Programming puzzles: processing lists!

[DavidM]

(05-31-2017 04:02 PM)pier4r Wrote:  Inspired by the commands of DavidM (that still I have to compare to GoFerList or to use for solutions, they are meant for it) I added some more challenges in the first post.

I'm still in the process of doing some performance testing on GoferList commands as compared to similar commands from my in-progress library, so it may take a few days before I can start to dig into your newest challenges.

I can share some preliminary findings of the performance testing, with the promise of more details to come:

Commands that are functionally equivalent are mostly on par with each other, though almost all of my library commands have a very slight speed advantage over their GoferList counterparts. Commands in this category include Copy/LMRPT, Repeat/LNDUP, Split/LSPLT, Chars/S->SL, and Strcat/LSUM.

There are a couple of exceptions to that rule, though:

GoferList's Nub performed better on larger lists than LDDUP, at least on my test data (which was intentionally chosen to have 50% duplicates). The reverse was true for lists of 0-100 elements, where LDDUP had the advantage. At 1000 elements, Nub averaged 78.4 seconds vs. LDDUP at 135.7 seconds. Since LDDUP is essentially a wrapper around a ROM routine, I'm thinking it might be worth coming up with my own version for comparison.

GoferList's Copy is orders of magnitude slower than LMRPT, and the performance degradation is stark as the quantity of element duplication grows. When duplicating a list of 1 item 1000 times (the worst case in my tests), Copy required 528.8 seconds. LMRPT completed the same task in 0.11 seconds.

I'm still in the process of comparing some others, but it's taking a bit more time because the commands aren't exactly the same. To get a more complete picture, I do two separate test suites for each of these: For the first I use a sequence of UserRPL commands (based on my library) to compare to a single GoferList command, and then I reverse the perspective and use a single command from my library to compare with a sequence of GoferList/UserRPL commands. There's been some interesting results with these tests, and a couple of surprises. Commands in this category are Chars/S->NL, Strcat/NL->S, Zip/LCLLT, and Unzip/LDST.

Then there's all the remaining commands that have no counterparts. That's a large list from GoferList, but only a few from my library (LCNT, LEQ, LGRP, LREPL, LROT, LRPCT, LSDIV, LSHF, LSWP). That may grow over time, though. The results of testing so far have convinced me that there's enough of a value proposition to warrant further development effort for "yet another list library".