This is similar to Stdlib.Seq.S.t but the suspended node is a result.

Consequently, the sequence of elements may be interrupted by an error. Specifically, there are two possible kinds of sequences:

• interrupted sequences where one of the suspended nodes is not returned and an Error _ is produced instead, and
• whole sequences where all the suspended nodes are actually returned inside an Ok _.

All the traversors below treat sequence interruption as an error that is returned as is.

Also note that nodes are suspended by a continuation rather than a lazy block. As a result, different traversals of the same sequence can lead to repeated evaluations of the same elements, or distinct sets of elements, or even a different kind of sequence. E.g., if a suspended sequence fails or succeeds depending on the content of a reference.

This is not recommended. You should use deterministic sequences that do not depend on state. Or you should use one-shot sequences that are used once and then never again. The documentation of this module is written assuming you adhere to these constraints.

A whole sequence of zero elements.

return x is a whole sequence containing the single element x.

return_e (Ok x) is a whole sequence containing the single element x. return_e (Error e) is a sequence immediately interrupted by the error e.

interrupted e is a sequence immediately interrupted by the error e.

nil is the node forming the empty sequence.

cons x s is the sequence containing x followed by s. It is a whole sequence if s is.

cons_e (Ok x) s is the sequence containing x followed by s. It is a whole sequence if s is.

cons_e (Error e) s is a sequence immediately interrupted by e.

append s1 s2 is a sequence s. If s1 is a whole sequence then s is composed of all the elements of s1 followed by s2. If s1 is an interrupted sequence then s is indistinguishable from s1.

s is a whole sequence if both s1 and s2 are.

first s is None if s is empty, it is Some (Error e) if s is immediately interrupted by e, it is Some (Ok x) where x is the first element of s otherwise.

Note that first forces the first element of the sequence, which can have side-effects or be computationally expensive. Consider, e.g., the case where s = filter (fun …) s': first s can force multiple of the values from s'.

fold_left f init seq is

• if seq is a whole sequence, then Ok x where x is the result of folding f over all the elements of seq starting with init, or
• if seq is interrupted by Error e, then Error e.

Note that, as with all other traversors below, if the sequence is interrupted, all the side-effects of f on the successful prefix of seq have already been applied before the traversal returns Error _.

fold_left_e f init seq folds f over the elements of seq with an accumulator set at init. It stops traversal (returning Error _) if f returns an Error _ or if the sequence is interrupted. Otherwise it returns Ok _.

It is the responsibility of the caller to differentiate between errors from the function and errors from the sequence. The function map_error may come in handy. E.g.,

fold_left_e
  (fun acc item ->
    f acc item |> Result.map_error (fun e -> `Traverse e))
  init
  (s |> map_error (fun e -> `Interrupt e))

fold_left_s f init seq is a promise that resolves to

• if seq is a whole sequence, then Ok x where x is the result of folding f over all the elements of seq starting with init, or
• if seq is interrupted by Error e, then Error e.

Note that if it returns Error _, the side-effects of f on previous elements have already been applied anyway.

The elements are traversed sequentially. Specifically, a node's suspension is called only when the f-promise of the previous node has resolved. Thus, there might be yielding in between suspensions being called.

fold_left_es f init seq is a promise that resolves to

• if seq is a whole sequence and f-promises always resolve successfully, then the result of folding f over all the elements of seq starting with init,
• otherwise, Error _ with the error that interrupts the sequence or with an error returned by f, whichever happens first.

The elements are traversed sequentially. Specifically, a node's suspension is called only when the f-promise of the previous node has resolved. Thus, there might be yielding in between suspensions being called.

See fold_left_e for a warning about traversal and interruption errors being indistinguishable.

iter f seq is fold_left (fun () x -> f x) () seq

iter_e f seq is fold_left_e (fun () x -> f x) () seq

iter_s f seq is fold_left_s (fun () x -> f x) () seq

iter_es f seq is fold_left_es (fun () x -> f x) () seq

iter_p f seq is a promise p.

• If seq is a whole sequence, then p resolves to Ok () once all the promises created by f on the elements of seq have resolved.
• If seq is interrupted by Error e, then p resolves to Error e once all the promises created by f on the elements of the successful prefix of seq have resolved.

Note that the behaviour for interrupted sequences is in line with the best-effort semantic of Lwtreslib.

map f seq is a sequence feq.

• If seq is a whole sequence, then feq is a whole sequence where the elements are the result of the application of f on the elements of seq.
• If seq is an interrupted sequence, then feq is a sequence interrupted by Error e where the elements of the successful prefix are the result of the application of f on the elements of the successful prefix of seq.

map_error f seq is a sequence feq.

• If seq is a whole sequence, then feq is the same whole sequence.
• If seq is an interrupted sequence, then feq is a sequence interrupted by Error (f e) where the elements of the successful prefix are the elements of the successful prefix of seq.

map_e f seq is a sequence feq.

• If seq is a whole sequence and if f is successful on all the elements of seq, then feq is a whole sequence where the elements are x where Ok x is the result of the application of f on the elements of seq.
• Otherwise feq is a sequence composed of elements of seq mapped by f and interrupted by f returning Error or by seq's interruption (whichever comes first).

filter f s is a sequence of the same kind as s with only the elements for which f returns true.

filter_e f s is a sequence that is interrupted like s or by f being unsuccessful (whichever comes first) or whole (if neither cases apply). Whichever is the case, the elements of the resulting sequence are the elements of s for which f returns Ok true.

filter_map f s is a sequence of the same kind as s where the elements are transformed by f (when it returns Some _) or dropped (when it returns None).

filter_map_e f s is a sequence that is whole or that is interrupted in the same way as s (if it is) or that is interrupted by f (if it happens). Whichever the case, the elements of the sequence or the successful prefix thereof are transformed by f (when it returns Some _) or dropped (when it returns None).