Distribution.Types.VersionInterval

A complementary representation of a VersionRange. Instead of a boolean version predicate it uses an increasing sequence of non-overlapping, non-empty intervals.

The key point is that this representation gives a canonical representation for the semantics of VersionRanges. This makes it easier to check things like whether a version range is empty, covers all versions, or requires a certain minimum or maximum version. It also makes it easy to check equality or containment. It also makes it easier to identify 'simple' version predicates for translation into foreign packaging systems that do not support complex version range expressions.

Inspect the list of version intervals.

Directly construct a VersionIntervals from a list of intervals.

Convert a VersionRange to a sequence of version intervals.

Convert a VersionIntervals value back into a VersionRange expression representing the version intervals.

View a VersionRange as a union of intervals.

This provides a canonical view of the semantics of a VersionRange as opposed to the syntax of the expression used to define it. For the syntactic view use foldVersionRange.

Each interval is non-empty. The sequence is in increasing order and no intervals overlap or touch. Therefore only the first and last can be unbounded. The sequence can be empty if the range is empty (e.g. a range expression like && 2).

Other checks are trivial to implement using this view. For example:

isNoVersion vr | [] <- asVersionIntervals vr = True
               | otherwise                   = False

isSpecificVersion vr
   | [(LowerBound v  InclusiveBound
      ,UpperBound v' InclusiveBound)] <- asVersionIntervals vr
   , v == v'   = Just v
   | otherwise = Nothing

VersionIntervals invariant:

• all intervals are valid (lower bound is less then upper bound, i.e. non-empty)

• intervals doesn't touch each other (distinct)