www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - understanding mir, generic nd array iterator

reply monkyyy <crazymonkyyy gmail.com> writes:
I asked

 does a preexisting nd array iterator exist?


got links to mir, docs I can barely read (srsly math language is 
anticomprehension)

ndslice allocates, so, hottake misnamed. A slice is a reference 
to an array, a dynmaic array is a gc allocated array 
https://opendlang.org/library/mir.ndslice.allocation.slice.4.html#Slice

Found `field`, says its a something defined by an opIndex, 
better, but no nd

`fieldnd.d`or something exists but it defines 2 types of nd 
fields and not a generic mapping, idk

`FieldIterator` exists but its opSlice only takes 2 arguments so 
I think I can rule it out as nd

etc.

So I dont think it exists in this tangled mess? Maybe, idk? I 
think in its naming scheme it would be called an ndfield
Sep 10
next sibling parent monkyyy <crazymonkyyy gmail.com> writes:
On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 I asked

 does a preexisting nd array iterator exist?


 got links to mir, docs I can barely read (srsly math language 
 is anticomprehension)

 ndslice allocates, so, hottake misnamed. A slice is a reference 
 to an array, a dynmaic array is a gc allocated array 
 https://opendlang.org/library/mir.ndslice.allocation.slice.4.html#Slice

 Found `field`, says its a something defined by an opIndex, 
 better, but no nd

 `fieldnd.d`or something exists but it defines 2 types of nd 
 fields and not a generic mapping, idk

 `FieldIterator` exists but its opSlice only takes 2 arguments 
 so I think I can rule it out as nd

 etc.

 So I dont think it exists in this tangled mess? Maybe, idk? I 
 think in its naming scheme it would be called an ndfield


Will need to gentricify this, but this seems promising

```d
import std;
struct Tuple(T...){
	enum istuple=true;
	T expand; alias expand this;
}
auto tuple(T...)(T args){
	return Tuple!T(args);
}
unittest{
	auto foo=tuple(1,"hi");
	assert(foo[0]==1);
	assert(foo[1]=="hi");
	auto bar=tuple();
}
enum istuple(T)=is(typeof(T.istuple));
unittest{
	assert(istuple!(typeof(tuple(1,2)))==true);
	assert(istuple!int==false);
}
auto iterate(T,S)(T t,S s){
	struct foreach_{
	int opApply(int delegate(int,int) dg){
		int result;
		static foreach(TI;0..t.length){
		foreach(x;t[TI]){
			static foreach(SI;0..s.length){
			foreach(y;s[SI]){
				result=dg(x,y);
				if(result>2){return result;}
				if(result==1){break;}
			}}
		}}
		return result;
	}}
	return foreach_();
}
unittest{
	loop:foreach(x,y;iterate(tuple(iota(0,3),iota(6,9)),tuple(iota(4,6)))){
		writeln(x,',',y);
		if(x==6){break;}
		if(x==7){break loop;}
}}
```
Sep 14
prev sibling next sibling parent reply Dejan Lekic <dejan.lekic gmail.com> writes:
On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 I asked

 does a preexisting nd array iterator exist?


 got links to mir, docs I can barely read (srsly math language 
 is anticomprehension)


I can barely read your posts, yet I do not complain...
Sep 15
parent reply monkyyy <crazymonkyyy gmail.com> writes:
On Monday, 15 September 2025 at 16:59:58 UTC, Dejan Lekic wrote:

 On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 I asked

 does a preexisting nd array iterator exist?


 got links to mir, docs I can barely read (srsly math language 
 is anticomprehension)


 I can barely read your posts, yet I do not complain...


you havnt? You sure?
Sep 15
parent reply Andy Valencia <dont spam.me> writes:
On Monday, 15 September 2025 at 17:07:36 UTC, monkyyy wrote:

 On Monday, 15 September 2025 at 16:59:58 UTC, Dejan Lekic wrote:

 On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 I asked

 does a preexisting nd array iterator exist?


 got links to mir, docs I can barely read (srsly math language 
 is anticomprehension)


 I can barely read your posts, yet I do not complain...


 you havnt? You sure?


