Change sentinel in find(first|next|prev|last) to `nothing` #25472

timholy · 2018-01-09T19:15:53Z

Closes JuliaLang/Juleps#47. One could discuss returning a NotFound value instead.

nalimilan · 2018-01-09T19:17:51Z

base/array.jl

@@ -1547,15 +1546,15 @@ julia> findfirst(A)
 2

 julia> findfirst(falses(3))
-0


Maybe do findfirst(...) == nothing giving true, that will be more explicit? Same below.

nalimilan · 2018-01-09T19:20:44Z

base/array.jl

 ```
 """
 findlast(testf::Function, A) = findprev(testf, A, endof(A))

+zero_sentinel(i) = i === nothing ? 0 : i


FWIW, this is just coalesce(i, 0), so probably not worth defining a new function?

There are some bootstrap issues (can't use coalesce in files in Core), but yes overall this is better.

nalimilan · 2018-01-09T20:10:03Z

base/bitset.jl

@@ -68,14 +68,14 @@ function _bits_findnext(b::Bits, start::Int)
 # start is 0-based
 # @assert start >= 0
 _div64(start) + 1 > length(b) && return -1
- unsafe_bitfindnext(b, start+1) - 1
+ zero_sentinel(unsafe_bitfindnext(b, start+1)) - 1


Why not keep unsafe_bitfindnext returning 0 for no match, since it's internal? Wouldn't that involve less changes?

It's one of each; I decided to change it for greater consistency. But indeed I followed this advice for other internal methods.

nalimilan · 2018-01-09T20:26:33Z

base/strings/search.jl

 throw(BoundsError(s, i))
 end
 @inbounds isvalid(s, i) || string_index_err(s, i)
 c = pred.x
- c ≤ '\x7f' && return _search(s, c % UInt8, i)
+ c ≤ '\x7f' && return nothing_sentinel(_search(s, c % UInt8, i))


~~Is it really useful to change _search too, given that it's internal?~~

EDIT: actually, I read it backwards. Shouldn't _search be changed too? It appears to be always used with nothing_sentinel (which could be removed since it's only used in this file AFAICT). BTW, looks like you missed this method:

julia/base/strings/search.jl

Line 265 in bf424d7

findnext(t::AbstractString, s::AbstractString, i::Integer) = _search(s, t, i)

Doesn't that one always return a range? _search caused me more headaches than any other here, so I could be misunderstanding.

Ah, right. Though my first remark about changing _search still holds I think.

Yeah, but I don't feel like doing it 😄. In truth this is what I tried first, spent half a day just trying to get Julia to build, and then punted and started from scratch. It's possible I was stymied by something very trivial, but at this stage even half a day is not worth it for an internal interface that can be changed later.

I think the problem is that the range-returning version calls the scalar-returning version and I don't know this part of Julia well enough to untangle it without investing more time than I can afford right now.

I know that feeling too, that code is complex.

Let's leave it that way, but please move nothing_sentinel to this file since it isn't used elsewhere.

nalimilan · 2018-01-09T20:28:55Z

test/bitarray.jl

@@ -1114,7 +1114,7 @@ timesofar("nnz&find")
 b1 = trues(v1)
 b2 = falses(v1)
 for i = 1:v1
- @test findprev(b1, i) == findprev(b1, true, i) == findprev(identity, b1, i)


Why not test findprev(equalto(true), b1, i)?

The same codepath is covered by identity.

I don't think so, given findnext(pred::EqualTo, B::BitArray, start::Integer) now exists.

nalimilan · 2018-01-09T20:32:04Z

base/combinatorics.jl

@@ -125,7 +125,7 @@ function invpermute!!(a, p::AbstractVector{<:Integer})
 count = 0
 start = 0
 while count < length(a)
- start = findnext(!iszero, p, start+1)
+ start = findnext(!iszero, p, start+1)::Int


Are we actually sure all AbstractArray implementations will return an Int? You could also try using notnothing(...), which throws when nothing is passed to it, which should allow the compiler to assert the type (need to check that).

BTW, how did you find the places where this assertion can be added?

