db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "martin (JIRA)" <j...@apache.org>
Subject [jira] Issue Comment Edited: (DERBY-3937) Select count(*) scans all the rows (and is therefore slow with big tables), is the amount of rows not available/known for example in index ?
Date Thu, 30 Jul 2009 12:47:15 GMT

    [ https://issues.apache.org/jira/browse/DERBY-3937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12737087#action_12737087
] 

martin edited comment on DERBY-3937 at 7/30/09 5:45 AM:
--------------------------------------------------------

> The current implementations of indexes and base tables do not maintain an exact count
of rows. 

Index could be useful even if it does not maintain "exact count". If scanning index can provide
the result, it should be faster than scanning the table itself. Is it feasible?


      was (Author: mstanik):
    > The current implementations of indexes and base tables do not maintain an exact count
of rows. 

Index could be useful even if it does not maintain "exact count". If scanning index can provide
the result, scanning index should be faster than scanning the table itself. Is it feasible?

  
> Select count(*) scans all the rows (and is therefore slow with big tables), is the amount
of rows not available/known for example in index ?
> --------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-3937
>                 URL: https://issues.apache.org/jira/browse/DERBY-3937
>             Project: Derby
>          Issue Type: Improvement
>          Components: Store
>         Environment: Any
>            Reporter: Martin Hajduch
>
> Create table with 5000000 rows. Create index on unique ID. Select count(*) on such table
is going to take quite some time.
> Shouldn't the index contain amount of indexed rows and the value taken from there ?
> Additionally, queries of the form select count(*) from table where col1=value; take lots
of time (depending on amount of rows satisfying WHERE clause) even if index on col1 exists.
Isn't it possible to find first and last occurence in the index, and then calculate amount
of rows more effectively then scanning through all of them ?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message