Create columns with tag semantic type

[killme2008]

What type of enhancement is this?

API improvement, User experience

What does the enhancement do?

If the users write data by influxdb, opentsdb or Prometheus remote write, GreptimeDB will auto-create the table with the correct semantic types:

• SemanticType::Tag for tags(or labels in Prometheus)
• SematicType::Field for fields(or values)
• SematicType::Timestamp for time index column

But when using SQL to create table, it doesn't provide a way to create columns with tag semantic type,all the columns not in the primary key except time index are field semantic type.

I think we can provide a SQL extension to create tables with tag columns, for example:

CREATE TABLE IF NOT EXISTS system_metrics (
    host STRING TAG,
    idc STRING TAG,
    cpu_util DOUBLE,
    memory_util DOUBLE,
    disk_util DOUBLE,
    ts TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    TIME INDEX(ts)
);

It's the same as:

CREATE TABLE IF NOT EXISTS system_metrics (
    host STRING,
    idc STRING,
    cpu_util DOUBLE,
    memory_util DOUBLE,
    disk_util DOUBLE,
    ts TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    primary key(host, idc),
    TIME INDEX(ts)
);

Then the host and idc will be created with tag semantics. The columns except time index without TAG option will be created with field semantics.

Implementation challenges

• Make the column option parser to support tag option.
• Pass the tag semantic type to create table request.

[NiwakaDev]

I'd like to work on this issue, but do I need to wait until migrate-region-server is merged?

[MichaelScofield]

@NiwakaDev You can checkout a branch from migrate-region-server and implement the feature on top of it. We will finally make the migrate-region-server branch the new "develop" branch.

[killme2008]

Thank you. But we don't want this feature yet. It needs more discussion.
The primary key constraint looks good enough to represent the columns in tag semantics and it contains the columns ordering info.
Not sure if we need a new syntax for it. As you know, less is more.

[NiwakaDev]

Until this discussion land, I think that we can add a section about relationship between primary keys and tags in the document so that user can easily understand:
GreptimeTeam/docs#452

[killme2008]

Until this discussion land, I think that we can add a section about relationship between primary keys and tags in the document so that user can easily understand: GreptimeTeam/docs#452

Can't agree more with it, thank u.