[
https://issues.apache.org/jira/browse/PHOENIX-4866?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16592535#comment-16592535
]
cherish peng commented on PHOENIX-4866:
---------------------------------------
In my scenes, I have two ways to solve this problem.
The preferred way:
I copied the org.apache.phoenix.compile.FromCompiler to my project lib, then
did some changes:
{code:java}
+protected boolean isAlwaysEmptyFunctionMap = false;
private BaseColumnResolver(PhoenixConnection connection, int tsAddition,
TableName mutatingTableName) {
this.connection = connection;
this.client = connection == null ? null : new MetaDataClient(connection);
this.tsAddition = tsAddition;
functionMap = new HashMap<String, PFunction>(1);
+isAlwaysEmptyFunctionMap = true;
this.functions = Collections.<PFunction>emptyList();
this.mutatingTableName = mutatingTableName;
}
@Override
public PFunction resolveFunction(String functionName) throws SQLException {
+if (functionMap.isEmpty() && isAlwaysEmptyFunctionMap) {
+ functions = createFunctionRef(Arrays.asList(functionName), true);
+ for (PFunction function : functions) {
+ functionMap.put(function.getFunctionName(), function);
+ }
+}
PFunction function = functionMap.get(functionName);
if(function == null) {
throw new FunctionNotFoundException(functionName);
}
return function;
}
{code}
then it works! I am newbie, so this way may be make more noise i am afraid...
Alternative way:
I have to split the whole SQL to two parts;
The first part CONTAINS the FIRST_VALUES function, then i have to insert it to
a temporary table which i should build before this query(because the VIEW's
dosen't support complex SQL)
The secondary part I will read from the temporary table, then use my custom
UDFs(ARRAY_DISTINCT AND OTHERS), but my data size is so large, so this way
drove me crazy!
> UDFs get error: org.apache.phoenix.schema.FunctionNotFoundException: ERROR
> 6001 (42F01): Function undefined
> -----------------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-4866
> URL: https://issues.apache.org/jira/browse/PHOENIX-4866
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 5.0.0
> Environment: [^ArrayDistinctFunction.java]
> Reporter: cherish peng
> Priority: Blocker
> Labels: User-Defined-Function, functions, udfs
> Attachments: ArrayDistinctFunction.java
>
> Original Estimate: 72h
> Remaining Estimate: 72h
>
> I built an udf `ARRAY_DISTINCT`, When i exec:
> {color:#d04437}①{color} SELECT ARRAY_DISTINCT(ARRAY['A', 'B', 'A', 'C'],
> TRUE);
> throw error:
>
> {code:java}
> org.apache.phoenix.schema.FunctionNotFoundException: ERROR 6001 (42F01):
> Function undefined. functionName=ARRAY_DISTINCT at
> org.apache.phoenix.compile.FromCompiler$1.resolveFunction(FromCompiler.java:134)
> at
> org.apache.phoenix.compile.ExpressionCompiler.visitLeave(ExpressionCompiler.java:327)
> {code}
>
> However when i exec:
> {color:#d04437}②{color} SELECT ARRAY_DISTINCT(ARRAY['A', 'B', 'A', 'C'],
> TRUE) {color:#f79232}FROM TABLE_NAME;{color}
> output:
> *+---------------------+*
> *|* *ARRAY['A','B','C']* *|*
> *+---------------------+*
> *+---------------------+*
>
> {color:#d04437}③{color} SELECT
> ARRAY_DISTINCT({color:#f79232}array_elements{color}, TRUE) as arr
> {color:#f79232}FROM TABLE_NAME;{color}**
> {color:#f79232}output:{color}
> *+---------------------+*
> *| ARR* * ** |*
> *+---------------------+*
> *| ARRAY['A','B','C'] |***
> *+---------------------+*
>
> The DDL:
> CREATE FUNCTION ARRAY_DISTINCT(VARCHAR[], BOOLEAN) RETURNS VARCHAR AS
> 'com.godalgo.udf.ArrayDistinctFunction';
> My Environment: apache-phoenix-5.0.0-HBase-2.0-bin hbase-2.0.0 hadoop-3.1.0
>
> I tried the built-in functions which provided by Phoenix, ALL OF THEM ARE OK .
> *I had putted ARRAY_DISTINCT's SOURCE TO attachment.*
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
