Spark Delta table restore to version

user_2130636 0
Viewed 1151

I am trying to restore a delta table to its previous version via spark java , am using local ide .code is as below

import io.delta.tables.*; 
DeltaTable deltaTable = DeltaTable.forPath(spark, <path-to-table>); 
DeltaTable deltaTable = DeltaTable.forName(spark, <table-name>); 
deltaTable.restoreToVersion(0) // restore table to     oldest version
deltaTable.restoreToTimestamp("2019-02-14") // restore to a specific timestamp

As per the documentation databricks doc the method given here is not available in delta-core version 0.8.0. The method is also not in the api docs .

Is this only available in Datbricks run time?

Currently i have to load with the previous version and rewrite the df using delta.Is there any better way to do it?

delta-lake databricks apache-spark
2 Answers

Deltalake version 0.8 does not have restoreToVersion and restoreToTimestamp methods. There is no trace of such methods in open-source deltalake 0.8 as you can check in delta-lake repository

So currently and as far as I know, there is no other method than rewriting from a previous version, as explained in answers of this question

EDIT

As commented by boyangeor, restoreToVersion and restoreToTimestamp are now available in DeltaLake from version 1.2

user_68077690

The rollback "restoreVersion" is very much this:

In python

delta_table_path = "/tmp/delta-table"
df = spark.read.format("delta").option("versionAsOf", 0).load(delta_table_path)     
df.show()

In Java:

String delta_table_path = "/tmp/delta-table"
Dataset<Row> df = spark.read().format("delta").option("versionAsOf", 0).load(delta_table_path);
df.show();

In Scala:

var delta_table_path = "/tmp/delta-table"
val df = spark.read.format("delta").option("versionAsOf", 0).load(delta_table_path)
df.show()
user_8445180
Related