feat(query): virtual column allow cast to other type

[b41sh]

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

Summary

allow virtual column cast to other type, avoiding jsonb value deserialization to improve performance.

• fix meta proto-conv v111 add glue meta test
• virtual column proto add data_types field
• allow virtual column cast to other types, this can avoids jsonb value deserialization to improve performance

From the following example, we can see that the execution time has been reduced from 291.674 sec to 235.888 sec, and the amount of data read has been reduced from 11.01 GiB to 9.16 GiB

root@0.0.0.0:48000/default> CREATE OR REPLACE TABLE user_activity_logs AS
SELECT 
    number % 100000000 AS id,
    JSON_OBJECT(
        'id', CAST(number % 100000000 AS STRING),
        'email', CONCAT('user', CAST(number % 100000000 AS STRING), '@example.com'),
        'nickname', CONCAT('user', CAST(number % 100000000 AS STRING)),
        'phone', CONCAT('******', CAST(number % 10000 + 1000 AS STRING))
    ) AS data
FROM numbers(100000000);

-- create virtual columns
root@0.0.0.0:48000/default> CREATE or replace VIRTUAL COLUMN (
    data['id'],
    data['email'],
    data['nickname'],
    data['phone']
) FOR user_activity_logs;

-- Refresh the virtual columns to activate them
root@0.0.0.0:48000/default> REFRESH VIRTUAL COLUMN FOR user_activity_logs;

root@0.0.0.0:48000/default> select data['id'], data['email'], data['nickname'], data['phone'] from user_activity_logs;
┌────────────────────────────────────────────────────────────────────────────────────────┐
│     data['id']    │        data['email']       │  data['nickname'] │   data['phone']   │
│ Nullable(Variant) │      Nullable(Variant)     │ Nullable(Variant) │ Nullable(Variant) │
├───────────────────┼────────────────────────────┼───────────────────┼───────────────────┤
│ "91732188"        │ "[email protected]" │ "user91732188"    │ "******3188"      │
│ "91732189"        │ "[email protected]" │ "user91732189"    │ "******3189"      │
│ ·                 │ ·                          │ ·                 │ ·                 │
│ ·                 │ ·                          │ ·                 │ ·                 │
│ ·                 │ ·                          │ ·                 │ ·                 │
│ "37499993"        │ "[email protected]" │ "user37499993"    │ "******10993"     │
│ 100000000 rows    │                            │                   │                   │
│ (1000 shown)      │                            │                   │                   │
└────────────────────────────────────────────────────────────────────────────────────────┘
100000000 rows read in 291.674 sec. Processed 100 million rows, 11.01 GiB (342.85 thousand rows/s, 38.67 MiB/s)


root@0.0.0.0:48000/default> CREATE OR REPLACE TABLE user_activity_logs2 AS
SELECT 
    number % 100000000 AS id,
    JSON_OBJECT(
        'id', CAST(number % 100000000 AS STRING),
        'email', CONCAT('user', CAST(number % 100000000 AS STRING), '@example.com'),
        'nickname', CONCAT('user', CAST(number % 100000000 AS STRING)),
        'phone', CONCAT('******', CAST(number % 10000 + 1000 AS STRING))
    ) AS data
FROM numbers(100000000);

-- create virtual columns
root@0.0.0.0:48000/default> CREATE or replace VIRTUAL COLUMN (
    data['id']::int,
    data['email']::string,
    data['nickname']::string,
    data['phone']::string
) FOR user_activity_logs2;

-- Refresh the virtual columns to activate them
root@0.0.0.0:48000/default> REFRESH VIRTUAL COLUMN FOR user_activity_logs2;

root@0.0.0.0:48000/default> select data['id'], data['email'], data['nickname'], data['phone'] from user_activity_logs2;
┌──────────────────────────────────────────────────────────────────────────────────┐
│    data['id']   │       data['email']      │ data['nickname'] │   data['phone']  │
│ Nullable(Int32) │     Nullable(String)     │ Nullable(String) │ Nullable(String) │
├─────────────────┼──────────────────────────┼──────────────────┼──────────────────┤
│        66666656 │ user66666656@example.com │ user66666656     │ ******7656       │
│        66666657 │ user66666657@example.com │ user66666657     │ ******7657       │
│               · │ ·                        │ ·                │ ·                │
│               · │ ·                        │ ·                │ ·                │
│               · │ ·                        │ ·                │ ·                │
│         4166665 │ user4166665@example.com  │ user4166665      │ ******7665       │
│  100000000 rows │                          │                  │                  │
│    (1000 shown) │                          │                  │                  │
└──────────────────────────────────────────────────────────────────────────────────┘
100000000 rows read in 235.888 sec. Processed 100 million rows, 9.16 GiB (423.93 thousand rows/s, 39.77 MiB/s)

• fixes: #[Link the issue here]

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test - Explain why

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

---

This change is 

[drmingdrmer]

Reviewed 10 of 29 files at r1, all commit messages.
Reviewable status: 10 of 29 files reviewed, 1 unresolved discussion (waiting on @b41sh and @sundy-li)

---

src/meta/proto-conv/src/virtual_column_from_to_protobuf_impl.rs line 54 at r1 (raw file):

            for (v, ty) in p.virtual_columns.iter().zip(p.data_types.iter()) {
                virtual_columns.push((v.clone(), TableDataType::from_pb(ty.clone())?));
            }

just a nit:
The length of virutal_columns and data_types should be identical but it's still be better to assert it here and return an Incompatible error if it does not match.

Code quote:

            for (v, ty) in p.virtual_columns.iter().zip(p.data_types.iter()) {
                virtual_columns.push((v.clone(), TableDataType::from_pb(ty.clone())?));
            }