feat(storage): update interpreter and storage support

[zhyass]

I hereby agree to the terms of the CLA available at: https://databend.rs/dev/policies/cla/

Summary

mysql> create table t(a int, b int);
Query OK, 0 rows affected (0.01 sec)

mysql> insert into t values(2,2);
Query OK, 1 row affected (0.01 sec)

mysql> insert into t values(1,1);
Query OK, 1 row affected (0.02 sec)

mysql> update t set a=2 where b=1;
Query OK, 1 row affected (0.03 sec)

mysql> select * from t;
+------+------+
| a    | b    |
+------+------+
|    2 |    1 |
|    2 |    2 |
+------+------+
2 rows in set (0.01 sec)
Read 2 rows, 16.00 B in 0.007 sec., 272.95 rows/sec., 2.13 KiB/sec.

The flow of Pipeline is as follows:

+---------------+      +-----------------------+
|MutationSource1| ---> |SerializeDataTransform1|   ------
+---------------+      +-----------------------+         |      +-----------------+      +------------+
|     ...       | ---> |          ...          |   ...   | ---> |MutationTransform| ---> |MutationSink|
+---------------+      +-----------------------+         |      +-----------------+      +------------+
|MutationSourceN| ---> |SerializeDataTransformN|   ------
+---------------+      +-----------------------+

UpdateSource: check whether need to update and generate a new block and write the object. The process is similar to DeletionSource.

UPDATE column = expression WHERE condition

Replace column to the result of the following expression:

if(condition, CAST(expression, type), column)

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Updated
databend	⬜️ Ignored (Inspect)		Jan 16, 2023 at 6:12AM (UTC)

[sundy-li]

This pr may be conflicted a lot with new expression, maybe we can slightly wait for that merged ?

I think it's better to create a pr based on expression branch.

[BohuTANG]

This pr may be conflicted a lot with new expression, maybe we can slightly wait for that merged ?

I think it's better to create a pr based on expression branch.

OK.
I convert this PR to a draft.
cc @zhyass @dantengsky

[dantengsky]

@zhyass DeletionSource and UpdateSource share some common code, it would be better if we could de-duplicate them. it is ok to me to do the refactoring in a dedicated PR.

[BohuTANG]

👍

[dantengsky]

@zhyass just a reminder:

after PR #9460 is merged, i think, the updation process needs to

• adjust the use of "column identity"
  not quite sure, seems bind_update, and UpdateInterpreter::execute2 should be adjusted.
• adjust the logic of conflict detection
  the simplest way, by now I can come up with, is just to abort if updating in parallel with a scheme evolution, and the update is committed later. but apparently, the strategy could be better for some cases (fields being updated not changed, e.g.)

[BohuTANG]

@soyeric128 looking forward to add update feature to documents, thanks.