Reorder arguments to index, adjust, and adjust'

[DaveBarton]

For other arrays, containers, strings, etc., including Sequence that users will often be converting from, index takes the container before the Int, and adjust and adjust' take the (a -> a) function as the first argument. In addition to being consistent with these cases, I think one can argue that the standard order is more often useful for currying.

I realize this is a breaking change, but I think it'd be much better to make it now, before there are too many users, rather than wait. Users that are surprised by the change should at least get a compiler error, plus you'll obviously bump the library's version number.

Thanks very much for this library!! Sorry to nit-pick.

[DaveBarton]

Whether you do this change or not, you could add a WARNING pragma to these functions, at least temporarily. Also, if you're not sure, you could include a request for feedback or an opinion/vote in the warning.

Finally, this issue applies to index and update in Data.AMT in extended-containers also.

[konsumlamm]

My reasoning for using a different argument order was consistency (the index is always the first argument). However, I do see the appeal of using the same argument order as Data.Sequence etc. to allow easy transitions. Can you give an example why you think the standard order is better for currying?

[DaveBarton]

(index v) can be passed to a linear or binary search algorithm, or anything else that expects a (possibly memoized) function, such as . (composition) or maybe or really anything else in functional programming. (index i) [for an Int i] would only be useful to pass to a function that works on potentially many vectors by always looking at just their ith element, which is rarer than a memoized function of Ints. For adjust and adjust', I guess the (a -> a) function could be something that you want done to a bunch of vector elements, like incrementing/decrementing them by 1 or some other fixed amount. Again, it's rarer that you want the ability to adjust a fixed-index set of elements by various (a -> a) functions. (I do think this currying argument is weaker for adjust/adjust' than for index, and the consistency argument with other container types is the main one.)

FWIW, the ListLike package uses the standard order for index also.