Finally I am moving forward with this
I have plenty of work still, with this RLCS Package idea.
One of them, indeed, is to make it into… A package.
I have been postponing this for quite some time, and well, I can only stall for so long.
Might actually be feasible
The thing is, I have been worried about how hard it would be. Well…
I use RStudio (side note: I hope it lives as product for still a long time…).
So I “just” created a “Package Project”, and then went through some of the chapters of the “R Packages” book once again this morning with a coffee.
Time has come, and so… What I needed to actually get it moving was to copy my current, non-package, functional code from my (usual) RLibs/ folder to the new project’s R/ folder…
Then I deleted some temporary code that shouldn’t have been there…
Then I know and accept this is far from finished and I have plenty more to do still but…
Long story short, I then made sure I had a clean environment, ran “devtools::load_all()” and ran a function I knew was stand-alone and should work (actually I “fudged” even the new project name, but this is just a first test :D):
Basically, it works, with “kinks”
So it looks like it “works” almost as-is:
However the above has for now a couple of issues at least:
That’s the version WITHOUT RCpp faster version of match function
And for some reason, I have to manually re-load the functions that set the print and plot for rules and population of RLCS, which… Well I don’t know yet why :D
But as-is, I can actually use this “as a package” to do supervised learning and it would in fact work, essentially.
Conclusions
It’s very early in that part of the process.
But while reviewing the book this morning, I realized I had actually been fairly organized following my own rules, and minimal changes would get me somewhere.
Indeed, I have a lot more to do. For instance:
what if I do not want to expose a function, and make it available only internally? Not sure yet about how to go about that.
Documentation, roxygen2? I’m not too familiar yet. Vignette? Sounds like a nice idea but, not there yet either.
RCpp? I have just the one function that did make a difference in terms of processing speed, and it makes sense it is kept as part of the package, so I need to make sure I include it, although I’m unclear on exactly how…
So yes, plenty yet to do. But it looks quite… Feasible? That’s good news.