Add array.zip_with and array.imap to the standard library

[vkleen]

This change adds the functions zip_with andimap to the array standard library. Both were quite useful in a recent project of mine and seem general enough to warrant inclusion.

[yannham]

This is a fair addition. I wonder about the naming though. OCaml and Scala use map2 instead of zip_with (and mapi instead of imap). Rust doesn't have them, I guess you use .enumerate() or .zip() and then do a normal map. In the JavaScript world, lodash (some kind of common utility extension of the stdlib) use zipWith.

While I understand why zip_with makes sense in a language with collections, tuples, curryfication and higher order functions, I wonder if the convention of all those variants being just map with a different suffix isn't more discoverable. But maybe that's just the stubborn Ocamler inside me 😕

[vkleen]

The biggest reason to go with zip_with alone and not with zip followed by map is that Nickel doesn't have first-class tuples. Both zip_with and imap as names betray my Haskell background, I guess. I wouldn't be opposed to map2 (and eventually map3 and so on), though.

On a side note, an error message in the manual currently depends on the line count of std.ncl, which scares the CI. I wonder if there's an easy way to hide that.

[yannham]

Let's discuss that further in the weekly meeting, and merge the implementation as it is. We just need to decide before the next version.

[yannham]

On a side note, an error message in the manual currently depends on the line count of std.ncl, which scares the CI. I wonder if there's an easy way to hide that.

Right, I had the same issue on a recent PR. Maybe we can use the same trick that we use for snapshot testing, where we strip off the local paths and replace them with a generic placeholder (or, rather, a co-placeholder 🤔 - a placetaker ?)