Ramda String Functions

[sanctuary-js]

Ramda String Functions

[davidchambers]

Before considering string slice functions, let's decide whether a sanctified version of R.slice is a good idea. We could implement S.tail and S.init in terms of S.slice:

S.tail = slice(1, Infinity);
S.init = slice(0, -1);

> S.slice(2, 4, [])
Nothing()
> S.slice(2, 4, ['a', 'b', 'c', 'd', 'e'])
Just(['c', 'd'])

I'm not sure how S.slice(2, 102, ['a', 'b', 'c', 'd', 'e']) would evaluate.

[svozza]

I guess there's not much difference between slice when toIndex > arr.length and tail so I think it's justified in only returning the remaining elements of the array. I don't think it would be that useful to return what amounts to a sparse array with 96 Nothings or Just(undefined)s.

[davidchambers]

I'm wondering how the following translates to the hypothetical S.slice:

> S.tail([1, 2, 3])
Just([2, 3])
> S.tail([1, 2])
Just([2])
> S.tail([1])
Just([])
> S.tail([])
Nothing()

It seems S.slice should only return a Just if the array contains the entire range of indexes (otherwise it isn't well differentiated from R.slice):

> S.slice(1, 1, ['a', 'b', 'c'])
Just([])
> S.slice(1, 2, ['a', 'b', 'c'])
Just(['b'])
> S.slice(1, 3, ['a', 'b', 'c'])
Just(['b', 'c'])
> S.slice(1, 4, ['a', 'b', 'c'])
Nothing()

This would guarantee that if startIndex >= 0 and endIndex >= 0 and startIndex <= endIndex, the resulting list will contain endIndex - startIndex elements.

[svozza]

Ah okay, I see where you're coming from. Seems like that's the implementation favoured here:

https://wiki.haskell.org/99_questions/Solutions/18

And here:

https://hackage.haskell.org/package/repa-array-4.1.0.1/docs/src/Data-Repa-Array-Auto-Operator.html

slice   :: Elem a => Int -> Int -> Array a -> Maybe (Array a)
slice from len arr
        | from >= 0, len >= 0
        , len  <= G.length arr - from
        = Just $ A.window from len arr

        | otherwise
        = Nothing

[davidchambers]

Thanks for investigating, @svozza. Let's add S.slice before deciding whether any string-specific functions are necessary.

[svozza]

Sounds good, I'll pick up this PR of you want.

[davidchambers]

Go for it! :)

[davidchambers]

We could define S.take and S.drop:

> S.take(2, ['a', 'b', 'c'])
Just(['a', 'b'])
> S.take(3, ['a', 'b', 'c'])
Just(['a', 'b', 'c'])
> S.take(4, ['a', 'b', 'c'])
Nothing()
> S.drop(2, ['a', 'b', 'c'])
Just(['c'])
> S.drop(3, ['a', 'b', 'c'])
Just([])
> S.drop(4, ['a', 'b', 'c'])
Nothing()

ramda/ramda#1159 deprecates R.substringTo and R.substringFrom. We could define S.take and S.drop in terms of S.slice so they would also work with strings.

[svozza]

Yep, sounds good. I can work on this later on today.

[davidchambers]

We now have S.slice, S.take, and S.drop, all of which can operate on strings (in addition to S.indexOf and S.lastIndexOf which are similarly polymorphic).