1
0
Fork 0
arangodb/Documentation/Books/Users/Aql/GraphFunctions.mdpp

343 lines
18 KiB
Plaintext

!CHAPTER Graph functions
AQL has the following functions to traverse graphs:
If you have created a graph in the general-graph module you may want to use
[Graph operations](../Aql/GraphOperations.md) instead.
- *PATHS(vertexcollection, edgecollection, direction, followcycles)*:
returns an array of paths through the graph defined by the nodes in the collection
*vertexcollection* and edges in the collection *edgecollection*. For each vertex
in *vertexcollection*, it will determine the paths through the graph depending on the
value of *direction*:
- *"outbound"*: Follow all paths that start at the current vertex and lead to another vertex
- *"inbound"*: Follow all paths that lead from another vertex to the current vertex
- *"any"*: Combination of *"outbound"* and *"inbound"*
The default value for *direction* is *"outbound"*.
If *followcycles* is true, cyclic paths will be followed as well. This is turned off by
default.
The result of the function is an array of paths. Paths of length 0 will also be returned. Each
path is a document consisting of the following attributes:
- *vertices*: array of vertices visited along the path
- *edges*: array of edges visited along the path (may be empty)
- *source*: start vertex of path
- *destination*: destination vertex of path
*Examples*
PATHS(friends, friendrelations, "outbound", false)
FOR p IN PATHS(friends, friendrelations, "outbound")
FILTER p.source._id == "123456/123456" && LENGTH(p.edges) == 2
RETURN p.vertices[*].name
If you have created a graph in the general-graph module you may want to use
[Graph operations](../Aql/GraphOperations.md) instead.
- *TRAVERSAL(vertexcollection, edgecollection, startVertex, direction, options)*:
Traverses the graph described by *vertexcollection* and *edgecollection*,
starting at the vertex identified by id *startVertex*. Vertex connectivity is
specified by the *direction* parameter:
- *"outbound"*: Vertices are connected in *_from* to *_to* order
- *"inbound"*: Vertices are connected in *_to* to *_from* order
- *"any"*: Vertices are connected in both *_to* to *_from* and in
*_from* to *_to* order
Additional options for the traversal can be provided via the *options* document:
- *strategy*: Defines the traversal strategy. Possible values are *depthfirst*
and *breadthfirst*. Defaults to *depthfirst*
- *order*: Defines the traversal order: Possible values are *preorder* and
*postorder*. Defaults to *preorder*
- *itemOrder*: Defines the level item order. Can be *forward* or
*backward*. Defaults to *forward*
- *minDepth*: Minimum path depths for vertices to be included. This can be used to
include only vertices in the result that are found after a certain minimum depth.
Defaults to 0
- *maxIterations*: Maximum number of iterations in each traversal. This number can be
set to prevent endless loops in traversal of cyclic graphs. When a traversal performs
as many iterations as the *maxIterations* value, the traversal will abort with an
error. If *maxIterations* is not set, a server-defined value may be used
- *maxDepth*: Maximum path depth for sub-edges expansion. This can be used to
limit the depth of the traversal to a sensible amount. This should especially be used
for big graphs to limit the traversal to some sensible amount, and for graphs
containing cycles to prevent infinite traversals. The maximum depth defaults to 256,
with the chance of this value being non-sensical. For several graphs, a much lower
maximum depth is sensible, whereas for other, more array-oriented graphs a higher
depth should be used
- *paths*: If *true*, the paths encountered during the traversal will
also be returned along with each traversed vertex. If *false*, only the
encountered vertices will be returned.
- *uniqueness*: An optional document with the following attributes:
- *vertices*:
- *none*: No vertex uniqueness is enforced
- *global*: A vertex may be visited at most once. This is the default.
- *path*: A vertex is visited only if not already contained in the current
traversal path
- *edges*:
- *none*: No edge uniqueness is enforced
- *global*: An edge may be visited at most once. This is the default
- *path*: An edge is visited only if not already contained in the current
traversal path
- *followEdges*: An optional array of example edge documents that the traversal will
expand into. If no examples are given, the traversal will follow all edges. If one
or many edge examples are given, the traversal will only follow an edge if it matches
at least one of the specified examples. *followEdges* can also be a string with the
name of an AQL user-defined function that should be responsible for checking if an
edge should be followed. In this case, the AQL function will is expected to have the
following signature:
function (config, vertex, edge, path)
The function is expected to return a boolean value. If it returns *true*, the edge
will be followed. If *false* is returned, the edge will be ignored.
