Port List functions from JavaScript to Elm

[jonathanhefner]

This builds off of #150, so that should be merged first.

In the same spirit as #130, I ported several List API functions from JavaScript to pure Elm. Some notes:

• I kept scanl in a separate commit because it is a tiny performance regression (1.55 -> 1.65 seconds on an old laptop for a 100k element list). The regression is due to the requisite extra reverse call.
• I did not port sortBy in terms of sortWith because that resulted in a larger performance regression (2.5 -> 3.0 seconds on an old laptop for a 100k element list). The regression is (likely) due to the extra string comparison required by the JavaScript sortWith implementation. If we decide the performance penalty is worth it, I can port it also.
• There are several native functions (any, drop, take, repeat) which were not converted, but could be by introducing an additional 1 or 2 native functions to drive looping. Either a function like foldlWhile that handles early termination, and / or a function like foldn (or Range.foldl) that handles counting without creating intermediary lists (i.e. without [1..n]). I can implement these in a separate experimental branch if anyone just wants to see what they would look like.

[evancz]

This looks great, nice work!

I think the last point about any, drop, etc. is interesting. It'd be good to move more code into Elm, so I'd be curious to see what this looks like and if it has performance implications.

[evancz]

This looks great, nice work!

I think the last point about any, drop, etc. is interesting. It'd be good to move more code into Elm, so I'd be curious to see what this looks like and if it has performance implications.

[jvoigtlaender]

Concerning the "implement some functions using native loops" aspect, there are already some candidate implementations for such loops in https://github.com/TheSeamau5/elm-loop.

[jvoigtlaender]

Concerning the "implement some functions using native loops" aspect, there are already some candidate implementations for such loops in https://github.com/TheSeamau5/elm-loop.

[jonathanhefner]

I created three branches to show what looping could be like for any, repeat, take, and drop:

• Using foldn
  • any cannot be implement via foldn, and thus is not ported
• Using foldlWhile
  • repeat cannot be implemented via foldlWhile, but a safe, less-efficient, pure-Elm version is included for consideration (note the ** in the benchmark data below)
• Using loopUntil
  • all 4 functions are possible, however the iteration logic was quite confusing to write (I say this as someone who grew up writing C, C#, and Java)

When benchmarking the branches, I first ran the combined tests for just the 4 functions on a 100k element list (that's the first column). That didn't provide a clear enough picture, so then I isolated the tests for each function (that's the remaining columns). For take and drop, there were multiple tests doing similar work, so I pared it down to just one test taking or dropping 99,999 elements from a 100k element list.

branch	All 4	any	repeat	take n-1	drop n-1
master	3.15s	1.0s	1.4s	1.4s	1.1s
foldn	4.0s	n/a	1.35s	1.65s	1.15s
foldlWhile	4.15s	1.05s	1.45s**	1.6s	1.3s
loopUntil	4.65s	1.4s	1.55s	1.65s	1.35s

[jonathanhefner]

I created three branches to show what looping could be like for any, repeat, take, and drop:

• Using foldn
  • any cannot be implement via foldn, and thus is not ported
• Using foldlWhile
  • repeat cannot be implemented via foldlWhile, but a safe, less-efficient, pure-Elm version is included for consideration (note the ** in the benchmark data below)
• Using loopUntil
  • all 4 functions are possible, however the iteration logic was quite confusing to write (I say this as someone who grew up writing C, C#, and Java)

When benchmarking the branches, I first ran the combined tests for just the 4 functions on a 100k element list (that's the first column). That didn't provide a clear enough picture, so then I isolated the tests for each function (that's the remaining columns). For take and drop, there were multiple tests doing similar work, so I pared it down to just one test taking or dropping 99,999 elements from a 100k element list.

branch	All 4	any	repeat	take n-1	drop n-1
master	3.15s	1.0s	1.4s	1.4s	1.1s
foldn	4.0s	n/a	1.35s	1.65s	1.15s
foldlWhile	4.15s	1.05s	1.45s**	1.6s	1.3s
loopUntil	4.65s	1.4s	1.55s	1.65s	1.35s