Started tests on real data.

dev/file_format/experiment.ipynb

@@ -56,7 +56,7 @@
       "prm = {\n",
       "       'nchannels': nchannels,\n",
       "       'nfeatures': nchannels*3,\n",
-      "       'nwavesamples': 20\n",
+      "       'waveforms_nsamples': 20\n",
       "       }\n",
       "prb = {'channel_groups': [\n",
       "            {\n",
@@ -316,7 +316,7 @@
      "cell_type": "code",
      "collapsed": false,
      "input": [
-      "exp.application_data.spikedetekt.nwavesamples"
+      "exp.application_data.spikedetekt.waveforms_nsamples"
      ],
      "language": "python",
      "metadata": {},

spikedetekt2/core/main.py

@@ -120,8 +120,7 @@ def run(raw_data=None, experiment=None, prm=None, probe=None):
     for chunk in raw_data.chunks(chunk_size=chunk_size, 
                                  chunk_overlap=chunk_overlap,):
         # Filter the (full) chunk.
-        # shape: (nsamples, nchannels)
-        chunk_raw = chunk.data_chunk_full
+        chunk_raw = chunk.data_chunk_full  # shape: (nsamples, nchannels)
         chunk_fil = apply_filter(chunk_raw, filter=filter)
 
         # Apply thresholds.

spikedetekt2/core/tests/test_main.py

@@ -22,17 +22,13 @@
 
 sample_rate = 20000
 duration = 1.
-nwavesamples = 10
 nchannels = 2
-chunk_size = 20000
 nsamples = int(sample_rate * duration)
 raw_data = .1 * np.random.randn(nsamples, nchannels)
 
 prm = get_params(**{
-    'nwavesamples': nwavesamples, 
     'nchannels': nchannels,
     'sample_rate': sample_rate,
-    'chunk_size': chunk_size,
     'detect_spikes': 'positive',
 })
 prb = {'channel_groups': [

spikedetekt2/dataio/experiment.py

@@ -57,6 +57,9 @@ def _get_child_id(child):
 def _print_instance(obj, depth=0, name=''):
     # Handle the first element of the list/dict.
     if isinstance(obj, (list, dict)):
+        if not obj:
+            r = []
+            return r
         if isinstance(obj, list):
             sobj = obj[0]
             key = '0'
@@ -293,7 +296,6 @@ def add(self, time_samples=None, time_fractional=0,
         self.cluster.append((cluster,))
         self.cluster_original.append((cluster_original,))
         self.features_masks.append(features_masks)
-
         self.waveforms_raw.append(waveforms_raw)
         self.waveforms_filtered.append(waveforms_filtered)
 

spikedetekt2/dataio/kwik.py

@@ -171,7 +171,7 @@ def create_kwx(path, prb=None, prm=None, has_masks=True):
     
     Arguments:
       * prb: the PRB dictionary
-      * nwavesamples (common to all channel groups if set)
+      * waveforms_nsamples (common to all channel groups if set)
       * nfeatures (total number of features per spike, common to all channel groups if set)
       * nchannels (number of channels per channel group, common to all channel groups if set)
     
@@ -185,20 +185,20 @@ def create_kwx(path, prb=None, prm=None, has_masks=True):
     nchannels = prm.get('nchannels', None)
     nfeatures_per_channel = prm.get('nfeatures_per_channel', None)
     nfeatures = prm.get('nfeatures', None)
-    nwavesamples = prm.get('nwavesamples', None)
+    waveforms_nsamples = prm.get('waveforms_nsamples', None)
 
     file = tb.openFile(path, mode='w')
     file.createGroup('/', 'channel_groups')
 
     for ichannel_group, chgrp_info in enumerate(prb.get('channel_groups', [])):
         nchannels_ = len(chgrp_info.get('channels', [])) or nchannels or 0
-        nwavesamples_ = chgrp_info.get('nwavesamples', nwavesamples) or 0
+        waveforms_nsamples_ = chgrp_info.get('waveforms_nsamples', waveforms_nsamples) or 0
         nfeatures_per_channel_ = chgrp_info.get('nfeatures_per_channel', nfeatures_per_channel) or 0
         nfeatures_ = chgrp_info.get('nfeatures', nfeatures) or nfeatures_per_channel_ * nchannels_
 
         assert nchannels_ > 0
         assert nfeatures_ > 0
-        assert nwavesamples_ > 0
+        assert waveforms_nsamples_ > 0
 
         channel_group_path = '/channel_groups/{0:d}'.format(ichannel_group)
 
