Decimal type?

[codalogic]

Decimal would be a nice type to have, as often protocols don't want to have to deal with something like: 1e+0.

The problem is that it would be nice to be able to specify the number of decimal places, AND the range, without conflicting with the short float definitions.

Copying the idea for our int? types, for an example price type, defined as a decimal with 2 decimal places, and that must be positive, one possibility is:

"price" : decimal.2  0.0..,

The ABNF might look like:

decimal-type = decimal-kw [ decimal-places ] [ spaces float-range ]
decimal-places = "." pos-integer

We could possibly add this as an experimental type in an appendix of the I-D if we're not comfortable with it yet.

[anewton1998]

Can this be done as an annotation? That would be my preference.

[codalogic]

As in annotations on a new decimal type:

"price" : @{decimal-places 2} @{range 0.0..} decimal,
; or maybe:
"price" : @{dp 2} @{range 0.0..} @{exclude-min} decimal,

Or annotations on the existing float types:

"price" : @{decimal} @{decimal-places 2} @{range 0.0..} float,

?

Which makes me wonder whether we should change the primitive-rule ABNF from:

primitive-rule   = annotations primitive-def

to:

primitive-rule   = annotations primitive-def annotations

so we can have:

"price" : decimal @{dp 2} @{range 0.0..} @{exclude-min},

[anewton1998]

I was thinking annotations on the existing floats. And we wouldn't need an annotation for range, floats already have that.

@{decimals 2} float

or

@{decimals 2} 0.00..1.00

[codalogic]

I think the purists (mentioning no names) would say that a decimal is not a float! Is it the likes of 1.1 that can't be represented exactly as a float, but can be as a decimal? (https://stackoverflow.com/questions/9014303/comparing-float-and-double).

I'm not sure of the consequences of that for JCR, a JCR validator, or a spec that might use JCR.

Perhaps we should park this issue for a while.

[anewton1998]

Yeah, let's think about this some more.