[
https://issues.apache.org/jira/browse/PHOENIX-1882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14515107#comment-14515107
]
James Taylor commented on PHOENIX-1882:
---------------------------------------
Patch looks good - I like the new test. I think you can get rid of this hacky
code in IndexMaintainer too (function name will be isRowDeleted there, though,
but you get the idea):
{code}
private DeleteType getDeleteTypeOrNull(Collection<Cell> pendingUpdates) {
int nDeleteCF = 0;
int nDeleteVersionCF = 0;
for (Cell kv : pendingUpdates) {
if (kv.getTypeByte() == KeyValue.Type.DeleteFamily.getCode()) {
nDeleteCF++;
boolean isEmptyCF = Bytes.compareTo(kv.getFamilyArray(),
kv.getFamilyOffset(), kv.getFamilyLength(),
dataEmptyKeyValueCF, 0, dataEmptyKeyValueCF.length) == 0;
// This is what a delete looks like on the client side for
immutable indexing...
if (isEmptyCF) {
return DeleteType.ALL_VERSIONS;
}
} else if (kv.getTypeByte() ==
KeyValue.Type.DeleteFamilyVersion.getCode()) {
nDeleteVersionCF++;
}
}
// This is what a delete looks like on the server side for mutable
indexing...
// Should all be one or the other for DeleteFamily versus
DeleteFamilyVersion, but just in case not
return nDeleteVersionCF >= this.nDataCFs ? DeleteType.SINGLE_VERSION :
nDeleteCF + nDeleteVersionCF >= this.nDataCFs ? DeleteType.ALL_VERSIONS : null;
}
{code}
Can just look like this now:
{code}
private DeleteType getDeleteTypeOrNull(Collection<Cell> pendingUpdates) {
int nDeleteCF = 0;
int nDeleteVersionCF = 0;
for (Cell kv : pendingUpdates) {
if (kv.getTypeByte() == KeyValue.Type.DeleteFamily.getCode()) {
nDeleteCF++;
} else if (kv.getTypeByte() ==
KeyValue.Type.DeleteFamilyVersion.getCode()) {
nDeleteVersionCF++;
}
}
// This is what a delete looks like on the server side for mutable
indexing...
// Should all be one or the other for DeleteFamily versus
DeleteFamilyVersion, but just in case not
return nDeleteVersionCF >= this.nDataCFs ? DeleteType.SINGLE_VERSION :
nDeleteCF + nDeleteVersionCF >= this.nDataCFs ? DeleteType.ALL_VERSIONS : null;
}
{code}
> Issue column family deletes instead of row deletes in PTableImpl
> ----------------------------------------------------------------
>
> Key: PHOENIX-1882
> URL: https://issues.apache.org/jira/browse/PHOENIX-1882
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: James Taylor
> Assignee: Thomas D'Silva
> Fix For: 5.0.0, 4.4.0
>
> Attachments: PHOENIX-1882-4.x-HBase-0.98-v2.patch,
> PHOENIX-1882-4.x-HBase-0.98.patch, PHOENIX-1882-master-v2.patch,
> PHOENIX-1882-master.patch
>
>
> Since PTable knows which column families are under control of Phoenix
> (table.getColumnFamilies()), we should issue a column family delete for each
> of them instead of a row delete (this is what it gets translated to anyway
> when it gets to the server). This is actually better, as Phoenix should not
> attempt to manage undeclared column families.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
