Issue #21148 has been updated by mame (Yusuke Endoh). Student (Nathan Zook) wrote:
Suppose someone deals five cards face down from a regulation poker deck, and we wish to reason about the number of aces. We know that that number is one of zero, one, two, three, or four. Therefore, if someone asks "is this number larger than negative one?", the answer is yes even without ever needing to resolve the actual value.
What would `ace_count > 2` be, for example? Return true or false according to its probability? Raise an exception? Something else? ---------------------------------------- Feature #21148: Class proposal: IndefiniteNumeric < Numeric https://bugs.ruby-lang.org/issues/21148#change-112256 * Author: Student (Nathan Zook) * Status: Open ---------------------------------------- Suppose someone deals five cards face down from a regulation poker deck, and we wish to reason about the number of aces. We know that that number is one of zero, one, two, three, or four. Therefore, if someone asks "is this number larger than negative one?", the answer is yes even without ever needing to resolve the actual value. This proposal is inspired by the proposed `Enumerable::Lazy::Length` class of Kevin Newton in response to #21135 The `IndefiniteNumeric` class would serve as a base class for such values. In particular, it relies on a subclass to define `#value`, which returns a definite `Numeric`. ```ruby class IndefiniteNumeric < Numeric def coerce(other) = value.coerce(other) def +(other) = value + other def -(other) = value - other def *(other) = value * other def /(other) = value / other def <=>(other) = value <=> other end ``` It is expected that in particular, `<=>` will be overridden for subclasses where `#value` might be slow to return. Usually, we will know more than just that the value in question is `Numeric`. It seems appropriate have `IndefiniteInteger` and so forth. What is not clear to me is if `IndefiniteInteger` should be a subclass of `IndefiniteNumeric` or of `Integer`. (Such a thing does not appear to be currently possible.) If subclassing cannot be, what about defining `#kind_of?(Integer)` to be `true`? Note: with the length of an `Enumerable` as inspiration, I would argue that an `IndefiniteInteger` might actually value to `Float::INFINITY`, so it should NOT define `<=>` against plus or minus `Float::INFINITY`. -- https://bugs.ruby-lang.org/