zhangbutao commented on code in PR #3276:
URL: https://github.com/apache/hive/pull/3276#discussion_r895032467
##########
standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/dataconnector/jdbc/MySQLConnectorProvider.java:
##########
@@ -90,10 +90,20 @@ protected String getDataType(String dbDataType, int size) {
// map any db specific types here.
switch (dbDataType.toLowerCase())
{
+ case "bit":
+ return toHiveBitType(size);
default:
mappedType = ColumnType.VOID_TYPE_NAME;
break;
}
return mappedType;
}
+
+ private String toHiveBitType(int size) {
+ if (size <= 1) {
+ return ColumnType.BOOLEAN_TYPE_NAME;
+ } else {
+ return ColumnType.BIGINT_TYPE_NAME;
Review Comment:
@nrg4878 Sorry for not updating the issue for a long time.
In my opinion, we have two ways to get origina MySQL BITs values, that is
b'111000' in MySQL will be read as '111000'(binary number) in hive.
First way:
In MySQL, we can get bit values using 'select bin(days) from work_calendar'.
So i was thinking how we use MySQL native language to get bit values in hive.
After some research, I found we could use jdbc table computation
pushdown(hive.sql.query) to rewrite query sql. That is to say, when we query a
bit type table using 'select days from work_calendar' in hive, we can rewrite
the pushdown query to 'select bin(days) from work_calendar'. In this way, we
may need check original MySQL data type and rewrite bit type query when
pushdown. Maybe we should do some change in JdbcSerde.java.
https://cwiki.apache.org/confluence/display/Hive/JDBC+Storage+Handler#JDBCStorageHandler-ComputationPushdown
Second way:
I was also thinking using udf to to cast valus to BITs as you suggested.
That is, 56(bigint) is cast to '111000'(binary number) in hive internal.
Fortunately, I found hive already has this udf bin:
https://github.com/apache/hive/blob/f29cb2245c97102975ea0dd73783049eaa0947a0/ql/src/java/org/apache/hadoop/hive/ql/udf/UDFBin.java#L28-L37
So, based on this pr, we can get bit values using hive query 'select
bin(days) from work_calendar'. And we no need change current jdbc
code(JdbcSerDe.java or others). I like this simple way as we can reuse bin udf
to acheive our goal.
wdyt? if you have better ways, feel free to come up with. thank you.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
