Add parsing functionality (possible in a new package)

[LaurentRDC]

One source of problems when dealing with units at the boundary between the user and a program, or between programs. However, there is currently no way to parse Quantity and Unit from unstructured text.

@bjornbm Would you be OK if this repository contained two packages, dimensional and possibly dimensional-parsing?

[bjornbm]

Could you show me an example of what you have in mind as API/usage?

Maybe take a look At @dmcclean’s
https://github.com/dmcclean/dimensional-parsers to check whether it seems complete and/or salvageable.

@dmcclean maybe you can comment?

[LaurentRDC]

Thanks for linking previous work.

Ideally, the parsing interface would be more type-safe than dimensional-parsers, avoiding dynamic quantities and units. Something like:

parseQuantity :: Text -> Either Error (Quantity d a)

where the user must specify what dimension d is expected, and parseQuantity failing if the input Text doesn't represent this dimension.

I haven't thought how that would work in practice though

[LaurentRDC]

Related issue: #8

[bjornbm]

Ideally, the parsing interface would be more type-safe than dimensional-parsers, avoiding dynamic quantities and units.

It seems it could make sense to parse to a DynQuantity and then fmap promoteQuantity to give you the type safety?

Convoluted (and untested) use case for DynQuantity parsing:

getVelocity :: Num a => Text -> Text -> Either Error (Power a)
getVelocity input1 input2 = do
   x <- parseDynQuantity input1
   y <- parseDynQuantity input2
   promoteQuantity (x * y)   -- see note

This allows the inputs to be a force and a velocity, or a torque and an angular velocity, or a current and a voltage.

Note: I took the liberty to disregard that promoteQuantity returns a Maybe – I think there would be no objection to changing this to a Monad m.

Anyway, I don't feel very strongly about this.

Maybe the first step is to determine your level of ambition… as indicated in #8 parsing units can get hairy.

I can also see how it could be nice to have a parse/read capability with lighter dependencies than dimensional-parsers for less demanding cases. If very light/no dependencies it could maybe be included in the dimensional package.