[ 
https://issues.apache.org/jira/browse/ATLAS-2816?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16575740#comment-16575740
 ] 

Apoorv Naik edited comment on ATLAS-2816 at 8/10/18 4:12 AM:
-------------------------------------------------------------

One suggestion, use the followReferences flag instead of hardcoding the 
ignoreRelationship param. This would make is easier to toggle if certain 
deployment scenario wants to use the relationship details to be captured in the 
entityText. Also follow this guideline for patch creation,

 
 # Work on a local branch
 # Commit the patch on local branch
 # Generate patch using "git format-patch origin/master" (this way you get 
credit by including author info in the patch)
 # Attach the patch to JIRA

 

HTH


was (Author: apoorvnaik):
One suggestion, use the followReferences flag instead of hardcoding the 
ignoreRelationship param. This would make is easier to toggle if certain 
deployment scenario wants to use the relationship details to be captured in the 
entityText.

 

HTH

> Allow ignoring relationship in EntityGraphRetriever for FullTextMapperV2
> ------------------------------------------------------------------------
>
>                 Key: ATLAS-2816
>                 URL: https://issues.apache.org/jira/browse/ATLAS-2816
>             Project: Atlas
>          Issue Type: Bug
>    Affects Versions: 1.0.0
>            Reporter: Chengbing Liu
>            Assignee: Apoorv Naik
>            Priority: Major
>         Attachments: ATLAS-2816.01.patch
>
>
> We encountered a problem when using Hive bridge in production. One database 
> has 5000+ tables. Importing the first table costs only tens of milliseconds, 
> and then it becomes slower with more tables. In the end, it costs 1~2 seconds 
> to import one table.
> After investigation, we realized that it is not necessary for the 
> {{FullTextMapperV2}} to retrieve all the relationship of the database each 
> time a table is imported. The time complexity of importing a whole database 
> actually goes to O(n^2) (n is number of tables).
> We propose to add a parameter to the constructor of {{EntityGraphRetriever}}: 
> {{ignoreRelationship}}. When set to true, {{mapVertexToAtlasEntity}} will 
> skip the {{mapRelationshipAttributes}} call. Since {{FullTextMapperV2}} will 
> not use relationship attributes of the entity, this can save plenty of time 
> when importing entities with a large number of relations.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to