plebeia

Functional storage using Merkle Patricia tree
package plebeia
Legend:
Library
Module
Module type
Parameter
Class
Class type
Library plebeia
Module Plebeia . Cursor

2 Types: Trail and cursor

type modified =
| Modified
| Unmodified of Node_type.indexed * Node_type.hashed
type trail = private
| Top
| Left of trail * Node_type.node * modified
| Right of Node_type.node * trail * modified
| Budded of trail * modified
| Extended of trail * Segment.t * modified

A trail represents the content of the memory stack when recursively exploring a tree. Constructing these trails from closure would be easier, but it would make it harder to port the code to C. The type parameters of the trail keep track of the type of each element on the "stack" using a product type.

The constructors are private. Use '_' prefixed functions with runtime invariant checks.

type info = Info.t
type cursor = private
| Cursor of trail * Node_type.node * Context.t * info(*

The cursor, also known as a zipper combines the information contained in a trail and a subtree to represent an edit point within a tree. This is a functional data structure that represents the program point in a function that modifies a tree. We use an existential type that keeps the .mli sane and enforces the most important: that the hole tags match between the trail and the Node

*)
type t = cursor

2 Constructor with invariant checks

val _Top : trail
val _Left : (trail * Node_type.node * modified) -> trail
val _Right : (Node_type.node * trail * modified) -> trail
val _Budded : (trail * modified) -> trail
val _Extended : (trail * Segment.t * modified) -> trail
val _Cursor : (trail * Node_type.node * Context.t * info) -> cursor

2 Creation

val empty : Context.t -> t

Creates a cursor to a new, empty tree.

2 Accessors

val context : t -> Context.t
val get_storage : t -> Storage.t
val index : t -> Index.t option

Get the index of the node pointed by the cursor, if indexed.

2 Segments

val path_of_trail : trail -> Path.t

Segment side list of the given trail, splitted by buds

val path_of_cursor : t -> Path.t

Segment side list of the given cursor, splitted by buds

val local_seg_of_trail : trail -> Segment.t

Segment side list of the given trail, splitted by buds

val local_seg_of_cursor : t -> Segment.t

Segment side list of the given cursor, splitted by buds

2 View

val view : t -> t * Node_type.view

Get the view of the cursor. Returns also the updated cursor with the view.

val may_forget : t -> t option

If the node pointed by the cursor is indexed, forget the details

2 Zipper functions

type Error.t +=
| Cursor_invariant of string
| Write of string
| Move of string

3 Simple 1 step cursor movement

val go_below_bud : t -> ( t option, Error.t ) Result.t

This function expects a cursor positionned on a bud and moves it one step below.

val go_down_extender : t -> ( t, Error.t ) Result.t

Go down an Extender node. The cursor must point to an Extender.

val go_side : Segment.side -> t -> ( t, Error.t ) Result.t

Go down an Internal node. The cursor must point to an Internal.

val go_up : t -> ( t, Error.t ) Result.t

Go up one level

3 Complex multi step cursor movement

Many of these functions fail when the given cursor does not point to a bud.

type access_result =
| Empty_bud
| Collide of cursor * Node_type.view
| Middle_of_extender of cursor * Segment.t * Segment.t * Segment.t
| Reached of cursor * Node_type.view
| HashOnly of cursor * Hash.t * Segment.t

Result of access_gen

type Error.t +=
| Access of access_result
val error_access : access_result -> ( 'a, Error.t ) Result.t

Make an access result into an error

val access_gen : t -> Segment.t -> ( access_result, Error.t ) Result.t

Follow a segment. t must point to a bud. The function first go below the bud, then follow the segment.

val access_gen' : t -> Segment.t -> ( access_result, Error.t ) Result.t

Follow a segment. t can be any node.

val go_top : t -> t

Move up to the top

val go_up_to_bud : t -> ( t, Error.t ) Result.t

Moves the cursor back to the bud above. Note that this is not like "cd ../".

