Add GSM encoding support to info (#1277)

test/torchaudio_unittest/backend/sox_io/info_test.py

@@ -221,6 +221,21 @@ class TestInfo(TempDirMixin, PytorchTestCase):
         assert info.bits_per_sample == 8
         assert info.encoding == "ALAW"
 
+    def test_gsm(self):
+        """`sox_io_backend.info` can check gsm file correctly"""
+        duration = 1
+        num_channels = 1
+        sample_rate = 8000
+        path = self.get_temp_path('data.gsm')
+        sox_utils.gen_audio_file(
+            path, sample_rate=sample_rate, num_channels=num_channels,
+            duration=duration)
+        info = sox_io_backend.info(path)
+        assert info.sample_rate == sample_rate
+        assert info.num_channels == num_channels
+        assert info.bits_per_sample == 0
+        assert info.encoding == "GSM"
+
     def test_htk(self):
         """`sox_io_backend.info` can check HTK file correctly"""
         duration = 1

torchaudio/csrc/sox/io.cpp

-#include <sox.h>
 #include <torchaudio/csrc/sox/effects.h>
 #include <torchaudio/csrc/sox/effects_chain.h>
 #include <torchaudio/csrc/sox/io.h>
+#include <torchaudio/csrc/sox/types.h>
 #include <torchaudio/csrc/sox/utils.h>
 
 using namespace torch::indexing;
@@ -10,41 +10,6 @@ using namespace torchaudio::sox_utils;
 namespace torchaudio {
 namespace sox_io {
 
-namespace {
-
-std::string get_encoding(sox_encoding_t encoding) {
-  switch (encoding) {
-    case SOX_ENCODING_UNKNOWN:
-      return "UNKNOWN";
-    case SOX_ENCODING_SIGN2:
-      return "PCM_S";
-    case SOX_ENCODING_UNSIGNED:
-      return "PCM_U";
-    case SOX_ENCODING_FLOAT:
-      return "PCM_F";
-    case SOX_ENCODING_FLAC:
-      return "FLAC";
-    case SOX_ENCODING_ULAW:
-      return "ULAW";
-    case SOX_ENCODING_ALAW:
-      return "ALAW";
-    case SOX_ENCODING_MP3:
-      return "MP3";
-    case SOX_ENCODING_VORBIS:
-      return "VORBIS";
-    case SOX_ENCODING_AMR_WB:
-      return "AMR_WB";
-    case SOX_ENCODING_AMR_NB:
-      return "AMR_NB";
-    case SOX_ENCODING_OPUS:
-      return "OPUS";
-    default:
-      return "UNKNOWN";
-  }
-}
-
-} // namespace
-
 std::tuple<int64_t, int64_t, int64_t, int64_t, std::string> get_info_file(
     const std::string& path,
     c10::optional<std::string> format) {

torchaudio/csrc/sox/types.cpp

@@ -102,5 +102,38 @@ BitDepth get_bit_depth_from_option(const c10::optional<int64_t> bit_depth) {
   }
 }
 
+std::string get_encoding(sox_encoding_t encoding) {
+  switch (encoding) {
+    case SOX_ENCODING_UNKNOWN:
+      return "UNKNOWN";
+    case SOX_ENCODING_SIGN2:
+      return "PCM_S";
+    case SOX_ENCODING_UNSIGNED:
+      return "PCM_U";
+    case SOX_ENCODING_FLOAT:
+      return "PCM_F";
+    case SOX_ENCODING_FLAC:
+      return "FLAC";
+    case SOX_ENCODING_ULAW:
+      return "ULAW";
+    case SOX_ENCODING_ALAW:
+      return "ALAW";
+    case SOX_ENCODING_MP3:
+      return "MP3";
+    case SOX_ENCODING_VORBIS:
+      return "VORBIS";
+    case SOX_ENCODING_AMR_WB:
+      return "AMR_WB";
+    case SOX_ENCODING_AMR_NB:
+      return "AMR_NB";
+    case SOX_ENCODING_OPUS:
+      return "OPUS";
+    case SOX_ENCODING_GSM:
+      return "GSM";
+    default:
+      return "UNKNOWN";
+  }
+}
+
 } // namespace sox_utils
 } // namespace torchaudio

torchaudio/csrc/sox/types.h

 #ifndef TORCHAUDIO_SOX_TYPES_H
 #define TORCHAUDIO_SOX_TYPES_H
 
+#include <sox.h>
 #include <torch/script.h>
 
 namespace torchaudio {
@@ -51,6 +52,8 @@ enum class BitDepth : unsigned {
 
 BitDepth get_bit_depth_from_option(const c10::optional<int64_t> bit_depth);
 
+std::string get_encoding(sox_encoding_t encoding);
+
 } // namespace sox_utils
 } // namespace torchaudio