Use dlopen

third_party/sox/CMakeLists.txt

@@ -1,7 +1,7 @@
 include(FetchContent)
 
 FetchContent_Declare(
-  sox_src
+  sox
   URL https://downloads.sourceforge.net/project/sox/sox/14.4.2/sox-14.4.2.tar.bz2
   URL_HASH SHA256=81a6956d4330e75b5827316e44ae381e6f1e8928003c6aa45896da9041ea149c
   PATCH_COMMAND ""
@@ -10,12 +10,9 @@ FetchContent_Declare(
   )
 # FetchContent_MakeAvailable will parse the downloaded content and setup the targets.
 # We want to only download and not build, so we run Populate manually.
-if(NOT sox_src_POPULATED)
-  FetchContent_Populate(sox_src)
+if(NOT sox_POPULATED)
+  FetchContent_Populate(sox)
 endif()
 
-add_library(sox SHARED stub.c)
-if(APPLE)
-  set_target_properties(sox PROPERTIES SUFFIX .dylib)
-endif(APPLE)
-target_include_directories(sox PUBLIC ${sox_src_SOURCE_DIR}/src)
+add_library(libsox INTERFACE)
+target_include_directories(libsox INTERFACE ${sox_SOURCE_DIR}/src)

tools/setup_helpers/extension.py

@@ -128,6 +128,7 @@ def build_extension(self, ext):
             "-DCMAKE_VERBOSE_MAKEFILE=ON",
             f"-DPython_INCLUDE_DIR={distutils.sysconfig.get_python_inc()}",
             f"-DBUILD_SOX:BOOL={'ON' if _BUILD_SOX else 'OFF'}",
+            "-DDLOPEN_SOX:BOOL=ON",
             f"-DBUILD_RIR:BOOL={'ON' if _BUILD_RIR else 'OFF'}",
             f"-DBUILD_RNNT:BOOL={'ON' if _BUILD_RNNT else 'OFF'}",
             f"-DBUILD_ALIGN:BOOL={'ON' if _BUILD_ALIGN else 'OFF'}",

torchaudio/_extension/__init__.py

@@ -2,7 +2,7 @@
 import os
 import sys
 
-from torchaudio._internal.module_utils import fail_with_message, is_module_available, no_op
+from torchaudio._internal.module_utils import eval_env, fail_with_message, is_module_available, no_op
 
 try:
     from .fb import _init_ffmpeg
@@ -51,23 +51,15 @@
     _IS_ALIGN_AVAILABLE = torchaudio.lib._torchaudio.is_align_available()
 
 
-# Similar to libtorchaudio, sox-related features should be importable when present.
-#
-# Note: This will be change in the future when sox is dynamically linked.
-# At that point, this initialization should handle the case where
-# sox integration is built but libsox is not found.
+# Initialize libsox-related features
+_SOX_ENABLED = eval_env("TORCHAUDIO_USE_SOX", True)
 _SOX_INITIALIZED = False
-if is_module_available("torchaudio.lib._torchaudio_sox"):
+if _SOX_ENABLED and is_module_available("torchaudio.lib._torchaudio_sox"):
     try:
         _init_sox()
         _SOX_INITIALIZED = True
     except Exception:
-        # The initialization of sox extension will fail if supported sox
-        # libraries are not found in the system.
-        # Since the rest of the torchaudio works without it, we do not report the
-        # error here.
-        # The error will be raised when user code attempts to use these features.
-        _LG.debug("Failed to initialize sox extension", exc_info=True)
+        _LG.debug("Failed to initialize libsox bindings", exc_info=True)
 
 
 # Initialize FFmpeg-related features
@@ -83,8 +75,13 @@
         # The error will be raised when user code attempts to use these features.
         _LG.debug("Failed to initialize ffmpeg bindings", exc_info=True)
 
+if _SOX_INITIALIZED:
+    fail_if_no_sox = no_op
+elif not _SOX_ENABLED:
+    fail_if_no_sox = fail_with_message("requires sox extension, but it is disabled. (TORCHAUDIO_USE_SOX=0)")
+else:
+    fail_if_no_sox = _fail_since_no_sox
 
-fail_if_no_sox = no_op if _SOX_INITIALIZED else _fail_since_no_sox
 fail_if_no_ffmpeg = _fail_since_no_ffmpeg if _FFMPEG_EXT is None else no_op
 
 fail_if_no_rir = (

torchaudio/_extension/utils.py

@@ -70,13 +70,6 @@ def _init_sox():
     _load_lib("libtorchaudio_sox")
     import torchaudio.lib._torchaudio_sox  # noqa
 
