Hello
I would like to generate predicates in the ExprNodeDesc form, as required by the
HiveStoragePredicateHandler.decomposePredicate
<https://github.com/apache/hive/blob/b8250ac2f30539f6b23ce80a20a9e338d3d31458/ql/src/java/org/apache/hadoop/hive/ql/metadata/HiveStoragePredicateHandler.java#L58>
API.
The goal would be to "break encapsulation" as little as possible.
Here is the context:
I have a map-reduce job that needs to read data from a hive table
Ideally I'd have my MR job (my choice of mapper, reducer, but also
partitioner and output format) integrated in a hive plan. Maybe some day...
Alas, the current solution interacts with HiveMetastoreClient and figures
out the Partitions, the StorageDescriptor-specified input formats and SerDes.
Long story short, I get my data out of object inspectors.
Now I would like to take advantage of predicate push down to optimize
filtering data.
I need to generate ExprNodeDesc objects for the
HiveStoragePredicateHandler.decomposePredicate API, and optionally to make a
ExprNodeEvaluator.
Any advice, tips, or hints are most welcome
thank you
Gabriel Balan
--
The statements and opinions expressed here are my own and do not necessarily
represent those of Oracle Corporation.
