The APL Source Code (2012)
computerhistory.org
computerhistory.org
Its introduction says: "Although it maintains the concept of array-driven computing and much of APL's array functionality, BQN discards all compatibility with other array languages and changes many fundamental concepts. It uses the based array model with dedicated array notation, distinguishes between data types and expression roles to give the language a context-free grammar with first-class functions, and uses a new set of glyphs with different primitive pairings."
Disclaimer: author
What is a fairly interesting observation (more people have made it, including Arthur Whitney); being able to reason about a program for which you can see all the code on one A4 without having to clickthrough into functions, search, browse through files, classes or even having to scroll is incredibly powerful. All you need is the cheatsheet (provided in your article) (which you know by heart in no time flat) and the program, which, often fits on 1 A4 or less. Modern example of that would be Aaron Hsu his 17 LoC parallel APL compiler[0] (from his thesis).
No, it’s quite a bit more than C offers, which has a paltry twenty-eight “operators”, none of which “work” on arrays. Most of the libraries in C and C-like languages have huge libraries because getting those twenty-eight operators to do anything can be a lot of work if you don’t think about it much.
In contrast, Iverson has more than 50 excellent operators, sensibly extended to arrays. Not all APLish languages have all the operators, but the choice about what they do have typically considers how all of Iversons operators can be/are used and how/when they can be implemented in terms of each other. This is largely a matter of taste.
And better symbols means better ideas. Some things are just more obvious in APL. And once you see it, you can take it with you.
When using C, I don’t typically use the C library, or libraries for any of those things you mentioned, because they just aren’t that complicated once you have seen APLish solutions to those problems. And that’s a good thing because less code means less bugs.
APL emphasizes some "formal external beauty" (of code) over the true beauty of an algorithm. Of course you can do such or such thing in 2½ characters of APL code rather than 60 boring lines of whatever language you want but it is generally full of unneeded hidden loops (or looping uselessly on whole parts of arrays when only a small part is required). Actual complexity of these "elegant" APL solutions is generaly insane.
On the other hand, the J language has a set of some "optimized" idioms -if I remember correctly. Not sure the thing is really better: "write it that way by pretending you are following the array-way-of-life; of course it would be insane to actually do it that way, but my engine will detect the very specific thing you want to do and will use a state-of-the-art algorithm for that purpose". Not sure it is a good thing because the task becomes over complicated: knowing the true (boring) algorithm, knowing the specific array-syntax encoding that algorithm, which falls back to looking in some "list" of what is optimized or not.
While I agree these languages are fascinating to think about computing, learn new things, etc. I finally went to more or less agree with Dijkstra (not the worst reference?) about these languages being "a mistake, carried through to perfection".
What is wrong about writing well commented, well indented, 30 lines of code if I can implement the exact algorithm that way?
Nothing, it's just interesting and I wish I could do a larger project with one of them outside just playing around.
Or an interactive cheatsheet website[1], where you can paste APL code, and mousing over symbols gives their definitions?
[1] http://I-wish-this-existed-but ...
Not so. Anyone who says these things simply does not know APL. It's like saying musical notation is write-only. Sure, if you don't know it reading the darn thing is excruciating. Yet, if you do, you can see music on the page.
What APL is not is a low-skill, low-knowledge language. A casual glance at the language isn't enough to know it. Just as you can't internalize musical notation without dedication, work and experience.
Back in the '80's I used APL professionally for about a decade. From that context, it reads like music. To this day I can read it just fine (though not as quickly). Interestingly enough the same is the case musical notation, which I could read without thinking twenty years ago, when I was studying and playing classical guitar and piano with great regularity.
That said, I would not recommend anyone get serious about APL today. Study with some degree of dedication? Yes, I think that would be of value. Real projects? No. Definitely not.
Back in the 80s I was an mainframe engineer at IBM. One of our customers (a bank?) was heavily into APL, and occasionally I would fix one of their keyboards.
(Sidenote: these were beautiful keyboards, and their keys were individually repairable. Also pleased to share my first mastodon link! https://mastodon.social/@hanshuebner/109412183054578487)
Watching people programming on these looked so intense and intimidating that I decided software wasn't for me, and it took years to find out today actually it was.
By the way, APL was the first language I learned, on the APL/360 system described here, using paper terminals.
Reading APL code, at least for a novice me, includes puzzling out "ok, now what's going on there?". Being able to highlight a bit of code, and get an explanation, as if sitting with the author, or reading annotated code, might usefully lower the barrier to reading APL?
I've seen several "here is an APLish program, and I'll go through and explain it" posts over the years, though I'm failing to find one now. If that could be somewhat automated...
