DOYUNG YOON created S2GRAPH-81:
----------------------------------
Summary: Separate Serializable's toKeyValues into 3, toRowKey,
toQualifier, toValue
Key: S2GRAPH-81
URL: https://issues.apache.org/jira/browse/S2GRAPH-81
Project: S2Graph
Issue Type: Sub-task
Reporter: DOYUNG YOON
Assignee: DOYUNG YOON
Priority: Minor
{{StorageSerializable}} trait under core.storage package only have following
interface.
{noformat}
trait StorageSerializable[E] {
def toKeyValues: Seq[SKeyValue]
}
{noformat}
This layer is responsible to build {{SKeyValue}} from {{Edge}}/{{Vertex}},
which will be used in each storage implementation to actually store them.
I am suggesting to separate this into 3 piece.
# toRowKey
# toQualifier
# toValue
The reason behind of this suggestion is for read path.
When user query comes in, we need to build rpc into storage which require us to
identify the row key.
Most of time, we only need to identify row key, when getEdges from certain
start vertex.
Sometimes, we additionally need qualifier, when check if edge exist between
vertex pair on specific Label.
We are identifying row, qualifier by using {{IndexEdgeSerializer}}, which
currently build bytes for row, qualifier, value altogether.
I think by separating {{toKeyValues}}, read path can avoid unnecessary
serialization.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
