Show of Vector{Particles}

[mauro3]

The show method seems to to some rounding with particles in an array:

julia> [10000+i ± 0.5 for i=1:10:50]
5-element Array{Particles{Float64,2000},1}:
 10000.0 ± 0.5
 10000.0 ± 0.5
 10000.0 ± 0.5
 10000.0 ± 0.5
 10000.0 ± 0.5

julia> ans[1]
Particles{Float64,2000}
 10001.0 ± 0.5

Is this intentional? If so, then it should take into account the significant digits according to the uncertainty.

[baggepinnen]

It's intentional, but poorly implemented at the moment

MonteCarloMeasurements.jl/src/particles.jl

Line 87 in 3bd31a3

# TODO: be smart and select sig digits based on s

Should probably use printf instead of the round function with sigdigits keyword?

[mauro3]

Should it be different between arrays and scalars?

[baggepinnen]

The idea is to not overwhelm the user with too many digits, similar to the difference

julia> rand()
0.22107097572682943

julia> rand(3,3)
3×3 Array{Float64,2}:
 0.831273  0.407677   0.509738
 0.924296  0.0520808  0.925542
 0.17117   0.44741    0.224877

But I think sigdigits does not do what I wanted, I wanted something similar to

julia> @printf("%4.3g", 10001.0)
1e+04

Ideally, the number of digits to show should depend on the standard deviation of the sample. There is no point in showing many digits of the mean if they are not significant.

[mauro3]

deally, the number of digits to show should depend on the standard deviation of the sample. There is no point in showing many digits of the mean if they are not significant.

Sounds good. So, for instance:

julia> 10001.0578 ± 0.01
Particles{Float64,2000}
 10001.0 ± 0.01

should print as 10001.05 ± 0.01.

[baggepinnen]

Something like that, yes :)
With a upper bound on the number of digits to show in the array case.