[
https://issues.apache.org/jira/browse/SPARK-26214?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16704475#comment-16704475
]
Thomas Decaux edited comment on SPARK-26214 at 11/30/18 9:32 AM:
-----------------------------------------------------------------
Sure, it's already possible to do it (like I said).
The same thing for "cache" / "persist" DataFrame method:
{code:java}
public DataFrame persist(StorageLevel newLevel) {
this.sqlContext().cacheManager().cacheQuery(this, scala.None..MODULE$,
newLevel);
return this;
}{code}
This is more a "short-cut" method as you can see, it's possible to use
cacheManager *OR* the DataFrame method.
Same thing for registerTempTable:
{code:java}
public void registerTempTable(String tableName) {
this.sqlContext().registerDataFrameAsTable(this, tableName);
}{code}
You can see here, this is a short-cut method.
I propose to do the same thing for broadcast.
was (Author: ebuildy):
Sure, it's already to do this (like I said).
The same thing for "cache" / "persist" DataFrame method:
{code:java}
public DataFrame persist(StorageLevel newLevel) {
this.sqlContext().cacheManager().cacheQuery(this, scala.None..MODULE$,
newLevel);
return this;
}{code}
This is more a "short-cut" method as you can see, it's possible to use
cacheManager *OR* the DataFrame method.
Same thing for registerTempTable:
{code:java}
public void registerTempTable(String tableName) {
this.sqlContext().registerDataFrameAsTable(this, tableName);
}{code}
You can see here, this is a short-cut method.
I propose to do the same thing for broadcast.
> Add "broadcast" method to DataFrame
> -----------------------------------
>
> Key: SPARK-26214
> URL: https://issues.apache.org/jira/browse/SPARK-26214
> Project: Spark
> Issue Type: Wish
> Components: SQL
> Affects Versions: 2.4.0
> Reporter: Thomas Decaux
> Priority: Trivial
> Labels: broadcast, dataframe
>
> As discussed at
> [https://stackoverflow.com/questions/43984068/does-spark-sql-autobroadcastjointhreshold-work-for-joins-using-datasets-join-op/43994022,]
> it's possible to force broadcast of DataFrame, even if total size is greater
> than ``*spark.sql.autoBroadcastJoinThreshold``.*
> But this not trivial for beginner, because there is no "broadcast" method (I
> know, I am lazy ...).
> We could add this method, with a WARN if size is greater than the threshold.
> (if it's an easy one, I could do it?)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
