SQL - DELETE EDGE
Removes edges from the database. This is the equivalent of the DELETE
command, with the addition of checking and maintaining consistency with vertices by removing all cross-references to the edge from both the in
and out
vertex properties.
Syntax
DELETE EDGE
( <rid>
|
[<rid> (, <rid>)*]
|
( [ FROM (<rid> | <select_statement> ) ] [ TO ( <rid> | <select_statement> ) ] )
|
[<class>]
(
[WHERE <conditions>]
[LIMIT <MaxRecords>]
[BATCH <batch-size>]
FROM
Defines the starting point vertex of the edge to delete.TO
Defines the ending point vertex of the edge to delete.WHERE
Defines the filtering conditions.LIMIT
Defines the maximum number of edges to delete.BATCH
Defines the block size for the operation, allowing you to break large transactions down into smaller units to reduce resource demands. Its default is100
. Feature introduced in 2.1.
Examples
Delete edges where the data is a property that might exist in one or more edges between two vertices:
orientdb>
DELETE EDGE FROM #11:101 TO #11:117 WHERE date >= "2012-01-15"
Delete edges filtering by the edge class:
orientdb>
DELETE EDGE FROM #11:101 TO #11:117 WHERE @class = 'Owns' AND comment LIKE "regex of forbidden words"
Delete edge filtering by the edge class and date:
orientdb>
DELETE EDGE Owns WHERE date < "2011-11"
Note that this syntax is faster than filtering the class through the
WHERE
clause.Delete edges where
in.price
shows the condition on theto
vertex for the edge:orientdb>
DELETE EDGE Owns WHERE date < "2011-11" AND in.price >= 202.43
Delete edges in blocks of one thousand per transaction.
orientdb>
DELETE EDGE Owns WHERE date < "2011-11" BATCH 1000
This feature was introduced in version 2.1.
For more information, see
Use Cases
Controling Vertex Version Increments
Creating and deleting edges causes OrientDB to increment versions on the involved vertices. You can prevent this operation by implementing the Bonsai Structure.
By default, OrientDB only uses Bonsai as soon as it reaches the threshold, in order to optimize operation. To always use Bonsai, configure it on the JVM or in the orientdb-server-config.xml
configuration file.
$ javac ... -DridBag.embeddedToSbtreeBonsaiThreshold=-1
To implement it in Java, add the following line to your application at a point before opening the database:
OGlobalConfiguration.RID_BAG_EMBEDDED_TO_SBTREEBONSAI_THRESHOLD.setValue(-1);
For more information, see Concurrency on Adding Edges.
NOTE: When using a distributed database, OrientDB does not support SBTree indexes. In these environments, you must set ridBag.embeddedToSbtreeBonsaiThreshold=Integer.MAX\_VALUE to avoid replication errors._ |
|
---|---|
Deleting Edges from a Sub-query
Consider a situation where you have an edge with a Record ID of #11:0
that you want to delete. In attempting to do so, you run the following query:
orientdb> DELETE EDGE FROM (SELECT FROM #11:0)
This does not delete the edge. To delete edges using sub-queries, you have to use a somewhat different syntax. For instance,
orientdb> DELETE EDGE E WHERE @rid IN (SELECT FROM #11:0)
This removes the edge from your database.
Deleting Edges through Java
When a User
node follows a company
node, we create an edge between the user and the company of the type followCompany
and CompanyFollowedBy
classes. We can then remove the relevant edges through Java.
node1 is User node,
node2 is company node
OGraphDatabase rawGraph = orientGraph.getRawGraph();
String[] arg={"followCompany,"CompanyFollowedBy"};
Set<OIdentifiable> edges=rawGraph.getEdgesBetweenVertexes(node1, node2,null,arg);
for (OIdentifiable oIdentifiable : edges) {
**rawGraph.removeEdge(oIdentifiable);
}
History
2.1
- Implements support for the option
BATCH
clause
1.4
- Command implements the Blueprints API. In the event that you are working in Java using the OGraphDatabase API, you may experience some unexpected results in how edges are managed between versions. To force the command to use the older API, change the GraphDB settings, as described on the [
ALTER DATABASE
])SQL-Alter-Database.md) command examples.
1.1
- First implementation of the feature.