@@ -216,9 +216,9 @@ def create_kwx(path, prb=None, prm=None, has_masks=True):
                               tb.Float32Atom(), (0, nfeatures_))
 
         file.createEArray(channel_group_path, 'waveforms_raw',
-                          tb.Int16Atom(), (0, nwavesamples_, nchannels_))
+                          tb.Int16Atom(), (0, waveforms_nsamples_, nchannels_))
         file.createEArray(channel_group_path, 'waveforms_filtered',
-                          tb.Int16Atom(), (0, nwavesamples_, nchannels_))
+                          tb.Int16Atom(), (0, waveforms_nsamples_, nchannels_))
 
     file.close()
 

spikedetekt2/dataio/tests/test_experiment.py

@@ -27,7 +27,7 @@
 
 def setup():
     # Create files.
-    prm = {'nfeatures': 3, 'nwavesamples': 10, 'nchannels': 3}
+    prm = {'nfeatures': 3, 'waveforms_nsamples': 10, 'nchannels': 3}
     prb = {'channel_groups': [
         {
             'channels': [4, 6, 8],
@@ -78,7 +78,7 @@ def test_experiment_channels():
         assert exp.application_data
         assert exp.user_data
         assert exp.application_data.spikedetekt.nchannels == 3
-        assert exp.application_data.spikedetekt.nwavesamples == 10
+        assert exp.application_data.spikedetekt.waveforms_nsamples == 10
         assert exp.application_data.spikedetekt.nfeatures == 3
 
         # Channel group.

spikedetekt2/dataio/tests/test_kwik.py

@@ -20,7 +20,7 @@
 DIRPATH = tempfile.mkdtemp()
 
 def setup_create():
-    prm = {'nfeatures': 3, 'nwavesamples': 10}
+    prm = {'nfeatures': 3, 'waveforms_nsamples': 10}
     prb = {'channel_groups': [
         {
             'channels': [4, 6, 8],
@@ -75,7 +75,7 @@ def test_create_kwik():
     path = os.path.join(DIRPATH, 'myexperiment.kwik')
 
     prm = {
-        'nwavesamples': 20,
+        'waveforms_nsamples': 20,
         'nfeatures': 3*32,
     }
     prb = {'channel_groups': [
@@ -99,12 +99,12 @@ def test_create_kwx():
     path = os.path.join(DIRPATH, 'myexperiment.kwx')
 
     # Create the KWX file.
-    nwavesamples = 20
+    waveforms_nsamples = 20
     nchannels = 32
     nchannels2 = 24
     nfeatures = 3*nchannels
     prm = {
-        'nwavesamples': 20,
+        'waveforms_nsamples': 20,
         'nfeatures': 3*nchannels,
     }
     prb = {'channel_groups': [
@@ -131,16 +131,16 @@ def test_create_kwx():
     wr1 = f.root.channel_groups.__getattr__('1').waveforms_raw
     wf1 = f.root.channel_groups.__getattr__('1').waveforms_filtered
     assert fm1.shape[1:] == (3*nchannels2, 2)
-    assert wr1.shape[1:] == (nwavesamples, nchannels2)
-    assert wf1.shape[1:] == (nwavesamples, nchannels2)
+    assert wr1.shape[1:] == (waveforms_nsamples, nchannels2)
+    assert wf1.shape[1:] == (waveforms_nsamples, nchannels2)
 
     # Group 2
     fm2 = f.root.channel_groups.__getattr__('2').features_masks
     wr2 = f.root.channel_groups.__getattr__('2').waveforms_raw
     wf2 = f.root.channel_groups.__getattr__('2').waveforms_filtered
     assert fm2.shape[1:] == (2*nchannels, 2)
-    assert wr2.shape[1:] == (nwavesamples, nchannels)
-    assert wf2.shape[1:] == (nwavesamples, nchannels)
+    assert wr2.shape[1:] == (waveforms_nsamples, nchannels)
+    assert wf2.shape[1:] == (waveforms_nsamples, nchannels)
 
     f.close()
 

spikedetekt2/utils/params_default.py

@@ -8,8 +8,8 @@
 
 # Chunks
 # ------
-chunk_size = int(sample_rate)
-chunk_overlap = int(.01 * sample_rate)
+chunk_size = int(1. * sample_rate)  # 1 second
+chunk_overlap = int(.01 * sample_rate)  # 10 ms
 
 # Spike detection
 # ---------------
@@ -27,8 +27,9 @@
 
 # Spike extraction
 # ----------------
-extract_s_before = 5
-extract_s_after = 5
+extract_s_before = 10
+extract_s_after = 10
+waveforms_nsamples = extract_s_before + extract_s_after
 
 # Features
 # --------