Allow autogeneration of singular data sources

[melinath]

Affected Resource(s)

• google_*

Data sources are currently all handwritten; however, they are fairly simple. We should allow autogeneration so that we can more cheaply support the existing data sources and also more easily add new ones.

[rileykarson]

Discussed in chat recently after @ScottSuarez brought this up. Context from there: most singular datasources (i.e. data.google_compute_address are of limited utility, and this has not proven the best use of our time to work on. Plural datasources (i.e. data.google_compute_addresses) would be slightly more useful in conventional configurations.

The most value comes from custom datasources generally, though, without corresponding resources.

[ScottSuarez]

My concern with datasources implemented manually is that sometimes they don't use the resource read or even schema. This leads to datasources becoming out of date with resource implementation. I would rather have flag to flip to turn on datasource* generation if we want to keep it an opt-in thing. Thus ensuring consistency in implementation design and less foreseeable maintenance.

[rileykarson]

Yes, older datasources were handwritten. We avoid those now for the reasons you've mentioned and use the datasource wrapper for singular datasources. That's a separate topic than my comment which addresses why generating these has not been particularly high priority- singular datasources are not particularly valuable for end users in most cases.

I would rather have flag to flip to turn on dataset generation if we want to keep it an opt-in thing.

Datasource?

[ScottSuarez]

mistype, mm gotcha !