Add optional label attribute protocols

[QMalcolm]

dbt-core currently allows for specifying an optional label property on metric nodes. However we don't currently specify this on the metric protocol. MetricFlow would like to begin using this property, and would like it to exist on more objects.

We should add an optional label property to the following protocols:

• Metric
• SemanticModel
• Dimension
• Measure

[QMalcolm]

@nhandel should it be added to any other protocols?

[tlento]

Do we need SemanticModel? It doesn't matter much to me either way, I think Metric, Dimension, and Measure are the most important.

[tlento]

In the originally filed #139 users are requesting the label property be added to entities as well, so we need to consider how that might work.

[ataft]

I think it's worth the extra effort to have consistency and add label to entities. At the bare minimum, every object in the semantic layer should have:

1. id/code (dbt calls this name)
2. label/title
3. description

Semantic layers should also have attributes for synonyms, formatting, and more, but I understand dbt doesn't want to be too opinionated. Like others, I'll just use the meta attribute for these. I don't see meta in the docs, but it's in the original spec proposal.

[nhandel]

@QMalcolm this should also be added to the Entity object. We'll also need to add validation that each Entity with the same name either has the same label specified.

[tlento]

@ataft support for meta isn't available yet, but it is coming. You can follow along here: dbt-labs/dbt-core#8511

[ataft]

@tlento I see from that issue that meta is being added for semantic models, but not dimensions and measures?

I was thinking/hoping that the new semantic layer could act as a "universal semantic layer" where I define everything (label, description, aggtype, meta for formatting and NLQ synonyms, etc.) that downstream tools could then just plug into. I was also hoping that the semantic layer YAML would replace the model and columns YAML, so I wouldn't need to define model columns and then separate measures and dimensions (seems redundant). Is the expectation that people will define a model and columns in one schema YAML file, then also define a semantic model with dimensions and metrics?

[emmyoop]

We've added labels in dbt-labs/dbt-core#8646. It is no longer a blocker on labels being supported.