What are peoples view on what ranges should be?

[crazymonkyyy]

There are 4? view points I see:

1. mine(i.e. correct one), ranges should be thought of as fragments of imperative "patterns"; a "slide" is when you use a += instead of a ++ in a for loop, as such you still have access to the index for several things that you dont in the current std, the 3 functions map pretty well to the 3 sections to the for loop and should probably not change, and in-place sorting just makes sense.

2. database query construction; map/reduce is something something database queries, there's all these jobs about data analysis; python data frames are considered to be great according to large sections of the general programming public; they probably dont see in-place sorting as sane, want more managed data structures and python compatibility

3. "source sink": walter wrote an article about how great ranges will be because bash scripting, maybe iostreams, idk if anyone still thinks that's the best metaphor.

4. functional; maybe someone will want immutable data structures to monad the nullable sumtypes; and will grab a random concept from lisp, haskell or apl and it will be named after a bird.

Are there more?

[pbackus]

Ranges are a common interface or "narrow waist" between algorithms and data sources.

The existence of a common interface means that, if you have N algorithms and M data sources, instead of having to write individual implementations for each of the N×M combinations, you can write each algorithm once, using ranges, and it will work with every data source. In other words, you can implement N×M use-cases with only O(N+M) code.

This view implies that the range interface should be designed first and foremost to give authors of generic algorithms the primitives they need to write correct, efficient code.

[crazymonkyyy]

O(n+m)

surely thats "why" and not "what"
whats the "correct, efficient code" Is it radix sort cause its just so great? is it downloading json from some mircoserverse that you then instantly turn around and make a minor change and host? Is it making bash scripts saner?

[crazymonkyyy]

Im asking so I can attempt to write a new std.range.interfaces that isnt based off a very slightly-modified stl concept(we airnt a 90's c++ and theres been decades of experience since, it shouldn't be hard to do better)

see: #13

I know theres other people talking about it but I cant parse verbose enterprise grade nonsense

can I summarize the "Copy Semantics" talk with one line of code:
https://forum.dlang.org/post/[email protected]

are there other considerations and people in this design space?