Anyplace where the code didn't seem to check for a 0 return but then indexed it, I assumed it was guaranteed to succeed.

alyst · 2018-01-09T22:01:59Z

base/bitarray.jl

@@ -1540,18 +1541,18 @@ function unsafe_bitfindprev(Bc::Vector{UInt64}, start::Integer)
 end
 end
 end
- return 0
+ return nothing
 end

 # returns the index of the previous non-zero element, or 0 if all zeros


"or nothing"

alyst · 2018-01-09T22:05:16Z

base/bitarray.jl

@@ -1459,13 +1459,13 @@ function unsafe_bitfindnext(Bc::Vector{UInt64}, start::Integer)
 end
 end
 end
- return 0
+ return nothing
 end

 # returns the index of the next non-zero element, or 0 if all zeros


"or nothing"

timholy · 2018-01-10T02:55:50Z

New version.

andyferris · 2018-01-10T10:45:08Z

This seems a safer than 0 and is infinitely better for offset arrays! :)

This was said in another forum but I’ll say it one time more - this seems like the kind of condition that can only be dealt with properly when the caller explicitly checks the result. To me this is a textbook case of using something like Nullable or the much discussed Some. Using some wrapped (and then explicitly unrwapped, by the caller) return type would allow the API to extend to other containers like AbstractDicts which allow anything as keys.

Just thoughts - don’t let this slow progress on AbstractArray.

nalimilan · 2018-01-10T10:58:23Z

@andyferris Better continue the discussion about Some at JuliaLang/Juleps#47. The problem with that solution (which is indeed more correct in general) is that it would be quite inconvenient when working with arrays, which are a major use case for find*.

andyferris · 2018-01-10T13:42:00Z

@nalimilan On one hand, I completely agree, let's just move forward with a simple thing which is correct for arrays (including offset arrays).

The problem with that solution (which is indeed more correct in general) is that it would be quite inconvenient when working with arrays

On the other hand, the underlying thing which bothers me here and in JuliaLang/Juleps#47 is that we discuss this "inconvenience" in a strange way - even for arrays, it is absolutely necessary for the caller to do something to deal with the fact that the index might not be found. Whichever way we disguise it (with sentinel values, union types, wrapping types, etc), the caller really needs to check some boolean condition whether a matching index was found or not. My judgement is that wrapping the return in a Nullable or Some is the best way to make the caller explicitly aware that this simply has to happen (or else an error is thrown). The union approach might mean that a suite of unit tests could pass if the author doesn't remember to cover the general case you might see with unknown, production data where no index is found.

But yes, perhaps this is something better discussed elsewhere! Like I tried to say earlier, I only want to write this for the record on JuliaLang/julia for future reference (say, for v2.0 development), rather than do anything to block this PR (which is clearly better than the status quo, and therefore quite acceptable).

timholy · 2018-01-10T14:01:03Z

OK, barring more objections I think this is ready. Let's first get a performance test:

@nanosoldier runbenchmarks(ALL, vs=":master")

nalimilan · 2018-01-10T14:11:59Z

Thanks Tim! BTW, I think you missed my question at #24774 (comment), which is now hidden.

nanosoldier · 2018-01-10T14:30:18Z

Something went wrong when running your job:

NanosoldierError: failed to run benchmarks against primary commit: failed process: Process(`sudo cset shield -e su nanosoldier -- -c ./benchscript.sh`, ProcessExited(1)) [1]

Logs and partial data can be found here
cc @ararslan

ararslan · 2018-01-10T19:40:04Z

Nanosoldier is hitting a MethodError related to the new return type from find. Perhaps something that will need to be addressed in BaseBenchmarks before Nanosoldier will work here.

