Support for size values (640 KB) and numbers with unit in general

[metopa]

This PR adds infrastructure for accepting inputs in form <number> <unit>. This covers things like size inputs, durations, and many more. <suffix> may or may not be separated from the number with space, must contain only letters and optionally can be matched case-insensitively.

CLI::AsNumberWithUnit takes a mapping from unit to factor. Transformation for durations may look like this:

map<string, double> m = {{"ns", 1e-9}, {"us", 1e-6}, {"ms", 1e-3}, {"s", 1}, {"m", 60}};
app.add_option("-d", duration)->transform(CLI::AsNumberWithUnit(mapping));

Transformation for size is already provided:

app.add_option("-s", size)->transform(CLI::AsSizeValue(true));

[codecov]

Codecov Report

Merging #253 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #253   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files          12     12           
  Lines        2566   2665   +99     
=====================================
+ Hits         2566   2665   +99

Impacted Files	Coverage Δ
include/CLI/StringTools.hpp	100% <100%> (ø)	⬆️
include/CLI/Validators.hpp	100% <100%> (ø)	⬆️

---

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 be8a08f...f6db0a7. Read the comment docs.

[codecov]

Codecov Report

Merging #253 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #253   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files          12     12           
  Lines        2782   2879   +97     
=====================================
+ Hits         2782   2879   +97

Impacted Files	Coverage Δ
include/CLI/StringTools.hpp	100% <100%> (ø)	⬆️
include/CLI/Validators.hpp	100% <100%> (ø)	⬆️

---

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 10e4b07...6c57111. Read the comment docs.

[henryiii]

I like the general version and adding an explicit one for common uses. I will probably break out the more specialized transforms (at least the size one here, IP4, etc.) into a separate utilities file so that single header file users won't have a growing file each release. It would be available for either normal or single header file use, but a separate file. Just a heads up for my plans, does not affect this PR. Will hopefully help with #194.

[metopa]

@henryiii Could you please take a look at the building error on GCC 4.7? I don't have it on my machine and the error is a bit of WTF

[henryiii]

I'll try it out in a 4.7 docker container later today and see if I can see what it's complaining about. GCC 4.7 is a pain, I'll probably be dropping it in a release or two. This looks fixable without too much work, I think.

[henryiii]

Note: This is on my todo list but I'm traveling a bit unexpectedly to cover for someone at a conference, so eventually will get to this!

[metopa]

@henryiii Sure thing, thank you a lot!

[henryiii]

I believe you are seeing this bug in GCC: https://stackoverflow.com/a/12086873/2402816. I believe you are only putting these methods in the class for organization? There are no members. We can either move them to put this in 1.8 or wait till 1.9 and drop GCC 4.7 support.

[metopa]

Tricky! Thank you for figuring this out.

[metopa]

After returning back to this after some time I got an idea. Should I rename SuffixedNumber to smth like NumberWithUnit and generally use "unit" term in docs? Right now it makes more sense to me.

[henryiii]

@phlptp , what do you think on the naming question? I like both. My physics backround leans toward Unit, but Suffix is a little more general

[phlptp]

I think I would have a preference toward "Unit". Given a significant chunk of my time recently has been on a runtime units library I may be bit biased though.

[henryiii]

@metopa Do you want to wait and target 1.9 instead, or try to get this in to 1.8?

[metopa]

@henryiii I've just updated it

[henryiii]

Thanks!