THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
- Efficient predicate pushdown. Can store all columns metadata together. Only query desired columns efficiently
- Parallel processing of splits natively supported
- Can provide UDF support if needed.
- This may be useful for geo queries commonly used
- Example: Table has latitude/longitude columns in table. But we can query data in hexagon/quad-tree efficiently using data skipping index. )
- Better storage compression
- We can try different layouts by sorting data on different parameters (partition/fileId/columnBeingIndexed etc)
...
- Doesn’t work well with hudi metadata table (because metadata table base format is HFile. HUDI table cannot support different file formats for different partitions)
- No fast single key lookup. So, may not be ideal for other types of index like UUID lookup?
...