nalimilan · 2018-01-10T19:46:49Z

Unfortunately, it's going to be difficult to support both the new and the old API. I guess we'll need to add a Compat.find function returning nothing even on 0.6.

timholy · 2018-01-11T02:58:35Z

I see slight regressions in

julia> @btime perf_findnext($x)    # length-1000 Bool vector
  1.234 μs (0 allocations: 0 bytes)

vs

julia> @btime perf_findnext($x)
  837.929 ns (0 allocations: 0 bytes)

There are a couple of apparent big regressions in what appears to be completely unrelated code (e.g., a ccall inside of resize!), I'm pretty puzzled about that, but it also doesn't seem to be very reproducible.

timholy · 2018-01-12T20:57:30Z

I'm not seeing any complaints about that regression. Any objections to merging? If not, I'll fix the conflicts.

garrison · 2018-01-20T15:06:55Z

I just noticed that indexin still uses 0 to denote missing values. I previously proposed some indexin improvements in #23845, but I am not sure how to handle this. Note that the function, with or without #23845, fails when an array with custom indices is passed as the second argument. A (relatively extensive) list of options I see:

Leave it as is (perhaps eventually incorporating Make indexin first argument accept any iterable #23845). But then there is pretty serious inconsistency in the API, as I know of no other function that still uses 0 to denote missing data.
Use nothing as the sentinel for indexin, which would allow expanding its usage to arrays with custom indices. But then the most common case of returning an array means that each element is boxed IIUC, as the data type of an element is Union{Nothing,Int}.
Error on missing data. This will be most efficient in the use cases where missing data is not anticipated, and will also allow using arrays with custom indices. But it is a break from current behavior.
Some combination of the above two options, based on a keyword argument (or a similar mechanism, if that were to break inference).
Remove indexin from Base altogether. It is not used in Base, and it seems to be used somewhat rarely throughout the package ecosystem (complete list of uses here). Given that there are multiple behaviors one might expect for missing data, perhaps it is better for each individual user to write this few-line function using the error handling that they expect. The (only) downside (that I can think of) is that this removes an interface that might be nice for some data structures, which would be missed by people rolling their own implementations. For instance, UniqueVectors implements this method in a way that skips the step of building the lookup dictionary (since it already has access to such a Dict). But this seems a relatively specialized use case, not a compelling case for keeping this function in Base.

My overall vote (at the moment) is the last option: removing indexin from Base. I think a decent case could also be made for the third or fourth options, but even so I doubt the function actually belongs in Base, rather it should be in a (standard?) library somewhere.

nalimilan · 2018-01-20T15:30:04Z

I had noticed that too, and I think we should return nothing instead of 0 for consistency with other functions. Union{Nothing,Int} arrays use an efficient memory representation on 0.7 already, and since we return nothing with all other find* functions, we already assume performance is OK (or can be fixed).

andreasnoack · 2018-02-15T09:45:00Z

This needs a NEWS entry

ararslan · 2018-02-15T19:03:19Z

#26067

timholy mentioned this pull request Jan 9, 2018

Deprecate a couple more find(next|prev) methods, and remove ::Integer on start #25468

Closed

ararslan added the domain:search & find The find* family of functions label Jan 9, 2018

nalimilan reviewed Jan 9, 2018

View reviewed changes

alyst reviewed Jan 9, 2018

View reviewed changes

