I had a look around but nothing obvious jumped out at me and I needed to join two 2D lists in a hurry so came up with this
but I'm wondering if there is an "official" library block to do the job - apart from the APL catenate which seems overkill for the job 
I was asking if there was an existing library block not how to make one - I'd already done that with my join block
Yes, catenate, from the APL library 
So what?
I couldn’t copy your join block, and was not sure whether it were the real thing or a mock-up. So I shook this code out of my sleeve, just in case. You’re welcome! 
The catenate block has a lot of dependicies whereas yours and my solutions are lean
Plus it doesn't have variadic inputs so would have to be nested to use it for more than 2 lists
This is the innards of my block
Yeah... I couldn't believe how many until I read its code. The main thing is that it works for any number of dimensions, not just two; that's why it uses RESHAPE. (I should look into whether the primitive RESHAPE, which was added after I wrote the library, would work.)
(I should also get someone who's better at optimization than I am to redo the library, after he finishes logic programming. :~) )
Ask @cymplecy, their join block is better than mine.
Warning: whataboutism … what about the Reference Manual? 
Seriously: I’m dreading it, somewhat.
Really? I would have thought that COLUMNS would be faster than MAP.
Oh. Let's talk about that. Maybe in another thread? I thought you were well under way!
Your right - did a test and @qwe version is about twice as fast
On the other hand, your code is simpler.
Yeah, it seems that MAP, especially nested MAPs, aren't as fast as many of the other primitives, so I do my best to avoid it. I try to use COMBINATIONS and COMBINE instead. Here are a couple examples of how to do so (examples taken from problems discussed here: What should this function be named? - (answer group), because it is a recent topic that I tried to figure out how to optimize):
First, we set the variable "Table" to:
Similarly, we can set "Table" to:
The reason why MAP is so slow is actually explained here.
Gah, I was with you until that headache-inducing code! (And, btw, how did you make a result pic that isn't a smart pic (loadable into Snap! as code)? I was going to try to solve this problem without such a mess by using DISTRIBUTION OF (one of the options in the portmanteau LENGTH block) but I'm not motivated enough to enter your TABLE list by hand.
I didn't, but the forum seems to have compressed the images. If you click on them and then drag&drop it should work.
The "fixed combine" block basically just makes sure that the first item is the same rank as the rest of the list when entered. With combine:
In iteration 1, when using a table, the first input will be a 1 dimensional list. However, if your combine then adds another 1 dimensional list to that, then in the next iteration the first input will be a 2 dimensional list. So we need to cover both cases.
With my "fixed" combine:
In iteration 1, when using a table, the first input will be a 2 dimensional list. When the combine then adds another 1 dimensional list to that, then in the next iteration will be a 2 dimensional list once again, so the code does not need to cover both options.
I'm not actually clear what that does, so I would be interested to see what you come up with (if you end up doing it).
I was wrong, DISTRIBUTION OF isn't flexible enough. I had to load the Frequency Distribution Analysis library and do this:
P.S. If the function you're mapping over the list is simple enough, you can right-click the MAP block and choose "Compile" to speed it up. Then it looks like this:
What it compiles isn't the MAP itself but the mapped function (the first input).
In my code, you can compile the inner map:
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.
