ProgressMeter Deprecation Warning and nothing comparison

[dylan-asmar]

• Updated the calls to Progress in line with the deprecation warning
• Changed comparisons to nothing to use === vs ==

Closes #533

[zsunberg]

Just out of curiosity, what is the advantage of === for nothing? Is there a link?

[dylan-asmar]

I thought there was a reference in the performance section of the Julia documentation, but I can't seem to find it.

I have run into an issue once where == wasn't checking against nothing correctly because a custom type == operator wasn't implemented 100% correctly (my fault). However, ever since then, I have defaulted to === for a few reasons:

• Efficiency: === is generally more efficient than == because it simply compares the memory addresses of the objects rather than checking for value equality. nothing I believe is a singleton and its identity is unique and constant.
• Clarity and Safety: Using === with nothing ensures that you are specifically checking for the absence of a value, rather than a value that might be considered equivalent to nothing. This can prevent unexpected behavior in cases where an object's value comparison (==) to nothing might be true even though the object isn't actually nothing. This is probably niche though.
• Avoiding Errors with Custom Types: If a custom type doesn't properly implement the == operator, comparing an instance of that type to nothing using == might result in errors or incorrect behavior. Since === does not rely on the potentially unimplemented or incorrectly implemented == method, it avoids this issue. This might be less of a thing in the last few versions of Julia...but in the back of my mind.

The Julia implementation of isnothing changed to just simply be === around 1.7 I think. I think there are some compile benefits of === as well...but I am not as certain about that one.

julia> x = [1, 3, 4];

julia> @benchmark x == nothing
BenchmarkTools.Trial: 10000 samples with 1000 evaluations.
 Range (min … max):  3.041 ns … 12.125 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     3.125 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   3.126 ns ±  0.214 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

             ▆           █          ▃           ▂            ▁
  █▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁█ █
  3.04 ns      Histogram: log(frequency) by time     3.25 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark x === nothing
BenchmarkTools.Trial: 10000 samples with 1000 evaluations.
 Range (min … max):  1.500 ns … 6.292 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     1.583 ns             ┊ GC (median):    0.00%
 Time  (mean ± σ):   1.569 ns ± 0.059 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▁            ▆▇             █             ▃               ▁
  █▁▁▁▁▁▁▁▁▁▁▁▁██▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁█▁▁▁▁▁▁▁▁▁▁▁▁▁█ █
  1.5 ns      Histogram: log(frequency) by time     1.67 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark isnothing(x)
BenchmarkTools.Trial: 10000 samples with 1000 evaluations.
 Range (min … max):  1.458 ns … 16.417 ns  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     1.583 ns              ┊ GC (median):    0.00%
 Time  (mean ± σ):   1.579 ns ±  0.172 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

          ▁       █       █▇       ▅       ▂       ▂         ▂
  ▃▁▁▁▁▁▁▁█▁▁▁▁▁▁▁█▁▁▁▁▁▁▁██▁▁▁▁▁▁▁█▁▁▁▁▁▁▁█▁▁▁▁▁▁▁█▁▁▁▁▁▁▁█ █
  1.46 ns      Histogram: log(frequency) by time     1.75 ns <

 Memory estimate: 0 bytes, allocs estimate: 0.

[zsunberg]

wow, thanks for the thorough explanation! But actually, why aren't we using isnothing?

[dylan-asmar]

I changed the code to use isnothing to begin with. Then changed it to === to keep the code similar while still achieving the same result. I think we should probably use isnothing (at least in the future).