If you are writing a GenericUDF or serde and want to return struct
types there are object inspectors to build structs. The Java type
return is an Object[]. Hive expects that if the struct has 5 fields
the object array will have a length of 5.
On Mon, Jul 23, 2012 at 1:34 PM, <kulkarni.swar...@gmail.com> wrote:
> Cool. Thanks :)
>
> Also was just curious what do people generally use to write struct data in
> hive tables? I see that there is a STRUCT function defined that takes
> parameters and creates structs off them. Can we use a custom class as well?
>
> Thanks again.
>
> Sent from my iPhone
>
> On Jul 23, 2012, at 12:05 PM, Edward Capriolo <edlinuxg...@gmail.com> wrote:
>
>> in your case hbase has a custom serde, the Deserializer interface is
>> what turns the value from the input format into something that hive
>> can understand. HBase support uses the user specified table property
>> columns.mapping as information for what it should parse out of the
>> hbase result.
>>
>> On Mon, Jul 23, 2012 at 12:27 PM, kulkarni.swar...@gmail.com
>> <kulkarni.swar...@gmail.com> wrote:
>>> Hello,
>>>
>>> I kind of have a pretty basic question here. I am trying to read structs
>>> stored in HBase to be read by Hive. In what format should these structs be
>>> written so that they can be read?
>>>
>>> For instance, if my query has the following struct:
>>>
>>> s struct<a: STRING, b: STRING>
>>>
>>> How should I be writing my data in HBase so that when read, it fits into
>>> this struct? In other words, can I create my own class 'MyStruct' which is
>>> something like:
>>>
>>> class MyStruct{
>>> string a;
>>> string b;
>>> }
>>>
>>> to create the struct bytes and read them using hive with the struct defined
>>> above? I hope I made my question clear. I will be glad to provide any
>>> clarifications.
>>>
>>> Thanks,
>>>
>>> --
>>> Swarnim
