fixed some usage edge cases with NCD files

Author: Teque5

sigmf/convert/wav.py

@@ -105,6 +105,7 @@ def wav_to_sigmf(
         If the wav file cannot be read.
     """
     wav_path = Path(wav_path)
+    out_path = None if out_path is None else Path(out_path)
 
     # auto-enable NCD when no output path is specified
     if out_path is None:
@@ -161,13 +162,21 @@ def wav_to_sigmf(
         global_info[SigMFFile.DATASET_KEY] = wav_path.name
         capture_info[SigMFFile.HEADER_BYTES_KEY] = header_bytes
 
-    # handle NCD case where no output files are created
-    if create_ncd and out_path is None:
+    if create_ncd:
         # create metadata-only SigMF for NCD pointing to original file
         meta = SigMFFile(global_info=global_info)
         meta.set_data_file(data_file=wav_path, offset=header_bytes)
         meta.data_buffer = io.BytesIO()
         meta.add_capture(0, metadata=capture_info)
+
+        # write metadata file if output path specified
+        if out_path is not None:
+            filenames = get_sigmf_filenames(out_path)
+            output_dir = filenames["meta_fn"].parent
+            output_dir.mkdir(parents=True, exist_ok=True)
+            meta.tofile(filenames["meta_fn"], toarchive=False)
+            log.info("wrote %s", filenames["meta_fn"])
+
         log.debug("created NCD SigMF: %r", meta)
         return meta
 

sigmf/sigmffile.py

@@ -216,7 +216,8 @@ def __init__(
         if global_info is not None:
             self.set_global_info(global_info)
         if data_file is not None:
-            self.set_data_file(data_file, skip_checksum=skip_checksum, map_readonly=map_readonly)
+            offset = self._get_ncd_offset()
+            self.set_data_file(data_file, skip_checksum=skip_checksum, map_readonly=map_readonly, offset=offset)
 
     def __len__(self):
         return self._memmap.shape[0]
@@ -392,6 +393,30 @@ def _is_conforming_dataset(self):
         # if we get here, the file exists and is conforming
         return True
 
+    def _get_ncd_offset(self):
+        """
+        Detect Non-Conforming Dataset files and return the appropriate header offset.
+
+        For NCD files that reference external non-SigMF files (e.g., WAV), the
+        core:header_bytes field indicates how many bytes to skip to reach the
+        actual sample data.
+
+        Returns
+        -------
+        int
+            Byte offset to apply when reading the dataset file. 0 for conforming datasets.
+        """
+        if self._is_conforming_dataset():
+            return 0
+
+        # check if this is an NCD with core:dataset and header_bytes
+        captures = self.get_captures()
+        dataset_field = self.get_global_field(self.DATASET_KEY)
+        if dataset_field and captures and self.HEADER_BYTES_KEY in captures[0]:
+            return captures[0][self.HEADER_BYTES_KEY]
+
+        return 0
+
     def get_schema(self):
         """
         Return a schema object valid for the current metadata