Tag Unitful v0.0.2 [https://github.com/ajkeller34/Unitful.jl]

[ajkeller34]

I have fixed a nasty promotion bug in v0.0.1 where [1.0m, 2.0m] ./ 3 would throw a Unitful.DimensionError(). I'm inclined to work on improving promotion—a few things still don't work quite like unitless numbers, for instance [1m, 2m]./3 will throw an InexactError()—but I'll probably wait to fix that until JuliaLang/julia#18465 is patched so that I know which @inferred errors are my fault when I play around with promote_op. Unitful v0.0.2 also adds a new unit conversion syntax.

Is there a summary of the state of promote_op as of v0.5, as well as how to write code that is amenable to type inference? I'm sorry if that's a bit vague. I feel my promote_op functions could be simpler and better but I have tried writing them other ways and I get @inferred failures. If I remove the promote_op definitions entirely then I get a lot of @inferred failures, although I have some vague notion that promote_op is being phased out... it's not even in the 0.5 manual anymore.

[tkelman]

@pabloferz why did you remove promote_op from the docs?

[pabloferz]

promote_op was documented as a possible way of fixing some limitations with the promotion system but the whole thing changed before 0.5 was released and it wasn't supposed to be a method to be preferably override outside Base. I deleted for that reason, I guess I could have changed the section by one explaining the changes instead.

[pabloferz]

@ajkeller34 If the problem was related to using promote_op with generated functions as you pointed out in https://groups.google.com/forum/#!topic/julia-users/EAoanGDSc6g, then the problem might be on the use of generated functions and not in promote_op itself. I just found out recently that are some restrictions when using generated functions, some of these are stated in the manual.