Absolute Zero

[esabol]

This is a very, very trivial thing, but I was reading the README.md and saw your Celsius and Fahrenheit examples. It was surprising to me that the Celsius range started at -273.14. Absolute zero is -273.15 °C, of course. Did you intentionally make the lower range -273.14 because absolute zero is not physically achievable? But -273.141 °C is achievable (although difficult, I'm sure).

https://en.wikipedia.org/wiki/Absolute_zero

Quasi-related, have you given any thought to supporting a syntax for open-ended interval numerical ranges like (a,b] or even ]a,b[?

https://en.wikipedia.org/wiki/Interval_(mathematics)#Including_or_excluding_endpoints

[happy-barney]

Range will be interesting add-on, definitely worth of discussions, but beyond MVP, my feeling is even beyond milestone after mvp

[Ovid]

There is some work for ranges discussed in parameterized checks. We went back and forth on this repeatedly. It was hard 😃 The parameterized arrays and parameterized nums give good examples.

[Ovid]

Oh, and the celsius thing was just a booboo on my part. Nice catch!

[Silyua]

I mean its true

[duncand]

Ranges should be actual value types which can be used in checks but also as actual data, rather than special syntax for checks. And then they can be used in checks in the same way one can have a check like "value must be a member of this array". As such, the whole range thing is really orthogonal to Oshun.

[happy-barney]

@druud what about this ?

Int[ max_inclusive ($MIN), min_exclusive ($MAX) ]

[zmughal]

Or, if we had a way to treat operators as functions (which could be curried), we could do:

INT & BinaryOp[ '>=' , $MIN ] & BinaryOp[ '<' , $MAX ]

This could be done via a look-up table and something like @tobyink's Hydrogen. The reason I'm emphasizing breaking up the types into simpler rules that can be built up together is that a tool can reason about them without having to know how to parse Perl expressions with proper precedence (which is the difficult part of the grammar that Damian's PPR and Kennedy's PPI both side step). I am actually working on a mix of Perl and Prolog right now to demonstrate the initial feasibility of this kind of specification of checks.

[happy-barney]

@zmughal intention vs implementation ... I prefer to name what I want to achieve.
Also with my approach context of [...] is Oshun grammar whereas context of (...) is Perl grammar.

[zmughal]

Yeah, both are valid and if I were writing a tool for doing extended checking, I don't think it would be too much work to support both to map to the same bit of logic. I would probably write a library of user-defined checks that implement BinaryOp and maybe more for other mathematical operations (especially for PDL which will be interesting because it includes shape as a parameter --- which is shares with APL).

[druud]

For building up config (and to me a check is similar), I like a chained syntax, where each link refines, and then returns, its input. The '&' approach is similar, if it has some localised $CHECK object that it can mutate at will.