[
https://issues.apache.org/jira/browse/TINKERPOP3-691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14551267#comment-14551267
]
Marko A. Rodriguez commented on TINKERPOP3-691:
-----------------------------------------------
I believe you are talking about these two sets of Exceptions:
{code}
public static UnsupportedOperationException
adjacentVertexLabelsCanNotBeRead() {
return new UnsupportedOperationException("The label of an adjacent
vertex can not be read");
}
public static UnsupportedOperationException
adjacentVertexPropertiesCanNotBeReadOrUpdated() {
return new UnsupportedOperationException("The properties of an
adjacent vertex can not be read or updated");
}
public static UnsupportedOperationException
adjacentVertexEdgesAndVerticesCanNotBeReadOrUpdated() {
return new UnsupportedOperationException("The edges and vertices of
an adjacent vertex can not be read or updated");
}
{code}
AND
{code}
public static UnsupportedOperationException
incidentAndAdjacentElementsCanNotBeAccessedInMapReduce() {
return new UnsupportedOperationException("The computer is in
MapReduce mode and a vertex's incident and adjacent elements can not be
accessed");
}
public static UnsupportedOperationException
vertexPropertiesCanNotBeUpdatedInMapReduce() {
return new UnsupportedOperationException("The computer is in
MapReduce mode and a vertex's properties can not be updated");
}
{code}
................
The first set is with respects to adjacent vertices. Thus, these should only be
thrown during `VertexProgram` execution. The second two are with respect to the
current starVertex (not adjacents). In MapReduce, you can't access edges or
update properties.
TinkerPop provides `ComputerGraph` as a helper to implement the behaviors.
`Giraph`, `Spark`, and `TinkerGraph` all use this thin wrapper as the graph is
always the same data structure, but its important that MapReduce behavior is
different from VertexProgram behavior.
{code}
vertexProgram.execute(ComputerGraph.vertexProgram(vertex, this.vertexProgram),
messenger, memory)
mapReduce.map(ComputerGraph.mapReduce(vertex), mapEmitter)
{code}
> TP3 is too prescriptive in exception
> ------------------------------------
>
> Key: TINKERPOP3-691
> URL: https://issues.apache.org/jira/browse/TINKERPOP3-691
> Project: TinkerPop 3
> Issue Type: Improvement
> Reporter: Matthias Broecheler
> Priority: Minor
>
> In its current state, TP3 is extremely prescriptive about the exact type of
> exception that should be thrown in certain situations. Sometimes these
> exceptions are symptoms of the same behavior (i.e. accessing properties you
> shouldn't) but are different based on the access path (i.e. there are
> separate exceptions for MapReduce and VertexProgram when such access happens).
> This makes it cumbersome for vendors to implement TP3 because you have to
> know in core datastructures where they are being used even though the
> datastructure can be reused between components.
> An example of that is PreloadedVertex in Titan where we have to store in what
> context such a vertex is used.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
