To certain functions, add the fancy_exponents kwarg, which allows the user to toggle the "fancy exponents" setting in a thread-safe way

[DilumAluthge]

This pull request adds the fancy_exponents::Bool kwarg to several methods of the following functions:

• show
• showval
• showrep
• superscript

We use the fancy Unicode exponents if and only if fancy_exponents is true.

The fancy_exponents::Bool kwarg is optional. The default value is determined as follows:

1. If the user explicitly provides the fancy_exponents kwarg, we use that value.
2. Otherwise, if the user has set the UNITFUL_FANCY_EXPONENTS environment, and the lowercased value parses to a Bool, we use that value.
3. Otherwise, if the user has set the UNITFUL_FANCY_EXPONENTS environment variable, but the lowercased value does not parse to a Bool, we default to false.
4. Otherwise, we default to true on macOS and false otherwise.

Therefore, if the user does not provide the fancy_exponents kwarg, this PR preserves the existing behavior.

Motivation

Currently, the only way to control the behavior of the fancy Unicode exponents is by the UNITFUL_FANCY_EXPONENTS environment variable. Unfortunately, as far as I understand, there is no thread-safe way to modify the environment while calling a Julia function. (The withenv function is not thread-safe. The addenv function is thread-safe, but as far as I can tell, it only lets you run a Cmd.)

The fancy_exponents kwarg allows the user to turn fancy Unicode exponents on/off in a thread-safe way.

[codecov-commenter]

Codecov Report

Merging #445 (94330bd) into master (cc3adef) will decrease coverage by 0.03%.
The diff coverage is 87.50%.

❗ Current head 94330bd differs from pull request most recent head be954f1. Consider uploading reports for the commit be954f1 to get more accurate results

@@            Coverage Diff             @@
##           master     #445      +/-   ##
==========================================
- Coverage   83.84%   83.80%   -0.04%     
==========================================
  Files          16       16              
  Lines        1337     1334       -3     
==========================================
- Hits         1121     1118       -3     
  Misses        216      216              

Impacted Files	Coverage Δ
src/logarithm.jl	69.42% <0.00%> (-0.13%)	⬇️
src/display.jl	96.34% <95.45%> (+0.13%)	⬆️
src/user.jl	94.07% <0.00%> (-0.19%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cc3adef...be954f1. Read the comment docs.

[giordano]

I can't say I'm a fan of using environment variables for this. Why not using IO context instead?

[DilumAluthge]

Sure, that would work. I just didn't want to break the existing behavior of the package.

[DilumAluthge]

I.e. the package already uses the environment variable, so removing it would be a breaking change.

[giordano]

How using IO context would be different from using an environment variable? You don't break anything, you don't even add any extra arguments

[DilumAluthge]

Ah I think I see what you mean.

[DilumAluthge]

Well just to clarify, we'd use IO contexts instead of the kwarg. The UNITFUL_FANCY_EXPONENTS environment still has to stay as a fallback.

[giordano]

Yes, the extra argument is redundant since you can pass extra stuff already from the IO. I use IO context in Measurements.jl, see for example https://github.com/JuliaPhysics/Measurements.jl/blob/master/src/show.jl#L42 and the tests

[DilumAluthge]

Yes, the extra argument is redundant since you can pass extra stuff already from the IO. I use IO context in Measurements.jl, see for example https://github.com/JuliaPhysics/Measurements.jl/blob/master/src/show.jl#L42 and the tests

It's so much cleaner! I like it.

Closing this PR in favor of #446, which uses the IO context property.