- *filterVertices*: An optional array of example vertex documents that the traversal will
treat specially. If no examples are given, the traversal will handle all encountered
vertices equally. If one or many vertex examples are given, the traversal will exclude
any non-matching vertex from the result and/or not descend into it. Optionally,
*filterVertices* can contain the name of a user-defined AQL function that should be responsible
for filtering. If so, the AQL function is expected to have the following signature:
function (config, vertex, path)
If a custom AQL function is used, it is expected to return one of the following values:
- *[ ]*: Include the vertex in the result and descend into its connected edges
- *[ "prune" ]*: Will include the vertex in the result but not descend into its connected edges
- *[ "exclude" ]*: Will not include the vertex in the result but descend into its connected edges
- *[ "prune", "exclude" ]*: Will completely ignore the vertex and its connected edges
- *vertexFilterMethod*: Only useful in conjunction with *filterVertices* and if no user-defined
AQL function is used. If specified, it will influence how vertices are handled that don't match
the examples in *filterVertices*:
- *[ "prune" ]*: Will include non-matching vertices in the result but not descend into them
- *[ "exclude" ]*: Will not include non-matching vertices in the result but descend into them
- *[ "prune", "exclude" ]*: Will neither include non-matching vertices in the result nor descend into them
The result of the TRAVERSAL function is a array of traversed points. Each point is a
document consisting of the following attributes:
- *vertex*: The vertex at the traversal point
- *path*: The path history for the traversal point. The path is a document with the
attributes *vertices* and *edges*, which are both arrays. Note that *path* is only present
in the result if the *paths* attribute is set in the *options*
*Examples*
TRAVERSAL(friends, friendrelations, "friends/john", "outbound", {
strategy: "depthfirst",
order: "postorder",
itemOrder: "backward",
maxDepth: 6,
paths: true
})
// filtering on specific edges (by specifying example edges)
TRAVERSAL(friends, friendrelations, "friends/john", "outbound", {
strategy: "breadthfirst",
order: "preorder",
itemOrder: "forward",
followEdges: [ { type: "knows" }, { state: "FL" } ]
})
// filtering on specific edges and vertices
TRAVERSAL(friends, friendrelations, "friends/john", "outbound", {
strategy: "breadthfirst",
order: "preorder",
itemOrder: "forward",
followEdges: [ { type: "knows" }, { state: "FL" } ],
filterVertices: [ { isActive: true }, { isDeleted: false } ],
vertexFilterMethod: [ "prune", "exclude" ]
})
// using user-defined AQL functions for edge and vertex filtering
TRAVERSAL(friends, friendrelations, "friends/john", "outbound", {
followEdges: "myfunctions::checkedge",
filterVertices: "myfunctions::checkvertex"
})
// to register the custom AQL functions, execute something in the fashion of the
// following commands in arangosh once:
var aqlfunctions = require("org/arangodb/aql/functions");
// these are the actual filter functions
aqlfunctions.register("myfunctions::checkedge", function (config, vertex, edge, path) {
return (edge.type !== 'dislikes'); // don't follow these edges
}, false);
aqlfunctions.register("myfunctions::checkvertex", function (config, vertex, path) {
if (vertex.isDeleted || ! vertex.isActive) {
return [ "prune", "exclude" ]; // exclude these and don't follow them
}
return [ ]; // include everything else
}, false);
If you have created a graph in the general-graph module you may want to use
[Graph operations](../Aql/GraphOperations.md) instead.
- *TRAVERSAL_TREE(vertexcollection, edgecollection, startVertex, direction, connectName, options)*:
Traverses the graph described by *vertexcollection* and *edgecollection*,
starting at the vertex identified by id *startVertex* and creates a hierarchical result.
Vertex connectivity is establish by inserted an attribute which has the name specified via
the *connectName* parameter. Connected vertices will be placed in this attribute as an
array.
The *options* are the same as for the *TRAVERSAL* function, except that the result will
be set up in a way that resembles a depth-first, pre-order visitation result. Thus, the
*strategy* and *order* attributes of the *options* attribute will be ignored.
*Examples*
TRAVERSAL_TREE(friends, friendrelations, "friends/john", "outbound", "likes", {
itemOrder: "forward"
})
When using one of AQL's graph functions please make sure that the graph does not contain cycles,
or that you at least specify some maximum depth or uniqueness criteria for a traversal.
If no bounds are set, a traversal may run into an endless loop in a cyclic graph or sub-graph,
and even in a non-cyclic graph, traversing far into the graph may consume a lot of processing
time and memory for the result set.
If you have created a graph in the general-graph module you may want to use
[Graph operations](../Aql/GraphOperations.md) instead.
- *SHORTEST_PATH(vertexcollection, edgecollection, startVertex, endVertex, direction, options)*:
Determines the first shortest path from the *startVertex* to the *endVertex*.
Both vertices must be present in the vertex collection specified in *vertexcollection*,
and any connecting edges must be present in the collection specified by *edgecollection*.
Vertex connectivity is specified by the *direction* parameter:
- *"outbound"*: Vertices are connected in *_from* to *_to* order
- *"inbound"*: Vertices are connected in *_to* to *_from* order
- *"any"*: Vertices are connected in both *_to* to *_from* and in
*_from* to *_to* order
The search is aborted when a shortest path is found. Only the first shortest path will be
returned. Any vertex will be visited at most once by the search.
Additional options for the traversal can be provided via the *options* document:
- *maxIterations*: Maximum number of iterations in the search. This number can be
set to bound long-running searches. When a search performs as many iterations as the
*maxIterations* value, the search will abort with an error. If *maxIterations* is not
set, a server-defined value may be used.
