The APL Orchard

« first day (729 days earlier) ← previous day next day → last day (2222 days later) »

dzaima

07:31

@nathanrogers I don't think there'd me any way to do that without modifying the code as the compiler easily could inline variables and stuff (aside from live compilation)..

2 hours later…

Adám

09:30

@nathanrogers Yeah, I'm with dzaima there. I don't think that's even possible no matter which language is trying to call that library. It is the nature of compilation, that things turn static, no?

Adám

09:53

@J.Sallé Problem is that cells can only depend on cells that precede them in ravel order.

chrispsn

10:38

@Adám The dependency ("topological order") problem is tricky. Can APL support lazy (ideally memoised) evaluation of array elements?

Adám

@chrispsn Not natively. However, if each cell is stored as a formula, it is possible to keep reevaluating until the result is stable. That's the method used here. Each cell effectively sets a flag when "done", and formulas that refer to cells that are not flagged as done are skipped in that pass.

3 hours later…

Feeds

13:24

Q: Assigning a niladic function

A non-niladic function F can be assigned to a variable G with G ← F However, if F is niladic, how do I prevent it from being evaluated?

apl

2 hours later…

Feeds

15:31

Join us at the first Dyalog Meetup in New York, February 7th 2019. Interested in meeting APL users in the New York area? Sign up here https://www.meetup.com/en-AU/New-York-Dyalog-APL-Meetup/ Note: The location has recently changed - be sure to sign up to receive details of the new location & access instructions.

Tweeted by dyalogapl on January 24, 2019 at 2:19 PM

Come along to the January British APL Association meeting on Friday 25th at The Hoop And Grapes in London. For more information go to https://groups.google.com/forum/#!topic/baa-london/ja0EtJmRYqg

Tweeted by dyalogapl on January 24, 2019 at 3:26 PM

nathan rogers

16:06

I guess that leaves me with two options, dramatically modifying 2 immense libraries, and keep them both maintained as they're updated, or creating my own library of equivalent size in APL. 0_0 Well, looks like I'd better get started

perhaps create a global state object in APL, and pass that object, modifying any function containing unbound local scope variables to take the global object in APL

Only thing is, that's still a lot of work just to see if it is feasible. Decisions decisions

1 hour later…

dzaima

17:40

@nathanrogers personally I wouldn't change the actual functions but make small wrappers that modify the state and call the respective thing. I don't wouldn't think there'd me that much to add, as you'd have to actually use all the functions, and at worst case, the process should be automizable

nathan rogers

But I don't see how that's possible. It's expected that there is some global variable flag to be set, and when you call functions it checks the state of this global variable

I don't see how that global variable flag will exist in such a way that the C library can "see" it

for example:

// c code:
globalVar = 1;

//elsewhere
void f(){
if(globalVar == 1) {
// do something here
}
}

then in apl I can't say globalVar←1 because f doesn't know about apl

I can't call my apl imported f without failing

because globalVar was never initialized

dzaima

@nathanrogers so make void fW(int globalVarDef) { globalVar = globalVarDef; f(); }

nathan rogers

if I were to create a wrapper like you said, won't the state created by that function be gone or garbage after it exits?

how do I know that it is being allocated correctly? and that it will persist?

all of the state that I have is being returned and held in APL variables

dzaima

@nathanrogers i don't actually know how ⎕NA works, i'm just giving general food for thought ¯\_(ツ)_/¯

nathan rogers

sure

I could give that a shot, I just don't know what happens to the state once a function exits

dzaima

17:46

how much of that global state actually is there? i.e. is it viable to pass & re-set it again for every function call?

nathan rogers

I'm not sure, but it appears to be prolific

I would like to see some kind of setter function like you're saying, or a means of extricating the global state into a central object that can be managed more centrally

because then I could create that global object in APL

« first day (729 days earlier) ← previous day next day → last day (2222 days later) »

all rooms