According to docs, 'names' entry in the spike_features dictionary must be a list of strings. Most fet* methods conform to this. However, features.combine returns numpy array of names. Moreover, ClusterAnalyzer._cluster relies on the fact that feature_data['names'] is actually a numpy array and uses advanced indexing which doesn't work on lists. This didn't cause many problems before feature hiding functionality was implemented
Because numpy arrays are generally not very well suited for storing variable length objects, It maybe worth accepting a "list of strings" convention completely.
According to docs, 'names' entry in the spike_features dictionary must be a list of strings. Most fet* methods conform to this. However,
features.combinereturns numpy array of names. Moreover,ClusterAnalyzer._clusterrelies on the fact thatfeature_data['names']is actually a numpy array and uses advanced indexing which doesn't work on lists. This didn't cause many problems before feature hiding functionality was implementedBecause numpy arrays are generally not very well suited for storing variable length objects, It maybe worth accepting a "list of strings" convention completely.
@btel or was that made on purpose?