[1] https://medium.com/@solarbreeze69/chatgpt-explains-arthur-wh... [2] In "J one-page interpreter fragment (1992)" https://news.ycombinator.com/item?id=34050715 , the brief comment https://news.ycombinator.com/item?id=34053932 .
https://podcasts.apple.com/us/podcast/the-array-cast/id15688...
So the original Perl?
APL is a completely different paradigm.
Every now and then I get the urge to take a deep dive into APL or J but I really need a use case.
Since they are array languages they seem like they might be good for machine learning?
APL is a tool for thought more than a real programming language so in my opinion it's best suited to boost your puzzle solving ability.
(opticl:write-png-file
"~/out.png" (april-c "2 0∘{(x s)←⍺←2 0 ⋄ (⌽[x],[x]⊢)⌽[x]⍣s⊢⍵↓[x]⍨(1-2×s)×⌊2÷⍨x⊃⍴⍵}"
(opticl:read-png-file "~/in.png")))
April was used to teach image filtering in a programming class for middle-schoolers, you can see a summary in this video: https://vimeo.com/504928819
For more APL-driven graphics, April's repo includes an ncurses demo featuring a convolution kernel powered by ⌺, the stencil operator: https://github.com/phantomics/april/tree/master/demos/ncurse...
[1] https://github.com/phantomics/aprilYou can create obscure Perl as you can create obfuscated C as in any language, but if you think about the reader you can create 'easy to read' program in Perl, even by a beginner.
But creating 'not obfuscated' APL? That's really hard!
I find it more natural to evaluate left-to-right, because that's the order we read things in. For example, to read a file and sort its lines (in D):
auto sortedLines = File("file.txt").byLineCopy.array.sort; ↑ nget "file.txt"
File("file.txt").byLineCopy.array.sort = sortedLines;
...which by first blush, seems bananas, but if you were to read it as English, would make sense: "Load the file 'file.txt' into an array of lines, sort them, then assign it to sortedLines"
Yes. Though one could use the . operator and a sink to make it completely left-to-right.
Especially if you add the |> operator (from Erlang I believe) to 'pipe' the data.
File("file.txt") |> readlines |> sort -> sortedLines;
We have to agree to these kinds of terms about something that is more than half a century old and has zero commercial value. Something in the copyright system is deeply broken.
I don't disagree.
The trouble is that, while almost everything 50 years old has almost no commercial value, a few things that old have tremendous commercial value.
My preferred solution is to have a very short automatic copyright period combined with a fee for registered works that increases exponentially based on the date of registration. Not that it really matters - I'm sure such a solution would never be enacted into law.
Each such source deck would have been assembled to make an object deck. There are assembler commands for formatting the printed assembly listing: TITLE to set a running title on the listing pages, PRINT OFF to avoid listing repetitive macros, PRINT ON, PRINT NOGEN to not print macro-generated statements, EJECT to start a new listing page.
A DSECT line introduces a data section, essentially a record. The pseudo-instruction USING tells the assembler which register contains the base address of a DSECT, so USING MKLOCLS,14 says, assume R14 has the address of the DSECT MKLOCLS. Reference to a name defined within that DSECT will generate an instruction using R14 plus the offset to that name.
A CSECT names a "control section" or named block of instructions. The linker handles CSECTs and DSECTs as units. A CSECT typically begins with USING *,15 or some other register, to inform the assembler that at run-time R15 will have the base address of this code. Local jump addresses will assemble as offsets from that register. A DROP <reg> statement tells the assembler to no longer assume that register is in use.
There are inside jokes: the code with TITLE 'THE AGORONOMIC ROUTINES' is memory management.
APL\360 managed memory on two levels. First, as a time-sharing system, it created a separate "workspace" for each logged-on user. A workspace was a contiguous block of memory (typically 32K as a practical size) that contained all of the user's current code definitions. There was not room in the typical 360 RAM for more than a few user workspaces at one time, so the system swapped whole workspaces in and out as it serviced each user. The code to do this operated the disk drives directly on a head and track level and was highly optimized. You'll see a number of references to "one-track workspace" as a special case; if a workspace fit on one track of the (2311 or 2315) disk, it could be read in or written out in minimum time.
On the user level, the interpreter while executing APL code is constantly creating and trashing values which are stored as memory objects within the current workspace. So there was memory allocation and garbage collection. Garbage collection was a simple sweep across the workspace moving all active objects to one end.
APL would be initiated as a normal job under OS/MFT and would immediately take control of the entire system, kicking the OS out and managing all aspects of the 360 hardware itself, including interrupt handling, storage protection and so on. This was before the day of hardware-supported virtual memory, so it didn't have that to deal with. See NAME=APLSAPLM and forward. I'm pretty sure in that code, TCB stands for Task Control Block, a dispatching object. APL had multiple internal tasks being time-shared, as necessitated by the fact that it had to handle asynchronous interrupts from both terminals and disk drives.