reverse searchsorted is inconsistent on 1.3

[rafaqz]

julia> searchsortedfirst(10:-1:1, 9.5; rev=true)
1

julia> searchsortedfirst(collect(10:-1:1), 9.5; rev=true)
2

julia> searchsortedlast(10:-1:1, 9.5; rev=true)
2

julia> searchsortedlast(collect(10:-1:1), 9.5; rev=true)
1

The range behaviour seems like the correct version, but needless to say it's extremely confusing getting different answers for these.

searchsortedfirst(a, x; by=, lt=, rev=false)

Return the index of the first value in a greater than or equal to x,
according to the specified order. Return length(a) + 1 if x is greater than
all values in a. a is assumed to be sorted.

It gets worse if you use all floats, now both the range and the array are wrong:

julia> searchsortedfirst(10.0:-1.0:1.0, 9.5; rev=true)
2

julia> searchsortedfirst(collect(10.0:-1.0:1.0), 9.5; rev=true)
2

julia> searchsortedlast(10.0:-1.0:1.0, 9.5; rev=true)
1

julia> searchsortedlast(collect(10.0:-1.0:1.0), 9.5; rev=true)
1

julia> versioninfo()
Julia Version 1.3.0
Commit 46ce4d7933 (2019-11-26 06:09 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM) i5-5300U CPU @ 2.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.1 (ORCJIT, broadwell)
Environment:
  JULIA_NUM_THREADS = 3

[rfourquet]

The range behaviour seems like the correct version

Actually the array version looks correct to me.

[rafaqz]

10 is the first value greater than or equal to 9.5. Which has an index of 1.

9 has an index of 2, but its not greater or equal to 9.5.

So 1 is the correct answer for searchsortedfirst

[rfourquet]

10 is the first value greater than or equal to 9.5

Yes with the usual interpretation of "greater than", but with rev=true, "greater than" means <=, so, in 10:-1:1, 9 is the first value <= 9.5.

[rafaqz]

That's extremely confusing! The doc should be rewritten, it's probably the reason this is broken.

It really makes it seem like search runs backwards until it finds a number greater than x.

[rfourquet]

Sure the docs should be inproved, the meaning of rev, by and lt is underspecified in the docstring. In that of searchsorted, it's a bit improved (at least they are mentioned), "... according to the order specified by the by, lt and rev keywords", but you have to know from somewhere else what they mean exactly, it's explained in the docstring of sort!, so there should be at least a link to that function to point where to find the information.