- *paths*: If *true*, the result will not only contain the vertices along the shortest
path, but also the connecting edges. If *false*, only the encountered vertices will
be returned.
- *distance*: An optional custom function to be used when calculating the distance
between a vertex and a neighboring vertex. The expected function signature is:
function (config, vertex1, vertex2, edge)
Both vertices and the connecting edge will be passed into the function. The function
is expected to return a numeric value that expresses the distance between the two
vertices. Higher values will mean higher distances, giving the connection a lower
priority in further analysis.
If no custom distance function is specified, all vertices are assumed to have the
same distance (1) to each other. If a function name is specified, it must have been
registered as a regular user-defined AQL function.
- *followEdges*: An optional array of example edge documents that the search will
expand into. If no examples are given, the search will follow all edges. If one
or many edge examples are given, the search will only follow an edge if it matches
at least one of the specified examples. *followEdges* can also be a string with the
name of an AQL user-defined function that should be responsible for checking if an
edge should be followed. In this case, the AQL function will is expected to have the
following signature:
function (config, vertex, edge, path)
The function is expected to return a boolean value. If it returns *true*, the edge
will be followed. If *false* is returned, the edge will be ignored.
- *filterVertices*: An optional array of example vertex documents that the search will
treat specially. If no examples are given, the search will handle all encountered
vertices equally. If one or many vertex examples are given, the search will exclude
the vertex from the result and/or not descend into it. Optionally, *filterVertices* can
contain the name of a user-defined AQL function that should be responsible for filtering.
If so, the AQL function is expected to have the following signature:
function (config, vertex, path)
If a custom AQL function is used, it is expected to return one of the following values:
- *[ ]*: Include the vertex in the result and descend into its connected edges
- *[ "prune" ]*: Will include the vertex in the result but not descend into its connected edges
- *[ "exclude" ]*: Will not include the vertex in the result but descend into its connected edges
- *[ "prune", "exclude" ]*: Will completely ignore the vertex and its connected edges
The result of the SHORTEST_PATH function is an array with the components of the shortest
path. Each component is a document consisting of the following attributes:
- *vertex*: The vertex at the traversal point
- *path*: The path history for the traversal point. The path is a document with the
attributes *vertices* and *edges*, which are both arrays. Note that *path* is only present
in the result if the *paths* attribute is set in the *options*.
*Examples*
SHORTEST_PATH(cities, motorways, "cities/CGN", "cities/MUC", "outbound", {
paths: true
})
// using a user-defined distance function
SHORTEST_PATH(cities, motorways, "cities/CGN", "cities/MUC", "outbound", {
paths: true,
distance: "myfunctions::citydistance"
})
// using a user-defined function to filter edges
SHORTEST_PATH(cities, motorways, "cities/CGN", "cities/MUC", "outbound", {
paths: true,
followEdges: "myfunctions::checkedge"
})
// to register a custom AQL distance function, execute something in the fashion of the
// following commands in arangosh once:
var aqlfunctions = require("org/arangodb/aql/functions");
// this is the actual distance function
aqlfunctions.register("myfunctions::distance", function (config, vertex1, vertex2, edge) {
return Math.sqrt(Math.pow(vertex1.x - vertex2.x) + Math.pow(vertex1.y - vertex2.y));
}, false);
// this is the filter function for the edges
aqlfunctions.register("myfunctions::checkedge", function (config, vertex, edge, path) {
return (edge.underConstruction === false); // don't follow these edges
}, false);
- *EDGES(edgecollection, startvertex, direction, edgeexamples)*:
Return all edges connected to the vertex *startvertex* as an array. The possible values for
*direction* are:
- *outbound*: Return all outbound edges
- *inbound*: Return all inbound edges
- *any*: Return outbound and inbound edges
The *edgeexamples* parameter can optionally be used to restrict the results to specific
edge connections only. The matching is then done via the *MATCHES* function.
To not restrict the result to specific connections, *edgeexamples* should be left
unspecified.
*Examples*
EDGES(friendrelations, "friends/john", "outbound")
EDGES(friendrelations, "friends/john", "any", [ { "$label": "knows" } ])
If you have created a graph in the general-graph module you may want to use
[Graph operations](../Aql/GraphOperations.md) instead.
- *NEIGHBORS(vertexcollection, edgecollection, startvertex, direction, edgeexamples)*:
Return all neighbors that are directly connected to the vertex *startvertex* as an array.
The possible values for *direction* are:
- *outbound*: Return all outbound edges
- *inbound*: Return all inbound edges
- *any*: Return outbound and inbound edges
The *edgeexamples* parameter can optionally be used to restrict the results to specific
edge connections only. The matching is then done via the *MATCHES* function.
To not restrict the result to specific connections, *edgeexamples* should be left
unspecified.
*Examples*
NEIGHBORS(friends, friendrelations, "friends/john", "outbound")
NEIGHBORS(users, usersrelations, "users/john", "any", [ { "$label": "recommends" } ] )