JacobZheng created SPARK-46071:
----------------------------------
Summary: TreeNode.toJSON may result in OOM when there are multiple
levels of nesting of expressions.
Key: SPARK-46071
URL: https://issues.apache.org/jira/browse/SPARK-46071
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 3.4.0
Reporter: JacobZheng
I am encountering an OOM exception when executing the following code:
{code:scala}
parser.parseExpression(sql).toJSON
{code}
This sql is a multiple nesting of {*}_CaseWhen_{*}. After testing I found that
the number of expressions in the json increases exponentially as the number of
nestings increases.
Here are some example:
sql:
{code:sql}
CASE WHEN(`cost` <= 275) THEN '(270-275]'
ELSE '----' END
{code}
json:
{code:json}
[
{
"class":"org.apache.spark.sql.catalyst.expressions.CaseWhen",
"num-children":3,
"branches":[
{
"product-class":"scala.Tuple2",
"_1":[
{
"class":"org.apache.spark.sql.catalyst.expressions.LessThanOrEqual",
"num-children":2,
"left":0,
"right":1
},
{
"class":"org.apache.spark.sql.catalyst.analysis.UnresolvedAttribute",
"num-children":0,
"nameParts":"[cost]"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"275",
"dataType":"integer"
}
],
"_2":[
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"(270-275]",
"dataType":"string"
}
]
}
],
"elseValue":[
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"----",
"dataType":"string"
}
]
},
{
"class":"org.apache.spark.sql.catalyst.expressions.LessThanOrEqual",
"num-children":2,
"left":0,
"right":1
},
{
"class":"org.apache.spark.sql.catalyst.analysis.UnresolvedAttribute",
"num-children":0,
"nameParts":"[cost]"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"275",
"dataType":"integer"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"(270-275]",
"dataType":"string"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"----",
"dataType":"string"
}
]
{code}
The child nodes of the *_CaseWhen_* expression are stored twice in JSON.
When *_CaseWhen_* is nested twice, the child expression of the first case when
is repeated 4 times, and so on.
{code:sql}
CASE WHEN(`cost` <= 270) THEN '(265-270]'
ELSE
CASE WHEN(`cost` <= 275) THEN '(270-275]'
ELSE '----' END END
{code}
Nesting the *_CaseWhen_* expression n times in this case will result in 2^n+11
expressions in the json.
The reason for this problem is that the field of *_CaseWhen_* cannot be
converted to children index of when executing method {*}_jsonFields_{*}.
Perhaps simplifying *_CaseWhen_* json a bit by overriding the *_jsonFields_*
method is a viable way to go.
{code:json}
[
{
"class":"org.apache.spark.sql.catalyst.expressions.CaseWhen",
"num-children":3,
"branches":[
{
"condition":0,
"value":1
}
],
"elseValue":2
},
{
"class":"org.apache.spark.sql.catalyst.expressions.LessThanOrEqual",
"num-children":2,
"left":0,
"right":1
},
{
"class":"org.apache.spark.sql.catalyst.analysis.UnresolvedAttribute",
"num-children":0,
"nameParts":"[cost]"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"275",
"dataType":"integer"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"(270-275]",
"dataType":"string"
},
{
"class":"org.apache.spark.sql.catalyst.expressions.Literal",
"num-children":0,
"value":"----",
"dataType":"string"
}
]
{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
