Add partition spec evolution to Table API

[rdblue]

Tables currently support schema evolution using updateSchema():

table.updateSchema()
    .addColumn("a", Types.IntegerType.get())
    .commit()

Iceberg should support a similar API for evolving a partition spec, which is currently a manual operation.

[timmylicheng]

@rdblue Hey, could you say more about the manual operation here? When you say iceberg should support updateSchema, which class or object are you referring to? Does it have something to do with issue #280 ?

[rdblue]

@timmylicheng, #280 should be done first so that metadata tables can handle schema evolution.

Updating a schema works fine, I'm referencing it so that we have a guide for building partition spec evolution. I think the two APIs should be similar.

Updating a partition spec is currently a manual operation because there is no API for making the change. You can build a replacement spec and add it to a table, though. Unfortunately, you might need to do that by editing JSON, depending on what methods we expose to update partition specs in TableMetadata.

[massdosage]

Is this done now? Just curious as it mentions closed and merged PRs.

[rdblue]

Yes. I'll close it. Thanks!

The partition evolution API was added in #1919 and #1942. Spark DDL commands were added in #1948:

ALTER TABLE db.table ADD PARTITION FIELD category;
ALTER TABLE db.table ADD PARTITION FIELD hours(event_ts) AS event_hour;
ALTER TABLE db.table DROP PARTITION FIELD bucket(16, id);