-    torchaudio.lib._torchaudio_sox.set_verbosity(0)
-
-    import atexit
-
-    torch.ops.torchaudio.sox_effects_initialize_sox_effects()
-    atexit.register(torch.ops.torchaudio.sox_effects_shutdown_sox_effects)
-
 
 def _try_access_avutil(ffmpeg_ver):
     libname_template = {
@@ -202,12 +195,12 @@ def _fail_since_no_sox(func):
     def wrapped(*_args, **_kwargs):
         try:
             # Note:
-            # We run _init_ffmpeg again just to show users the stacktrace.
+            # We run _init_sox again just to show users the stacktrace.
             # _init_ffmpeg would not succeed here.
             _init_sox()
         except Exception as err:
             raise RuntimeError(
-                f"{func.__name__} requires sox extension which is not available. "
+                f"{func.__name__} requires libsox extension which is not available. "
                 "Please refer to the stacktrace above for how to resolve this."
             ) from err
         # This should not happen in normal execution, but just in case.

torchaudio/csrc/sox/CMakeLists.txt

@@ -1,17 +1,23 @@
 set(
   sources
+  stub.cpp
   io.cpp
   utils.cpp
   effects.cpp
   effects_chain.cpp
   types.cpp
   )
+set(
+  compiler_definitions
+  DLOPEN_SOX
+  )
+
 torchaudio_library(
   libtorchaudio_sox
   "${sources}"
   ""
-  "torch;sox"
-  ""
+  "torch;libsox"
+  "${compiler_definitions}"
   )
 
 if (BUILD_TORCHAUDIO_PYTHON_EXTENSION)
@@ -20,6 +26,6 @@ if (BUILD_TORCHAUDIO_PYTHON_EXTENSION)
     "pybind/pybind.cpp;"
     ""
     "libtorchaudio_sox"
-    ""
+    "${compiler_definitions}"
     )
 endif()

torchaudio/csrc/sox/effects.cpp

@@ -1,49 +1,10 @@
 #include <sox.h>
 #include <torchaudio/csrc/sox/effects.h>
 #include <torchaudio/csrc/sox/effects_chain.h>
+#include <torchaudio/csrc/sox/stub.h>
 #include <torchaudio/csrc/sox/utils.h>
 
 namespace torchaudio::sox {
-namespace {
-
-enum SoxEffectsResourceState { NotInitialized, Initialized, ShutDown };
-SoxEffectsResourceState SOX_RESOURCE_STATE = NotInitialized;
-std::mutex SOX_RESOUCE_STATE_MUTEX;
-
-} // namespace
-
-void initialize_sox_effects() {
-  const std::lock_guard<std::mutex> lock(SOX_RESOUCE_STATE_MUTEX);
-
-  switch (SOX_RESOURCE_STATE) {
-    case NotInitialized:
-      TORCH_CHECK(
-          sox_init() == SOX_SUCCESS, "Failed to initialize sox effects.");
-      SOX_RESOURCE_STATE = Initialized;
-      break;
-    case Initialized:
-      break;
-    case ShutDown:
-      TORCH_CHECK(
-          false, "SoX Effects has been shut down. Cannot initialize again.");
-  }
-};
-
-void shutdown_sox_effects() {
-  const std::lock_guard<std::mutex> lock(SOX_RESOUCE_STATE_MUTEX);
-
-  switch (SOX_RESOURCE_STATE) {
-    case NotInitialized:
-      TORCH_CHECK(false, "SoX Effects is not initialized. Cannot shutdown.");
-    case Initialized:
-      TORCH_CHECK(
-          sox_quit() == SOX_SUCCESS, "Failed to initialize sox effects.");
-      SOX_RESOURCE_STATE = ShutDown;
-      break;
-    case ShutDown:
-      break;
-  }
-}
 
 auto apply_effects_tensor(
     torch::Tensor waveform,
@@ -91,7 +52,7 @@ auto apply_effects_file(
     const c10::optional<std::string>& format)
     -> c10::optional<std::tuple<torch::Tensor, int64_t>> {
   // Open input file
-  SoxFormat sf(sox_open_read(
+  SoxFormat sf(SOX sox_open_read(
       path.c_str(),
       /*signal=*/nullptr,
       /*encoding=*/nullptr,
@@ -129,21 +90,15 @@ auto apply_effects_file(
       dtype,
       normalize.value_or(true),
       channels_first_);
-
   return std::tuple<torch::Tensor, int64_t>(
       tensor, chain.getOutputSampleRate());
 }
 
 namespace {
 
 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);
 }
-
 } // namespace
 } // namespace torchaudio::sox