Julia: NaN in -Inf:Inf

Created on 18 May 2017 · 10Comments · Source: JuliaLang/julia

I would like to have a infinite range, the code do not return error, but have a NaN. This can easily create bugs in program.

julia> a = -Inf:Inf
-Inf:1.0:NaN

versioninfo()
Julia Version 0.5.1-pre+31
Commit 6a1e339* (2016-11-17 17:50 UTC)
Platform Info:
  System: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz
  WORD_SIZE: 64
  BLAS: libmkl_rt
  LAPACK: libmkl_rt
  LIBM: libimf
  LLVM: libLLVM-3.7.1 (ORCJIT, haswell)

help wanted

Source

jingpengw

Most helpful comment

Maybe you are looking for an infinite interval, rather than a range.
This is available in e.g. the IntervalArithmetic.jl package:

julia> using IntervalArithmetic

julia> X = -∞..∞   # .. constructs intervals of real numbers
[-∞, ∞]

julia> 1.5e15 ∈ X
true

julia> typeof(X)
IntervalArithmetic.Interval{Float64}

dpsanders on 18 May 2017

👍6

All 10 comments

This should probably be an error.

What do you expect your range to do?

simonbyrne on 18 May 2017

I would expect an infinity range as an initialization.
the expected output should be exactly: -Inf:1.0:Inf

jingpengw on 18 May 2017

What would that mean though? How long would it be? What values would you
get when iterating over it?

On 18 May 2017 4:51 pm, "Jingpeng Wu" notifications@github.com wrote:

I would expect an infinity range as an initialization.
the expected output should be exactly: -Inf:Inf

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/JuliaLang/julia/issues/21941#issuecomment-302446949,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABnRaafRQBtC4YaAq-eZBDMQTjLv1w8Xks5r7GjygaJpZM4Nfa7K
.

simonbyrne on 18 May 2017

These should be an error on construction. There's no sensible meaning to such ranges.

StefanKarpinski on 18 May 2017

👍3

Maybe you are looking for an infinite interval, rather than a range.
This is available in e.g. the IntervalArithmetic.jl package:

julia> using IntervalArithmetic

julia> X = -∞..∞   # .. constructs intervals of real numbers
[-∞, ∞]

julia> 1.5e15 ∈ X
true

julia> typeof(X)
IntervalArithmetic.Interval{Float64}

dpsanders on 18 May 2017

👍6

-Inf:Inf is an error since 0.5.1
Inf:Inf is an error on v0.6.0-rc1

Julia-0.5.0> -Inf:Inf
-Inf:1.0:NaN

Julia-0.5.1> -Inf:Inf
ERROR: InexactError()
 in trunc(::Type{Int64}, ::Float64) at .\float.jl:463
 in colon(::Float64, ::Float64, ::Float64) at .\range.jl:183
 in colon(::Float64, ::Float64) at .\range.jl:191



md5-9933ee6d569eb12697615ec87fa3e9d4



Julia-0.5.2> Inf:Inf
Inf:1.0:Inf



md5-9933ee6d569eb12697615ec87fa3e9d4



Julia-0.6.0-rc1> Inf:Inf
ERROR: InexactError()
Stacktrace:
 [1] trunc(::Type{Int64}, ::Float64) at .\float.jl:672
 [2] colon(::Float64, ::Float64, ::Float64) at .\twiceprecision.jl:156
 [3] colon(::Float64, ::Float64) at .\range.jl:67

GregPlowman on 19 May 2017

not that the 0.5.0 behavior is that useful, but that's a bit more of a behavior change between 0.5.0 and 0.5.1 than I usually would have preferred for backporting. any guesses which backport commit it was?

tkelman on 19 May 2017

20396 ??

GregPlowman on 20 May 2017

Someone is welcome to figure out a way to restore the pre-0.5.2 behavior without breaking what that change fixed but it does not seem important enough to me to warrant much effort.

StefanKarpinski on 20 May 2017

Looks like this be closed now? Or is there any work / discussion remaining?