timholy force-pushed the teh/find_sentinel branch from bf424d7 to 8deab51 Compare January 10, 2018 02:55

nalimilan mentioned this pull request Jan 10, 2018

Unifying search & find functions #10593

Closed

timholy force-pushed the teh/find_sentinel branch 2 times, most recently from f6d7053 to d195c91 Compare January 10, 2018 13:57

nalimilan approved these changes Jan 10, 2018

View reviewed changes

timholy added 2 commits January 13, 2018 05:50

Remove deprecated findnext/findprev signatures

338110c

Change sentinel in find(first|next|prev|last) to nothing

b6bd419

timholy force-pushed the teh/find_sentinel branch from d195c91 to b6bd419 Compare January 13, 2018 11:57

timholy merged commit 5c3f580 into master Jan 13, 2018

timholy deleted the teh/find_sentinel branch January 13, 2018 21:17

samoconnor added a commit to JuliaWeb/HTTP.jl that referenced this pull request Jan 17, 2018

Quick hack to work around JuliaLang/julia#25472

f7cbe5f

garrison mentioned this pull request Jan 20, 2018

RFC: Change indexin sentinel to nothing #25662

Merged

andreasnoack mentioned this pull request Feb 15, 2018

Clean up the log JuliaLang/METADATA.jl#13367

Merged

andreasnoack added the needs news A NEWS entry is required for this change label Feb 15, 2018

ararslan mentioned this pull request Feb 15, 2018

Add a NEWS entry for find* sentinel change #26067

Merged

KristofferC removed the needs news A NEWS entry is required for this change label Nov 13, 2018

c42f pushed a commit to JuliaWeb/URIs.jl that referenced this pull request Oct 9, 2020

Quick hack to work around JuliaLang/julia#25472

c36e65f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Change sentinel in find(first|next|prev|last) to `nothing` #25472

Change sentinel in find(first|next|prev|last) to `nothing` #25472

timholy commented Jan 9, 2018

nalimilan Jan 9, 2018

nalimilan Jan 9, 2018

timholy Jan 10, 2018 •

edited

Loading

nalimilan Jan 9, 2018

timholy Jan 10, 2018

nalimilan Jan 9, 2018 •

edited

Loading

timholy Jan 10, 2018

nalimilan Jan 10, 2018

timholy Jan 10, 2018

nalimilan Jan 10, 2018

nalimilan Jan 9, 2018

timholy Jan 10, 2018

nalimilan Jan 10, 2018

nalimilan Jan 9, 2018

timholy Jan 10, 2018

alyst Jan 9, 2018

alyst Jan 9, 2018

timholy commented Jan 10, 2018

andyferris commented Jan 10, 2018

nalimilan commented Jan 10, 2018

andyferris commented Jan 10, 2018

timholy commented Jan 10, 2018

nalimilan commented Jan 10, 2018

nanosoldier commented Jan 10, 2018

ararslan commented Jan 10, 2018

nalimilan commented Jan 10, 2018

timholy commented Jan 11, 2018

timholy commented Jan 12, 2018

garrison commented Jan 20, 2018 •

edited

Loading

nalimilan commented Jan 20, 2018

andreasnoack commented Feb 15, 2018

ararslan commented Feb 15, 2018

Change sentinel in find(first|next|prev|last) to nothing #25472

Change sentinel in find(first|next|prev|last) to nothing #25472

Conversation

timholy commented Jan 9, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timholy Jan 10, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nalimilan Jan 9, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

timholy commented Jan 10, 2018

andyferris commented Jan 10, 2018

nalimilan commented Jan 10, 2018

andyferris commented Jan 10, 2018

timholy commented Jan 10, 2018

nalimilan commented Jan 10, 2018

nanosoldier commented Jan 10, 2018

ararslan commented Jan 10, 2018

nalimilan commented Jan 10, 2018

timholy commented Jan 11, 2018

timholy commented Jan 12, 2018

garrison commented Jan 20, 2018 • edited Loading

nalimilan commented Jan 20, 2018

andreasnoack commented Feb 15, 2018

ararslan commented Feb 15, 2018

Change sentinel in find(first|next|prev|last) to `nothing` #25472

Change sentinel in find(first|next|prev|last) to `nothing` #25472

timholy Jan 10, 2018 •

edited

Loading

nalimilan Jan 9, 2018 •

edited

Loading

garrison commented Jan 20, 2018 •

edited

Loading