Hi Danny, Thanks for preparing this design document. IMO It's a very useful feature, especially combined with time attribute support to specify watermark in DDL.
The design doc looks quite good, but I would suggest to reduce the scope of the first version. Like we don't have to support "STORED" in the first MVP version, and you can also delete related content in document to make it more clean and easier to understand. Best, Kurt On Tue, Sep 17, 2019 at 9:18 PM Qi Luo <luoqi...@gmail.com> wrote: > Fantastic! We're also very interested in this feature. > > +Boxiu > > On Tue, Sep 17, 2019 at 11:31 AM Danny Chan <yuzhao....@gmail.com> wrote: > > > In umbrella task FLINK-10232 we have introduced CREATE TABLE grammar in > > our new module flink-sql-parser. And we proposed to use computed column > to > > describe the time attribute of process time in the design doc FLINK SQL > > DDL, so user may create a table with process time attribute as follows: > > create table T1( > > a int, > > b bigint, > > c varchar, > > d as PROCTIME, > > ) with ( > > 'k1' = 'v1', > > 'k2' = 'v2' > > ); > > > > The column d would be a process time attribute for table T1. > > > > Besides that, computed columns have several other use cases, such as > > these [2]: > > > > > > • Virtual generated columns can be used as a way to simplify and unify > > queries. A complicated condition can be defined as a generated column and > > referred to from multiple queries on the table to ensure that all of them > > use exactly the same condition. > > • Stored generated columns can be used as a materialized cache for > > complicated conditions that are costly to calculate on the fly. > > • Generated columns can simulate functional indexes: Use a generated > > column to define a functional expression and index it. This can be useful > > for working with columns of types that cannot be indexed directly, such > as > > JSON columns. > > • For stored generated columns, the disadvantage of this approach is that > > values are stored twice; once as the value of the generated column and > once > > in the index. > > • If a generated column is indexed, the optimizer recognizes query > > expressions that match the column definition and uses indexes from the > > column as appropriate during query execution(Not supported yet). > > > > > > > > Computed columns are introduced in SQL-SERVER-2016 [1], MYSQL-5.6 [2] and > > ORACLE-11g [3]. > > > > This is the design doc: > > > > > https://docs.google.com/document/d/110TseRtTCphxETPY7uhiHpu-dph3NEesh3mYKtJ7QOY/edit?usp=sharing > > > > Any suggestions are appreciated, thanks. > > > > [1] > > > https://docs.microsoft.com/en-us/sql/relational-databases/tables/specify-computed-columns-in-a-table?view=sql-server-2016 > > [2] > > > https://dev.mysql.com/doc/refman/5.7/en/create-table-generated-columns.html > > [3] https://oracle-base.com/articles/11g/virtual-columns-11gr1 > > > > Best, > > Danny Chan > > >
