Interval.Make
is a functor that takes a type that you'd like to create intervals for
and returns a module with functions over intervals of that type.
For example, suppose you had a Percent.t
type and wanted to work with intervals over
it, i.e., inclusive ranges like 40-50% or 0-100%. You would create your
Percent_interval
module by calling:
module Percent_interval = Interval.Make(Percent)
You now have a module with lots of functionality ready to use. For instance you could
call Percent_interval.empty
to create an empty interval, or:
Percent_interval.create (Percent.of_percentage 3) (Percent.of_percentage 30)
to get an actual interval that ranges from 3%
to 30%
. You can then ask questions
of this interval, like whether it's a subset of another interval or
whether it contains a particular value.
NB. In order to use the Interval.Make
functor, your type must satisfy
Comparable and support bin-io and s-expression conversion. At a minimum, then,
Percent
must look like this:
module Percent = struct
module T = struct
type t = float [@@deriving bin_io, compare, sexp]
end
include T
include Comparable.Make_binable(T)
end
Bound : sig ... end
include sig ... end
val bin_t : t Bin_prot.Type_class.t
val bin_read_t : t Bin_prot.Read.reader
val __bin_read_t__ : (int ‑> t) Bin_prot.Read.reader
val bin_reader_t : t Bin_prot.Type_class.reader
val bin_size_t : t Bin_prot.Size.sizer
val bin_write_t : t Bin_prot.Write.writer
val bin_writer_t : t Bin_prot.Type_class.writer
val bin_shape_t : Bin_prot.Shape.t
val t_of_sexp : Base.Sexp.t ‑> t
val sexp_of_t : t ‑> Base.Sexp.t
type 'a bound
bound
is the type of points in the interval (and therefore of the bounds).
bound
is instantiated in two different ways below: in module type S
as a
monotype and in module type S1
as 'a
.
create l u
returns the interval with lower bound l
and upper bound u
, unless
l > u
, in which case it returns the empty interval.
val empty : 'a t
val is_empty : 'a t ‑> bool
val is_empty_or_singleton : 'a t ‑> bool
convex_hull ts
returns an interval whose upper bound is the greatest upper bound
of the intervals in the list, and whose lower bound is the least lower bound of the
list.
Suppose you had three intervals a
, b
, and c
:
a: ( ) b: ( ) c: ( ) hull: ( )
In this case the hull goes from lbound_exn a
to ubound_exn c
.
bound t x
returns None
iff is_empty t
. If bounds t = Some (a, b)
, then
bound
returns Some y
where y
is the element of t
closest to x
. I.e.:
y = a if x < a y = x if a <= x <= b y = b if x > b
map t ~f
returns create (f l) (f u)
if bounds t = Some (l, u)
, and empty
if
t
is empty. Note that if f l > f u
, the result of map
is empty
, by the
definition of create
.
If you think of an interval as a set of points, rather than a pair of its bounds,
then map
is not the same as the usual mathematical notion of mapping f
over that
set. For example, map ~f:(fun x -> x * x)
maps the interval [-1,1]
to [1,1]
,
not to [0,1]
.
val are_disjoint : 'a t list ‑> bool
are_disjoint ts
returns true
iff the intervals in ts
are pairwise disjoint.
val are_disjoint_as_open_intervals : 'a t list ‑> bool
Returns true iff a given set of intervals would be disjoint if considered as open
intervals, e.g., (3,4)
and (4,5)
would count as disjoint according to this
function.
Assuming that ilist1
and ilist2
are lists of disjoint intervals, list_intersect
ilist1 ilist2
considers the intersection (intersect i1 i2)
of every pair of
intervals (i1, i2)
, with i1
drawn from ilist1
and i2
from ilist2
,
returning just the non-empty intersections. By construction these intervals will be
disjoint, too. For example:
let i = Interval.create;;
list_intersect [i 4 7; i 9 15] [i 2 4; i 5 10; i 14 20];;
[(4, 4), (5, 7), (9, 10), (14, 15)]
Raises an exception if either input list is non-disjoint.
val half_open_intervals_are_a_partition : 'a t list ‑> bool
Returns true if the intervals, when considered as half-open intervals, nestle up
cleanly one to the next. I.e., if you sort the intervals by the lower bound,
then the upper bound of the n
th interval is equal to the lower bound of the
n+1
th interval. The intervals do not need to partition the entire space, they just
need to partition their union.