Gating needs to do a proper job of handling timestamps associated with each frame

[jmtayloruk]

The code is not doing a great job of precise timing. It determines a delay time before sending the trigger, but then executes a bunch more code.

Oh and, more importantly, that delay time is then treated relative to “current_time”, which is set after doing the phase-matching. That is going to reduce accuracy and precision, and also makes me even more uncomfortable in terms of future-proofing.

I think it would be much better to pass around absolute times, not deltas.

Chas writes:
There was a reason, three years ago, to use relative times... but it could just have been because that was of interest to me in my offline analyses?

Originally posted by @ChasNelson1990 in https://github.com/_render_node/MDE3OlB1bGxSZXF1ZXN0UmV2aWV3NDY4MjcxOTQw/pull_request_reviews/more_threads

Jonny writes: I suspect any emulation concern you might have will be resolved if we pass around proper timestamp metadata with each frame

[jmtayloruk]

So: I believe the solution is to associated timestamp metadata with our frame data, and use those timestamps throughout the trigger prediction code. I will try and tackle that as part of the refactor.

[ChasNelson1990]

So, for prosperity, are you still trying to decide between a np.ndarray subclasses and something like h5py and hdf5 file sinks?

[jmtayloruk]

Yes, I have not given this any more thought since we spoke, and don't feel any option is yet jumping out as self-evidently the best solution. Was going to have a bit of a read back through what we actually do with the frame objects, and think about it some more.

[jmtayloruk]

Yes, I have not given this any more thought since we spoke, and don't feel any option is yet jumping out as self-evidently the best solution. Was going to have a bit of a read back through what we actually do with the frame objects, and think about it some more.

[jmtayloruk]

Yes, I have not given this any more thought since we spoke, and don't feel any option is yet jumping out as self-evidently the best solution. Was going to have a bit of a read back through what we actually do with the frame objects, and think about it some more.

[jmtayloruk]

Yes, I have not given this any more thought since we spoke, and don't feel any option is yet jumping out as self-evidently the best solution. Was going to have a bit of a read back through what we actually do with the frame objects, and think about it some more.