Compared to other doubly-linked lists, in this one:
1. Calls to modification functions (insert*, move*, ...) detect if the list is being iterated over (iter, fold, ...), and if so raise an exception. For example, a use like the following would raise.
iter t ~f:(fun _ -> ... remove t e ...)
2. There is a designated "front" and "back" of each list, rather than viewing each element as an equal in a ring.
3. Elements know which list they're in. Each operation that takes an
t, first checks that the
Elt belongs to the
t, and if not, raises.
4. Related to (3), lists cannot be split, though a sort of splicing is available as
transfer. In other words, no operation will cause one list to become two. This
makes this module unsuitable for maintaining the faces of a planar graph under edge
insertion and deletion, for example.
5. Another property permitted by (3) and (4) is that
length is O(1).
of_list l returns a doubly-linked list
t with the same elements as
l and in the
same order (i.e. the first element of
l is the first element of
t). It is always
the case that
l = to_list (of_list l).
fold_elt t ~init ~f is the same as fold, except
f is called with the
from the list instead of the contained
Note that like other iteration functions, it is an error to mutate
t inside the
fold. If you'd like to call
remove on any of the
'a Elt.t's, use
filter_inplace t ~f removes all elements of
t that don't satisfy
unchecked_iter t ~f behaves like
iter t ~f except that
f is allowed to modify
t. Adding or removing elements before the element currently being visited has no
effect on the traversal. Elements added after the element currently being visited
will be traversed. Elements deleted after the element currently being visited will
not be traversed. Deleting the element currently visited is an error that is not
detected (presumably leading to an infinite loop) .