Closed milktrader closed 10 years ago
Okay, we're up to 72 tests now. https://github.com/JuliaStats/TimeSeries.jl/pull/69
Updated the file list to reflect the new organization
Closing this because it's too umbrella-ish (of course, it did admit to that when it opened)
@karbarcca @carljv
Ye Olde Umbrella Issue ...
Probably the best workflow is to create a branch calling it something along the lines
timearrayrefactor
and focus on that group of methods.There are currently 65 tests that cover much of the functionality, but certainly more can be written. The mathematical operators tests for example are not exhaustive, and probably should be. I made the comparison operators exhaustive and discovered that I neglected a
.<=
method instead calling it a<=
method.I feel like the number of methods in the package is complete (covers 99% of time series related functionality) and fulfills the goal of being a lightweight package. Probably those in utilities can be removed (deprecated).
Ideas for other must-have methods?
Some notes about the testing framework
The testing framework uses the registered FactCheck package, and
const
objects from the un-registered MarketData package.Wait a minute, I thought you said FactCheck is registered? Well it is, but the package sha1 hasn't been bumped to METADATA for a few weeks, and the
@runtest
macro isn't available if the package is installed viaPkg.add()
. More on this usage below.I'm going to rewrite the MarketTechnicals package to support the current implementation of TimeSeries, and then go on to writing a backtesting framework. There is also work to be done in TimeModels, if anyone is interested.