stdlib: add List.intersperse and List.intercalate

[craigfe]

Adds a new function – List.intersperse : 'a -> 'a list -> 'a list – to the standard library. This transforms a list l by adding a separating element between each adjacent pair of elements in l. For example:

# List.intersperse 0 [1; 2; 3] ;;
- : int list = [1; 0; 2; 0; 3]

# List.intersperse 0 [] ;;
- : int list = []

I re-define this function reasonably often, and it's slightly annoying to do so due to the asymmetry of the empty list case. Example use-cases I've had in the past:

• adding comma / tab separators to a list representing .csv / .tsv fields;
• adding yield points to a list of computations;
• adding "spacing" elements to a DSL term (for file separators, break hints, etc.).

Notes:

• Other standard libraries that include this function: Jane Street base, Haskell base and Elm core.

• If intersperse is deemed useful, I think there's a strong argument for (at least) Seq.intersperse as well. I'm happy to add that to this PR.

[nojb]

Left some general comments. I don't have a strong opinion on whether to include this or not in the stdlib; let's wait for other developers to chime in.

[nojb]

Probably better to make the function tail-recursive.

[craigfe]

Done 🙂

[nojb]

Why not go further and allow the separator to be a 'a list?

[craigfe]

The 'a -> 'a list -> 'a list form has a more descriptive type and I think more intuitively matches the name "intersperse".

Anecdotally, I've never found I needed the more general form either. YMMV.

[dbuenzli]

Personally I'm not sure I had that much of a need for intersperse (but I don't mind). I rather often missed the signature of String.concat for List (which would be 'a list -> 'a list list -> 'a list).

[craigfe]

@dbuenzli: Interesting. I've never found a need for that one, but I know Haskell developers like it, and they call it intercalate. I'm happy to add that along with this PR if it's deemed worthy; it's in some sense the natural sibling to intersperse.

[craigfe]

Rebased and added List.intercalate as per @dbuenzli's suggestion.

[damiendoligez]

You should investigate the CI failures.

I don't have a strong opinion on adding these functions. Maybe @gasche or @alainfrisch would like to comment?

[craigfe]

AppVeyor is now green. A maintainer will need to explicitly approve triggering the other CIs (thanks to GitHub's new security policy).

[dbuenzli]

Still missing that fun. Btw. I'm not sure I see the interest of adding intersperse once you have intercalate

[Hirrolot]

What's the status of this issue? I need interleave quite often. You can also find it in the standard library of Haskell and in libraries extending OCaml std, such as containers.

[nojb]

What's the status of this issue? I need interleave quite often. You can also find it in the standard library of Haskell and in libraries extending OCaml std, such as containers.

The status is that the PR is stalled. Judging from the discussion, there wasn't a lot of support for the addition, nor was there strong opposition. There was some related discussion in #10583 (comment).

Again judging from the discussion, there does not seem to be a lot of consensus about which signaturess are the most useful in practice.

In any case, I think any addition in this direction should be done in both List and Seq simultaneusly, and evaluated accordingly.

[Hirrolot]

@nojb, I can make another PR that borrows code from this PR and introduces the same functions for List and Seq simultaneously, and then we can re-evaluate it. How does that sound?

[nojb]

@nojb, I can make another PR that borrows code from this PR and introduces the same functions for List and Seq simultaneously, and then we can re-evaluate it. How does that sound?

No opposition in principle.

[craigfe]

Closed as I don't have the bandwidth to un-stall this at the moment. Best of luck with the follow-up PR @Hirrolot :-)