@rak1507 @Razetime Nah, I can't find it online. I may have only had it in actual print at the office. Meanwhile, let me make up a clue for you for 15→: Indicate where the integer ratio between M and N equals ME.
And rather than making up your own, it'd be much easier to port the originals.
And as you can see, I really need people to proof-read.
The format should be pretty self-explanatory, with each puzzle just being a very simple index.html file in a dir, and all the complicated stuff being in JS.
I don't get it. Isn't that exactly how the pages work?
@Razetime You don't need the actual numbers, only indication of which cells are numbered, which are empty and which are black. The numbers are in ravel order.
And the answer could conveniently be represented as another matrix, just of the proper characters, with spaces for the black cells, and with special characters for multi-choice cells.
I'm building a simple automation tool that will let me script a Dyalog APL session and run it as if I were typing. ]demo won'd do because I want to script )ED, other system commands and using the Tracer.
The core is working but I need to translate APL characters to prefix-char format as I have to send raw keystrokes. Has anyone got a text file that maps APL chrs to their prefixed representation (so ⍴ maps to r, for instance)?
I can extract stuff from a web page if necesary but a text file would be super if one exists.
D.bq and D.bqbqc from RIDE console: https://dzaima.github.io/paste/#0lVppk9y2Ef3uX2FtEjmR13buU8fGue/7tBwGQ2JmoCUJLgjuzuoqx3LWulaJZ0o@PtiRtauytHaVy6cUOYqr5H@SX5IGhpLRTTYmmS94xfcGaFyNRpMnlj679PWl/5x9bWl56XOA7l4H8HkH3gTwBQAHofyi05zbBfQlQIeg/LJ/cg3QVwAdhvKr/skVQF/zNbqKHgT0IJR/dk8uPAPoEUAfvQDgkAM3AawBOALlhtNsvwdI@v@fB2T8s3cBWUCnodx0T7aeA9R4NAOkvOodQNqhFy4AqgAdgPIJr/o7oCc9mgI6etTB81cBCv/X24Bqhy6eBZR55JofAkqgHHmDngU09mgL0DGPXgS0CughKHP35NJlQN/wtV4C9JBH7tlx3@bTgCYenQGU@jpuAFr3aA/QwLNuYEuP3KAXgE5CuexrexnKRz16CtBj3toPAZ32yPVln0fvA1rxutcBfcIj1/1PeuQG6VNedwfQn/yzc4D2@1afB3TAP/sHoE979Bagz3j0NqAE0D4oH/Z1…
Hi folks! My query engine saga continues :) So far my prototype eats 250MB of RAM (which I assume is mostly ⎕MAP'ed data that I precalculated earlier) and the core calculation for a query (not including JSON parsing) takes 120 us. My colleagues' code eats 9GB of RAM (they do precalculations but perhaps don't store the bool matrix as efficiently) and the core calculation takes 0.2s (!!!). So, while my code still has some aspects hardcoded, it has ~90% of the full functionality needed. Great!
I'm now looking at JSONServer. All the examples initialize data in each request, it seems. Is it possible to pre-initialize data (⎕MAP from files for example) once and then let the request handlers just use those variables?
(The stuff that's currently in production and that we'd like to replace is Go + some horrible hack on top of ElasticSearch)
In addition to the JSONServer question above, does anybody know of a way to )load dfns without nuking everything else that was in the current namespace? (This might be my misunderstanding of how namespaces etc. work in Dyalog APL...)
@dzaima Thanks, I think 'cmpx'⎕CY'dfns' is all I need currently. BTW I've just (through searching this chatroom for JSONServer) found your Android IDE. Just installed it - thanks for making it! :)
@MartinJaniczek no, that's unrelated - it's for the Dyalog bot and IRC replies
@dzaima (a reply is just :message-id-of-target, so you can even manually reply to yourself by just copying the ID from the message transcript link (which is what i used to do for months if not years until i finally wrote the userscript))
@MartinJaniczek Talking to myself: that InitializeFn indeed did exactly what I needed. Now I just need to make JSONServer give me JSON matrix instead of namespace 🙃
@dzaima I'm looking at the corner cases in iteration modifiers now. In dzaima/BQN, ≡˘ 2 returns 0 and ≡˘ <2 returns 1, as a plain number in both cases. My implementation has both enclosed. That seems obviously better for ≡˘ <2 although I'm not sure about the case with an atomic argument. Any thoughts?
I've also noticed that in my specification and implementation, ⚇¯1 is slightly different from ¨ because it doesn't map when all arguments are atoms, so that 3-¨2 is <1 but 3 -⚇¯1 2 is just 1. I think that's okay but I'm kind of wondering if I should change Each to match Depth as the non-enclosed version seems a little more useful.
@dzaima Even if one argument has rank >0 and the other has rank 0? That case is pretty useful, and I think if I allow that it only makes sense to open it up generally.
@Marshall what that's really asking for is a ˘-left and ˘-right, which is unsustainable, so i guess allowing rank 0 args is the only option to allow that
I'll let the backend guys react to it. It would need that last 10% to be fully generic and handle all requests, but ... so far it seems better than anything else we've tried. I'd need to talk to some higher-ups about buying a Dyalog license, of course. That and relative obscurity/unfamiliarity of APL is probably the biggest downside I expect to hear. But, I've learned APL enough to write this in a week or two of evenings.
So I'm hopeful, and excited by the results, but it might still be a (politics) battle :)
@MartinJaniczek I believe there is a setting for performing something once at server start. The method I used for TryAPL to do things once is to let each callback check if initialised, and initialise if not done yet, i.e. upon first callback only.
@Adám setting srv.InitializeFn←'Init' and then having Source/Init.dyalog with whatever I need has worked for me. And it runs that code before any request.
All assignments made in that init fn are visible from the request handler functions too, which is ideal for me
@MartinJaniczek Awesome. That was what I had in mind, I just didn't know the name offhand. Let us know if you need any help with JSONServer/Jarvis. I can get its main developer in here if you need anything, and I'm sure he'll be happy to add minor features too, should you need it. This all really sounds promising.
@MartinJaniczek Yeah, I thought that wasn't a built-in option. But obviously, it could be made so. In fact, one should probably be able to give any of the options to ⎕JSON to tweak functionality. E.g. to make it use JSON5 or return non-compact JSON. I'll pass your request forward. Also, have you tried using Jarvis? It is JSONServer's successor and is actively maintained.
@Adám I have looked at Jarvis a little bit, but it seemed to do much more than I wanted (REST stuff etc.). But if there is some kind of streamlined mode in which I can use it like JSONServer, then I'm more than willing to use that (if it's maintained as you say)
@dzaima Ok, is dzaima/APL parallelized? If so did you write custom bit-shifting implementations of array operations like ↑ and ⌽ to handle changing boolean arrays in parallel? I'm looking at options for parallelizing boolean array ops in April.
@dzaima I see, thanks, I've gained substantially from parallelizing regular array ops but I could just do boolean operations in serial. Doing ⌽ in parallel nets a 10x speed boost for me.
@phantomics dyadic or monadic ⌽? (my monadic ⌽ falls back to casting up do a double array, and it would need some special implementation to get it done bitwise)