Not going to lie, I had no idea what "mir" meant, nor what an "nd 
array" is.  For such things I mostly wait and see if I can pick 
up my answers from the context of any responses.

Mostly! :-)

Andy
Sep 15
parent reply monkyyy <crazymonkyyy gmail.com> writes:
On Monday, 15 September 2025 at 20:10:22 UTC, Andy Valencia wrote:

 On Monday, 15 September 2025 at 17:07:36 UTC, monkyyy wrote:

 On Monday, 15 September 2025 at 16:59:58 UTC, Dejan Lekic 
 wrote:

 On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy 
 wrote:

 I asked

 does a preexisting nd array iterator exist?


 got links to mir, docs I can barely read (srsly math 
 language is anticomprehension)


 I can barely read your posts, yet I do not complain...


 you havnt? You sure?


 Not going to lie, I had no idea what "mir" meant, nor what an 
 "nd array" is.  For such things I mostly wait and see if I can 
 pick up my answers from the context of any responses.

 Mostly! :-)

 Andy


mir is a lib that has nd math; that allot of people know is meta 
programming to hell and back but no one can tell ya how to use it
nd is "N" as anonymous int and D as in dimension, 2d vs 3d vs ... 
99d
Sep 15
parent Andy Valencia <dont spam.me> writes:
On Monday, 15 September 2025 at 20:23:23 UTC, monkyyy wrote:


 mir is a lib that has nd math; that allot of people know is 
 meta programming to hell and back but no one can tell ya how to 
 use it
 nd is "N" as anonymous int and D as in dimension, 2d vs 3d vs 
 ... 99d


Cool... thanks.  A world I never even knew existed!

Andy
Sep 15
prev sibling parent reply jmh530 <john.michael.hall gmail.com> writes:
On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 [snip]


Documentation for mir iterators is here:
http://mir-algorithm.libmir.org/mir_ndslice_iterator.html

If you want to use mir without using the GC. You can allocate 
with malloc, custom allocator, or use an RC allocator.
http://mir-algorithm.libmir.org/mir_ndslice_allocation.html
Sep 17
parent reply monkyyy <crazymonkyyy gmail.com> writes:
On Wednesday, 17 September 2025 at 12:06:32 UTC, jmh530 wrote:

 On Wednesday, 10 September 2025 at 20:23:28 UTC, monkyyy wrote:

 [snip]


 Documentation for mir iterators is here:
 http://mir-algorithm.libmir.org/mir_ndslice_iterator.html


I see nd-iota but it seems under powered(handling only the single 
argument case of iota)
I could spell that `auto fixediota(I)(I start,I 
end)=>iota(end-start).map!(a=>a+start)` to get the 2 argument 
version (3 arguments could be a little tricky)

soooo
```d
fixedndiota(???)=>...
ndfield(???)=>fixediota(???).map!(a=>foo.opIndex(a));
```

That doesnt seem premade anymore


 If you want to use mir without using the GC. You can allocate 
 with malloc, custom allocator, or use an RC allocator.
 http://mir-algorithm.libmir.org/mir_ndslice_allocation.html


Not about allocating, user defined opIndex on user defined data

If mir only defines a single data-type-family and is deeply 
opinionated about controlling it(which seems to be increasingly 
the case) its not generic and leaving some level of simplicity on 
the table.
Sep 17
parent reply jmh530 <john.michael.hall gmail.com> writes:
On Wednesday, 17 September 2025 at 14:57:19 UTC, monkyyy wrote:

 [snip]


Sorry, I'm having trouble understanding what you're trying to do.
Sep 17
parent monkyyy <crazymonkyyy gmail.com> writes:
On Wednesday, 17 September 2025 at 19:02:22 UTC, jmh530 wrote:

 On Wednesday, 17 September 2025 at 14:57:19 UTC, monkyyy wrote:

 [snip]


 Sorry, I'm having trouble understanding what you're trying to 
 do.


I was hoping this code already existed

https://gist.github.com/crazymonkyyy/bb6c52dd08ffb55b8e38b3f66896f459
Sep 25