@sprintf throws a bound error - but only on some inputs.

[sbromberger]

Perhaps related to #22126 ?

julia> @sprintf("%.330f", 2)
"2.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000"

julia> @sprintf("%.330f", 2.1)
ERROR: BoundsError: attempt to access 326-element Array{UInt8,1} at index [327]
Stacktrace:
 [1] bignumdtoa(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/bignum.jl:68
 [2] grisu(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/grisu.jl:51
 [3] fix_dec(::Float64, ::Int64) at ./printf.jl:953
 [4] fix_dec(::Base.AbstractIOBuffer{Array{UInt8,1}}, ::Float64, ::String, ::Int64, ::Int64, ::Char) at ./printf.jl:801

julia> versioninfo()
Julia Version 0.7.0-DEV.373
Commit 82e9638225 (2017-05-28 01:51 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin16.5.0)
  CPU: Intel(R) Core(TM) i5-6267U CPU @ 2.90GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, skylake)

[omus]

Issue also exists on 0.5 and 0.6.

julia> @sprintf("%.325f", nextfloat(0.0))
"0.0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000049"

julia> @sprintf("%.325f", 1.0)
ERROR: BoundsError: attempt to access 326-element Array{UInt8,1} at index [327]
Stacktrace:
 [1] bignumdtoa(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/bignum.jl:68
 [2] grisu(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/grisu.jl:51
 [3] fix_dec(::Float64, ::Int64) at ./printf.jl:953
 [4] fix_dec(::Base.AbstractIOBuffer{Array{UInt8,1}}, ::Float64, ::String, ::Int64, ::Int64, ::Char) at ./printf.jl:801

The exception can occur with printing less decimal places when the input is large:

julia> v = prevfloat(typemax(Float64))
1.7976931348623157e308

julia> @sprintf("%.16f", v)
"179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.0000000000000000"

julia> @sprintf("%.17f", v)
ERROR: BoundsError: attempt to access 326-element Array{UInt8,1} at index [327]
Stacktrace:
 [1] bignumdtoa(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/bignum.jl:68
 [2] grisu(::Float64, ::Int64, ::Int64, ::Array{UInt8,1}, ::Array{Base.Grisu.Bignums.Bignum,1}) at ./grisu/grisu.jl:51
 [3] fix_dec(::Float64, ::Int64) at ./printf.jl:953
 [4] fix_dec(::Base.AbstractIOBuffer{Array{UInt8,1}}, ::Float64, ::String, ::Int64, ::Int64, ::Char) at ./printf.jl:801

[vtjnash]

Off by one error? looks like

julia/base/printf.jl

Line 952 in ecc3a81

if n > length(DIGITS)-1; n = length(DIGITS)-1; end

is supposed to prevent this, but also seems wrong that grisu is trying to append a nul-byte

[sbromberger]

This is still an issue in 1.1.0

[StefanKarpinski]

@quinnj, I think you're the most recent person to have dived into the Printf code, would it be possible for you to take a look at this?

[quinnj]

So, I somehow totally missed that I was pinged here; sorry about that. But, as fate would have it, I'm going over old numeric display issues in light of #32799 and it looks like this would be solved by that PR. Now, we still need to work out the exact details of @printf + new ryu algorithm details, but essentially calling Ryu.writefixed or Ryu.writeexp with all these examples works in that PR just fine.

[quinnj]

Fixed in #32859