 
   const sox_effect_handler_t* (*sox_find_effect)(char const* name);
@@ -59,7 +59,6 @@ struct LSX {
   size_t (*sox_write)(sox_format_t* ft, const sox_sample_t* buf, size_t len);
 };
 
-void init_libsox();
 LSX& lsx();
 
 } // namespace torchaudio::sox

torchaudio/csrc/sox/pybind/pybind.cpp

@@ -7,22 +7,13 @@ namespace sox {
 namespace {
 
 PYBIND11_MODULE(_torchaudio_sox, m) {
-  m.def("init_libsox", &torchaudio::sox::init_libsox, "Initialize libsox.");
   m.def("set_seed", &torchaudio::sox::set_seed, "Set random seed.");
+  m.def("set_verbosity", &torchaudio::sox::set_verbosity, "Set verbosity.");
+  m.def("set_use_threads", &torchaudio::sox::set_use_threads, "Set threading.");
   m.def(
-      "set_verbosity", &torchaudio::sox::set_verbosity, "Set verbosity.");
+      "set_buffer_size", &torchaudio::sox::set_buffer_size, "Set buffer size.");
   m.def(
-      "set_use_threads",
-      &torchaudio::sox::set_use_threads,
-      "Set threading.");
-  m.def(
-      "set_buffer_size",
-      &torchaudio::sox::set_buffer_size,
-      "Set buffer size.");
-  m.def(
-      "get_buffer_size",
-      &torchaudio::sox::get_buffer_size,
-      "Get buffer size.");
+      "get_buffer_size", &torchaudio::sox::get_buffer_size, "Get buffer size.");
   m.def(
       "list_effects",
       &torchaudio::sox::list_effects,

torchaudio/csrc/sox/utils.cpp

@@ -1,7 +1,7 @@
 #include <c10/core/ScalarType.h>
 #include <sox.h>
-#include <torchaudio/csrc/sox/types.h>
 #include <torchaudio/csrc/sox/libsox.h>
+#include <torchaudio/csrc/sox/types.h>
 #include <torchaudio/csrc/sox/utils.h>
 
 namespace torchaudio::sox {
@@ -44,7 +44,8 @@ std::vector<std::vector<std::string>> list_effects() {
 
 std::vector<std::string> list_write_formats() {
   std::vector<std::string> formats;
-  for (const sox_format_tab_t* fns = lsx().sox_get_format_fns(); fns->fn; ++fns) {
+  for (const sox_format_tab_t* fns = lsx().sox_get_format_fns(); fns->fn;
+       ++fns) {
     const sox_format_handler_t* handler = fns->fn();
     for (const char* const* names = handler->names; *names; ++names) {
       if (!strchr(*names, '/') && handler->write)
@@ -56,7 +57,8 @@ std::vector<std::string> list_write_formats() {
 
 std::vector<std::string> list_read_formats() {
   std::vector<std::string> formats;
-  for (const sox_format_tab_t* fns = lsx().sox_get_format_fns(); fns->fn; ++fns) {
+  for (const sox_format_tab_t* fns = lsx().sox_get_format_fns(); fns->fn;
+       ++fns) {
     const sox_format_handler_t* handler = fns->fn();
     for (const char* const* names = handler->names; *names; ++names) {
       if (!strchr(*names, '/') && handler->read)