module Vect:Extensible vectors with constant-time append/prepend.sig..end
This module implements extensible arrays which work very much like ropes as described in Boehm, H., Atkinson, R., and Plass, M. 1995. Ropes: an alternative to strings. Softw. Pract. Exper. 25, 12 (Dec. 1995), 1315-1330.
These vectors have some interesting properties:
All operations but destructive_set (provided for efficient ephemeral usage)
are non-destructive: the original vect is never modified. When a new vect is
returned as the result of an operation, it will share as much data as possible
with its "parent". For instance, if a vect of length n undergoes m
operations (assume n >> m) like set, append or prepend, the modified vector
will only require O(m) space in addition to that taken by the original vect.
However, Vect is an amortized data structure, and its use in a persistent setting
can easily degrade its amortized time bounds. It is thus mainly intended to be used
ephemerally. In some cases, it is possible to use Vect persistently with the same
amortized bounds by explicitly rebalancing vects to be reused using balance.
Special care must be taken to avoid calling balance too frequently; in the limit,
calling balance after each modification would defeat the purpose of amortization.
Author(s): Mauricio Fernandez
type 'a t
exception Out_of_bounds
val max_length : intval empty : 'a tval singleton : 'a -> 'a tval of_array : 'a array -> 'a tof_array s returns a vect corresponding to the array s.
Operates in O(n) time.val to_array : 'a t -> 'a arrayto_array r returns an array corresponding to the vect r.val to_list : 'a t -> 'a listval make : int -> 'a -> 'a tmake i c returns a vect of length i whose elements are all equal to
c; it is similar to Array.makeval is_empty : 'a t -> boolval height : 'a t -> intval length : 'a t -> intO(1)).val balance : 'a t -> 'a tbalance r returns a balanced copy of the r vect. Note that vects are
automatically rebalanced when their height exceeds a given threshold, but
balance allows to invoke that operation explicity.val concat : 'a t -> 'a t -> 'a tconcat r u concatenates the r and u vects. In general, it operates
in O(log(min n1 n2)) amortized time.
Small vects are treated specially and can be appended/prepended in
amortized O(1) time.val append : 'a -> 'a t -> 'a tappend c r returns a new vect with the c element at the end
in amortized O(1) time.val prepend : 'a -> 'a t -> 'a tprepend c r returns a new vect with the c character at the
beginning in amortized O(1) time.val get : int -> 'a t -> 'aget n r returns the (n+1)th element from the vect r; i.e.
get 0 r returns the first element.
Operates in worst-case O(log size) time.
Raises Out_of_bounds if a character out of bounds is requested.val set : int -> 'a -> 'a t -> 'a tset n c r returns a copy of the r vect where the (n+1)th element
(see also get) has been set to c.
Operates in worst-case O(log size) time.val destructive_set : int -> 'a -> 'a t -> unitdestructive_set n e v sets the element of index n in the v vect
to e. This operation is destructive, and will also affect vects
sharing the modified leaf with v. Use with caution.val sub : int -> int -> 'a t -> 'a tsub m n r returns a sub-vect of r containing all the elements
whose indexes range from m to m + n - 1 (included).
Raises Out_of_bounds in the same cases as Array.sub.
Operates in worst-case O(log size) time.val insert : int -> 'a t -> 'a t -> 'a tinsert n r u returns a copy of the u vect where r has been
inserted between the elements with index n and n + 1 in the
original vect. The length of the new vect is
length u + length r.
Operates in amortized O(log(size r) + log(size u)) time.val remove : int -> int -> 'a t -> 'a tremove m n r returns the vect resulting from deleting the
elements with indexes ranging from m to m + n - 1 (included)
from the original vect r. The length of the new vect is
length r - n.
Operates in amortized O(log(size r)) time.val iter : ('a -> unit) -> 'a t -> unititer f r applies f to all the elements in the r vect,
in order.val iteri : (int -> 'a -> unit) -> 'a t -> unitval rangeiter : ('a -> unit) -> int -> int -> 'a t -> unitrangeiter f m n r applies f to all the elements whose
indices k satisfy m <= k < m + n.
It is thus equivalent to iter f (sub m n r), but does not
create an intermediary vect. rangeiter operates in worst-case
O(n + log m) time, which improves on the O(n log m) bound
from an explicit loop using get.
Raises Out_of_bounds in the same cases as sub.val fold_left : ('a -> 'b -> 'a) -> 'a -> 'b t -> 'afold_left f a r computes f (... (f (f a r0) r1)...) rN-1
where rn = Vect.get n r and N = length r.val fold_right : ('a -> 'b -> 'b) -> 'a t -> 'b -> 'bfold_right f r a computes f (r0 ... (f rN-2 (f rN-1 a)) ...))
where rn = Vect.get n r and N = length r.val map : ('a -> 'b) -> 'a t -> 'b tmap f v returns a vect isomorphic to v where each element of index
i equals f (get v i). Therefore, the height of the returned vect
is the same as that of the original one. Operates in O(n) time.val id_map : ('a -> 'a) -> 'a t -> 'a tid_map f v returns a vect isomorphic to v where each element of index
i equals f (get v i). It is very similar to map, but tries to share
as much data as possible with the original vect; for example,
id_map (fun x -> x) v == v. This can lead to significative space savings
if f leaves many values unmodified.
For each element, the new value f x and the old one x are compared
with <>. Operates in O(n) time.val filter : ('a -> bool) -> 'a t -> 'a tfilter f v returns a vect with the elements x from v such that
f x returns true. Operates in O(n) time.module type RANDOMACCESS =sig..end
module Make: