Closed brownworth closed 6 years ago
Right now, the data is 3 columns, ostensibly indexed by datestamp. To me, it makes more sense to be indexed by minute (or hour) and the columns are each individual computer, with the intersection being the state of that machine at that point in time.
8b991e44b5aa6a2a81d3c4a3e97b769702aa04e5 for my attempt at this.
I didn't see that you had something like this in place, and I have done something similar. We can compare code to see which suits our needs.
As an update I just pushed a change that improves the speed of what i had before immensely.
That's in efa5140
After looking at this yet further, I found some potential issues with the implementation currently in develop
. I revisited my alternate implementation and managed to get the runtime down quite a bit. I've pointed out some discrepancies in the results between the two methods in the nick-data-import-issues
branch.
Already merged.
Timestamps are datetime with ms precision, this needs to be expanded to minutes to be able to do summation stats.