If the cursor is already at a bud, the cursor will move to its parent bud.

val parent : t -> ( t, Error.t ) Result.t

Moves the cursor back to the bud above. Like "cd ../". The cursor must point to a bud otherwise parent fails.

val subtree : t -> Segment.t -> ( t, Error.t ) Result.t

Moves the cursor down a segment, to the root of a sub-tree. Think "cd segment/"

val create_subtree : t -> Segment.t -> ( t, Error.t ) Result.t

Create a subtree (bud). Think "mkdir segment". The cursor does NOT move from the original position.

val subtree_or_create : t -> Segment.t -> ( t, Error.t ) Result.t

Same as subtree but create a subtree if not exists

val get : t -> Segment.t -> ( t * [ `Leaf of Node_type.view | `Bud of Node_type.view ], Error.t ) Result.t

Gets a value if present in the current tree at the given segment.

val get_value : t -> Segment.t -> ( t * Value.t, Error.t ) Result.t

Gets a value or a bud at the given segment.

val insert : t -> Segment.t -> Value.t -> ( t, Error.t ) Result.t

Inserts a value at the given segment in the current tree. It fails if a value already exists at the segment. The cursor does NOT move from the original position.

val upsert : t -> Segment.t -> Value.t -> ( t, Error.t ) Result.t

Upserts. If a value alrady exists at the segment, it is overwritten. This can still fail if the segment points to a subtree. The cursor does NOT move from the original position.

val update : t -> Segment.t -> Value.t -> ( t, Error.t ) Result.t

Update. A value must be bound at the segment.

val set_hashonly : t -> Segment.segment -> Hash.t -> ( t, Error.t ) Stdlib.result
val delete : t -> Segment.t -> ( t, Error.t ) Result.t

Delete a leaf or subtree. The cursor does NOT move from the original position.

val delete' : t -> Segment.t -> ( t, Error.t ) Result.t

Delete a node. The cursor does NOT move from the original position.

val remove_empty_bud : t -> ( t, Error.t ) Result.t

Remove the empty Bud pointed by the cursor. If the non-root parent becomes empty by the removal, remove_empty_bud recursively removes it, too. If the cursor points non empty bud, it does nothing.

The cursor in the result points to the parent bud of the upmost removed bud.

val alter : t -> Segment.segment -> ( Node_type.view option -> ( Node_type.node, Error.t ) Result.t ) -> ( t, Error.t ) Result.t

alter can easily break the invariants.

val forget_info : t -> t
val fold : init:'acc -> t -> ( 'acc -> t -> [< `Continue | `Exit | `Up ] * 'acc ) -> 'acc

Folding over the node tree. The function can choose the node traversal from the given cursor: either continuing into its sub-nodes, not traversing its sub-nodes, or quit the entire folding.

If a node is shared at multiple places it is visited MORE THAN ONCE. If you want to avoid visiting a shared node at most once, carry a set of visited nodes by indices and check a node is visited or not.

val traverse : 'a -> t list -> ( 'a -> t -> [< `Exit | `Up | `Continue ] * 'a ) -> 'a * t list

More generic tree traversal than fold, which has a step-by-step visitor interface: it does not recurse the structure by itself.

If a node is shared at multiple places it is visited MORE THAN ONCE.

2 Cursor hash compuation

val compute_hash : t -> t * Hash.t

Returns the hash of the node pointed by the cursor. It traverses nodes and compute hashes if necessary.

It also returns the updated cursor with the hash information.

val compute_hash' : ( Node_type.t -> [< `Hashed of Hash.t * Node_type.t | `Not_Hashed of Node_type.view ] ) -> t -> t * Hash.t

Compute the node hash of the node pointed by the given cursor. Tail recursive.

The function argument for short cutting, especially for Disk _. It can do either

* Obtain the node hash of the node somehow from an external source * Retrieve the view of the node to let compute' calculate the node hash of it

2 Statistics

val stat : t -> Stat.t

2 Debug

val dot_of_cursor_ref : ( t -> string ) Stdlib.ref

Placeholder of Graphviz rendering of cursors

module Monad : sig ... end
module Cursor_storage : sig ... end
val deep_stat : int32 -> t -> unit Lwt.t

debug