Etienne Ackermann

coerce ASA computations to scalars when possible

3

I have already done this with mean, but it needs to be done for std, max, min, ...? ```python try: means = np.mean(self._ydata, axis=axis).squeeze() if means.size == 1: return np.asscalar(means)...

Implement generic irregular smoothing

4

Savva has been working on this for SpikeTrainArrays, but I'm parking it here as a generally useful feature that we should support as generically and as efficiently as possible. Ideally,...

Confusing "'fs' has been deprecated; use 'step' instead" warning when creating ASA

6

This happens because ASA calls get_contiguous_segments using both step and fs as arguments. We need to sort out which one to actually use, and update ASA and get_contiguous_segments implementations. My...

Implement core object mean, min, max, etc. to SMEVA pattern

2

I'm parking this issue to discuss an example use case for a SMEVA, and to remind us to sanitize all the core object arithmetic ops like mean, median, min, max,...

multidimensional slice indexing

3

Basic idea is now in place, but implementation can still be tricky... Will implement for AnalogSignalArray first...

Remove requested interpolation points that fall outside a signal's support in asarray

1

Currently `asarray`, `get_interpobj`, and `simplify` all work as expected, with the exception of `asarray`, which does not yet remove interpolation points outside of a signal's support. This is not a...

Interpolation can consume too much memory: chunk it

1

This is especially true for many-channel wideband data. We should ideally (1) parallelize over individual signals, but also consider (2) chunking the interpolation, so that our memory footprint stays reasonable.

Add EventArray indexing to AnalogSignalArray

9

This should facilitate trajectory operations, plotting AnalogSignals with different sampling rates, etc. Careful: how should interpolation work with support epochs? Careful: should an interp object be LAZILY computed and then...

filtering error message when some epochs are too short is noninformative

8

It is unclear if we should even raise any errors here, or when some epoch is too short, if we should simply leave it as is, and perhaps raise a...

relax unit_ids to not have to be integer?

6

Alternatively, we should re-think how to showcase our raster flexibility against that of matplotlib's eventplot()