HBase datastore does not serialize table name.

[tomaszguzialek]

I created a sample job using DataCleaner desktop application and saved it as .analysis.xml file. The definition of the input datastore has been serialized without the table name:

    <source>
        <data-context ref="CDH VM HBase"/>
        <columns>
            <column id="col_0" path="mainFamily:iso2" type="VARCHAR"/>
            <column id="col_1" path="mainFamily:iso3" type="VARCHAR"/>
            <column id="col_2" path="mainFamily:country_name" type="VARCHAR"/>
        </columns>
    </source>

HBase database can have multiple tables, so the information where to look for the columns should be included in the column path.

[kaspersorensen]

Actually this is a feature and not a bug. Although it can of course be discussed. It's not specific to HBase.

It is currently like this: We check if the consumed table is 1) from the default schema and) the default schema only contains this table. If that's the case we only save the column names and the job is also loaded properly still. The reason is to maximise portability of a job to other datastores. Often people have multiple files or databases that share the same structure. By saving as little information about the physical structure as possible, we can rather rely on inferring the rest at runtime.

[tomaszguzialek]

I have 3 tables in my HBase. According to your description, shouldn't the table prefix be included?

[kaspersorensen]

Yes. Then I don't understand. Can you try debugging it - I believe this goes on in the JaxbJobWriter class.