On Thursday, 20 August 2020 at 08:26:59 UTC, Ali Çehreli wrote:
On 8/19/20 9:11 PM, data pulverizer wrote:
Thanks. I might go for a design like this:

```
struct View(T){
   T* data;
   long[2][] ranges;
}
```
[...]

I implemented the same idea recently; it's a fun exercise. :) I didn't bother with opIndex because my use case was happy with just the InputRange primitives (and .length I think).

And I had to implement it because std.range.chain works only with statically known number of sub-ranges. :/ If the number of ranges are known, then this works:

import std.stdio;
import std.range;

void main() {
  auto x = [1,  2,  3, 4,  5,
            6,  7,  8, 9,  10,
            11, 12, 13, 14, 15];
  auto y = chain(x[0..5], x[9..14]);
  writeln(y);
}

[1, 2, 3, 4, 5, 10, 11, 12, 13, 14]

Ali

Many thanks for confirming the internal structure of D's arrays and for the tip on std.range's chain function. It's exactly what I need. In fact the number of sub-ranges are statically known ...

As an aside, the reason for this query is that I have written a small module for multidimensional arrays to be included in a GLM (generalized linear models) package I am writing in D (in fact this will be the second major version of the package in D) - I know about Mir but my array doesn't have to be feature rich and only requires very few methods, I'm also probably going to write an article about it's internal structure and I've learned a lot creating it, and it means the GLM library won't have dependencies outside D's standard library and BLAS/LAPACK.

The multidimensional array is structured like this:

```
Array(long N, T)
if(isFloatingPoint!T && (N >= 1))
{
  T[] data;
  long[N] dim;
  ...
}
```

Where N is the number of dimensions. The indexing and stuff works fine but I wasn't happy that subsetting the array with slices returns a copy (for example an 2D array `A[0..2, 1..$]`) and since the subsetting is not necessarily contiguous, directly slicing from the data array was not feasible. But now N is known at compile time (`A[r[0][0]..r[0][1], r[1][0]..r[1][1], ... , r[N][0]..r[N][1]]`) so doing a subset using std.range's `chain` on `data` is easy. It also means I don't have to implement a separate `View` struct. This chain function will also simplify my indexing code, at the moment I am using string mixins to generate code for for loops over all the different dimensions where the function creating string the is recursive ... which was fun to write [:laugh:]!

Reply via email to