Note: I'm by no means done, I wanted to get feedback from you before doing the grunt work. :-)

Okay, I've gotten farther on this. The concept works and seems to scale. I've converted all the indicators and am about 2/3 of the way through the tests.

The major downside is literals in tests. Integers generally do the right thing but floats don't, so I had to write a lit! (literal) macro and use it on literals. This adds a trivial tax on writing indicators (there are about 7-8 literals total) but a noticeable tax on writing tests, which use literals liberally. I still think this is a worthwhile endeavor and hope you'll accept it, but wanted to be transparent.

Okay @greyblake, this is done (with one caveat) and ready for review. I've detailed the overall approach above.

One thing that's not done is the benchmark, which requires rust_decimal to support rand. As it happens, they've just added that (three days ago; I did not do it), so I'm just waiting for a 1.24 release (which I have also politely inquired about).

EDIT: It turns out the rand implementation was partial; I've sent paupino/rust-decimal#519 to provide what the benchmark needs.

• Tests all pass both with and without the decimal feature.
• Examples only pass without decimal, and I've configured CI accordingly. It did not seem proper to me to clutter up your documentation / examples.

@greyblake This is now ready! :-)

@greyblake Ping! :-)

@lukesneeringer Pong. Sorry for delays, I barely have time these days because of personal reasons.