I've also been searching for a way to convert tick-level data (exported ASCII from Multicharts) to a Range, Renko and Heikin Ashi series that I'd be able to store once and leverage across a lot of back-testing scenarios in R since MC back-testing doesn't take advantage of the 8 cores available on my I7 (among other things). I've invested a lot of time in writing a lot of EasyLanguage over the years but figure it's going to be worth converting to R since I seem to be at an impasse with either NT or MC for testing. Long term, I'd like to be able make things a lot more parallel than they are today and distribute the workload across a cluster of 48-core boxes, but that's AFTER I get the code converted .
That being said, having spent a lot of time with Perl throughout my IT career, R internals appears at first glance very similar to Perl (from a data structure / C usage standpoint) so that has me thinking I should write a "harrr" (Heikin-Ashi, Renko, Range) package to create xts-like OHLC time series that can be either used in back-testing or passed to whatever TA charting tools R has to offer (quantmod will most likely suffice here once I have the data converted).
Given this thread was started 8+ months ago, has anyone seen this implemented yet or does it sound like something worth pursuing and uploading to CRAN?
Heiken Ashi is in my IKTrading package. Renko and other such "brick only when the price rises above a certain point" is something I've never done. Heiken Ashi requires a little Rcpp knowledge, however. R's for loops are way too slow to compute it.
The following user says Thank You to IlyaKipnis for this post:
Thanks Ilya - I'll take a look there for HA. I'm also poking around with the xts package ("to.period()" stuff) which looks to also use a lot of Rcpp for performance. I've read a bunch already about R's performance (or lack thereof) for loops, but fortunately I know C/C++ and am just learning how Rcpp makes use of them (hopefully won't take too long).