/**
- Created: 2014-08-16
- Author: Oscar Campbell
- Licence: MIT (Expat) - http://opensource.org/licenses/mit-license.html **/
2014-10-23 - There are bugs around: OHLC-candles, EMA(1) on tick basis and periodicalMean doesn't line up in time. They differ by 1 bar and 2 bars respectively. ALSO, making a selection of time to render gives unexpected results: Selecting 8:59... vs 09:00... gives completely different sets of data - not one minutes diff!!!??? WTF gives??
- ** PERHAPS - the date set (15:57) becomes the epochian date, even though it isn't.
2014-11-04 - Been re-structuring and testing a lot the last couple days. Time-stomping different periodizations, time-stones (mini epochs), ticks, etc is getting to where it should be. Needs to be verified even more with edge cases (however esoteric they may be). The one-off mystery (no more off-by-two's anyway) is still lurking.
2014-12-03 - The one-offs for periodizations is fixed. See git log instead. This file isn't maintained well.