Make Float extensible

[loldrup]

If the Float type was extensible, a language user could easily create distinct types for eg. temperatures, distances, weights.
As of current, such values all have to be merely Floats, giving us no type safety against mixing them up when doing computations.

PS. I know that it's possible to define one's own Floaty type, but I also know that it is quite cumbersome. Simply extending Float would give me all that Floatyness for free.
PPS. F# has ability to mingle Floats of various units. Eg. if I divide a value that has a 'lengthy' unit (eg. meters) by a value that has a 'timy unit' (eg. seconds), I get a length per time type. If I multiply a length by a length, I get a 'length squared' type. If I divide a length by a length, I get a scalar value. That's all quite cool. Could Ceylon support something similar, and would that feature be in good harmony with the ability to extend the Float type?

[gavinking]

If by "extending", you mean "subclassing", that would mean we can't erase the type Float to float in the JVM.

[loldrup]

Yes I meant subclassing.

An, most likely, ignorant question: Why would it no longer be possible to type erase Float to float?

[RossTate]

Heads up: To achieve the F# feature you describe, it’ll take a lot more than just subclassing.

…

On Sat, Apr 7, 2018 at 7:17 PM loldrup ***@***.***> wrote: Yes I meant subclassing. An, most likely, ignorant question: Why would it no longer be possible to type erase Float to float? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#7348 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABFj24Yyw-HoeFjeElQZ-uaSC-PC8mv7ks5tmUkggaJpZM4TLRJF> .

[gavinking]

Why would it no longer be possible to type erase Float to float?

Because I could assign a value of type Length to a variable of type Float, but I would lose the concrete type.

[gavinking]

To achieve the F# feature you describe, it’ll take a lot more than just subclassing.

Right, a proper units-of-measure system is really surprisingly complex. For example, F# lets you write numerical functions that abstract over the unit of measure.

[xkr47]

To make things better, simply embedding a Float in a Length class like

shared class Length(shared variable Float val) {}

will take about the same amount of memory as a subclass of Float since the compiler can use the primitive type float to store the value. The downside of course is that you have to refer to the value as:

Length len = Length(5);
print(len.val);

[loldrup]

Thanks for the tip! Kind regardsJon Loldrup 28. Aug 2018 21:07 af notifications@github.com <mailto:notifications@github.com>:

…

To make things better, simply embedding a Float in a Length class like shared> > class> > Length> (> shared> > variable> > Float> > val> ) {} will take about the same amount of memory as a subclass of Float since the compiler can use the primitive type > float> to store the value. The downside of course is that you have to refer to the value as: Length> > len> = > Length> (> 5> );> print> (> len> .> val> ); — You are receiving this because you modified the open/close state. Reply to this email directly, > view it on GitHub <#7348 (comment)>> , or > mute the thread <https://github.com/notifications/unsubscribe-auth/AAP_NBOlxYyekupV1CXqVn2HDzttF1tHks5uVZUCgaJpZM4TLRJF>> .