Keep invalid samples invalid in smooth_frames

smooth_frames averaged the invalid sample value with the other samples
in a frame, which gave a large valid-looking value instead of an invalid
one (NaN in physical units). As in the WFDB library's meansamp(), a
frame that contains an invalid sample now gives the invalid value.

Author: TheNetherWatcher

docs/changes.rst

@@ -41,6 +41,9 @@ Unreleased
 **Fix resample_sig for signals starting with NaN**
   ``resample_sig`` (and the functions that use it, such as ``resample_singlechan`` and ``resample_multichan``) returned only NaN values if the signal started with NaN. Leading NaN values now take the first valid value before resampling, as trailing ones already took the last.
 
+**Fix smoothing of frames with invalid samples**
+  When a signal with more than one sample per frame is read with ``smooth_frames=True`` (the default), a frame that contains an invalid sample now gives an invalid value (NaN in physical units), as in the WFDB library. The invalid sample value used to be averaged with the other samples, which gave a large but valid-looking value.
+
 Version 4.3.1 (Feb 2026)
 -----------------------------
 

src/wfdb/io/_signal.py

@@ -1060,15 +1060,20 @@ def smooth_frames(self, sigtype="physical"):
             if spf[ch] == 1:
                 signal[:, ch] = expanded_signal[ch]
             else:
+                # As in the WFDB library, a frame with an invalid sample gives
+                # an invalid value. Physical samples are already NaN.
+                invalid = None
+                if sigtype == "digital" and self.fmt and self.fmt[ch]:
+                    invalid = _digi_nan(self.fmt[ch])
                 frames = expanded_signal[ch].reshape(-1, spf[ch])
                 for chunk_start in range(0, sig_len, CHUNK_SIZE):
                     chunk_end = chunk_start + CHUNK_SIZE
-                    signal_sum = np.sum(
-                        frames[chunk_start:chunk_end, :],
-                        axis=1,
-                        dtype=intermediate_dtype,
-                    )
+                    chunk = frames[chunk_start:chunk_end, :]
+                    signal_sum = np.sum(chunk, axis=1, dtype=intermediate_dtype)
                     signal[chunk_start:chunk_end, ch] = signal_sum / spf[ch]
+                    if invalid is not None:
+                        has_invalid = np.any(chunk == invalid, axis=1)
+                        signal[chunk_start:chunk_end, ch][has_invalid] = invalid
 
         return signal
 

tests/test_record.py

@@ -857,6 +857,22 @@ def test_write_smoothed(self, temp_path):
         )
         np.testing.assert_array_equal(record.d_signal, record2.d_signal)
 
+    def test_smooth_frames_invalid_sample(self, temp_path):
+        """A frame with an invalid sample is invalid after smoothing."""
+        # Two samples per frame. -32768 is the invalid sample value for fmt 16.
+        raw = np.array([2, 4, -32768, 100, 6, 8], dtype="<i2")
+        raw.tofile(os.path.join(temp_path, "invalid.dat"))
+        with open(os.path.join(temp_path, "invalid.hea"), "w") as f:
+            f.write("invalid 1 100 3\n")
+            f.write("invalid.dat 16x2 200/mV 16 0 0 0 0 sig\n")
+        record_name = os.path.join(temp_path, "invalid")
+
+        record = wfdb.rdrecord(record_name, physical=False)
+        np.testing.assert_array_equal(record.d_signal[:, 0], [3, -32768, 7])
+
+        record = wfdb.rdrecord(record_name)
+        np.testing.assert_allclose(record.p_signal[:, 0], [0.015, np.nan, 0.035])
+
     def test_to_dataframe(self):
         record = wfdb.rdrecord("sample-data/test01_00s")
         df = record.to_dataframe()