Add takewhile, dropwhile to iterators

[o314]

Constating that python, ruby, rust, elixir all offer takewhile, dropwhile

language	take while	drop while
python	takewhile	dropwhile
ruby	take_while	drop_while
rust	take_while	skip_while
elixir	take_while	drop_while

And also dotnet, java, scala !

I think Julia should have them too.

IterTools.jl, and rust have been consulted.

[ararslan]

Lots of stylistic fixes needed but I agree that the functionality would be nice to have.

[ararslan]

Should rebase on master to get the whitespace fix. Also takewhile and dropwhile should be added to the Iterators submodule exports along with the other functions there. They need docstrings with a compat annotation and should be added to the manual with the other iterators.

[o314]

Thanks for the feedback and openness !
As stated, i have some work ahead for now, i will fix remaining parts asap,
tomorrow the better.

[o314]

Possible Enhancement

• optimize nested TakeWhile(TakeWhile()), DropWhile(DropWhile())
DropWhile(TakeWhile()), TakeWhile(DropWhile())

[JeffBezanson]

Code looks good now, thanks. Just needs news and docs.

[o314]

news update

@ new library functions

• takewhile and dropwhile have been added to the Iterators submodule ([Add takewhile, dropwhile to iterators #33437]).

[o314]

compat update

compat has been set to Julia 1.4

[ararslan]

Julia 1.3 is feature frozen, so the compatibility notice here should be set to 1.4. As for a NEWS item, I'd just say

* `takewhile` and `dropwhile` have been added to the Iterators submodule ([#33437]).

[o314]

I've updated the news and update sections according to your remarks.

[ararslan]

Looks like the NEWS file hasn't been updated.

Requested changes implemented

[o314]

A filter(bitvector,itr) has been evocated (think compress @ python).

A oneliner can do the thing.
(May be two oneliners)

# zip variant. clone of python compress
@inline Base.Iterators.filter(markers::Union{Vector{Bool},BitVector}, s) =
    (i for (m,i) in zip(markers,s) if m)

# enumerate variant. julia thing
@inline Base.Iterators.filter(markers::Union{Vector{Bool},BitVector}, s) =
    (i for (n,i) in enumerate(s) if markers[n])

Benches

Bench initialization

using Random
bs = bitrand(10_000);
is = rand(1:100,10_000)

Bench of filter (zip variant)

julia> @benchmark Iterators.filter(bs, is)
BenchmarkTools.Trial:
  memory estimate:  80 bytes
  allocs estimate:  4
  --------------
  minimum time:     35.092 ns (0.00% GC)
  median time:      39.066 ns (0.00% GC)
  mean time:        54.293 ns (23.81% GC)
  maximum time:     41.544 μs (99.83% GC)
  --------------
  samples:          10000
  evals/sample:     1000

Bench of filter (enumerate variant)

julia> @benchmark Iterators.filter(bs, is)
BenchmarkTools.Trial:
  memory estimate:  80 bytes
  allocs estimate:  4
  --------------
  minimum time:     31.782 ns (0.00% GC)
  median time:      35.423 ns (0.00% GC)
  mean time:        50.731 ns (25.69% GC)
  maximum time:     42.052 μs (99.82% GC)
  --------------
  samples:          10000
  evals/sample:     1000

The enumerate variant seems 10-15% faster but does not checkbounds!
I don't know if we should have something in Base for this.