hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From lars hofhansl <lhofha...@yahoo.com>
Subject Re: Column Prefix delete markers (HBASE-5268)
Date Wed, 25 Jan 2012 02:01:43 GMT
This needs more work. See comment from Todd and Mikhail on the jira. I might do a more complicated
patch or just close as "won't fix".

 From: lars hofhansl <lhofhansl@yahoo.com>
To: hbase-dev <dev@hbase.apache.org> 
Sent: Tuesday, January 24, 2012 4:24 PM
Subject: Column Prefix delete markers (HBASE-5268)
In https://issues.apache.org/jira/browse/HBASE-5268 I propose a new type of delete marker
that in scops sits between a
column delete marker (marking all version of a column older than the delete's TS for deletion)
and family delete marker (marking all versions
of all columns of a family older than the delete's TS).

This new delete marker would mark all versions of all columns that start with a certain prefix
older than the marker's TS for deletion.

With that (together with ColumnRangeFilter) it is possible to model columns and column prefixes
in the exact same way we would model row-keys and their prefixes,
providing more flexibility for HBase key modeling. Columns inside a row can be organized by
prefix and efficiently be targeted for delete.

The Delete object gets a new method: deleteColumnsByPrefix (parameters identical to deleteColumns,
but marking all columns that start with that prefix).

The change is small (there's an indentation change that makes it look bigger) and has no impact
on the read path unless prefix markers are used.

Please share your comments/concerns.


-- Lars
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message