spanner: ReadOnlyTransaction().ReadRowUsingIndex()

[majelbstoat]

ReadRow is useful method for getting a single row. However, if you want to get all the columns of a single row using an index (a common case for unique fields, like, say finding a user by username), it's much harder. ReadUsingIndex requires that you define STORING for every column name, which is duplicative and brittle to maintain (remembering to add every new column as you go*) .

The alternative is to use a SELECT ... {FORCE_INDEX=index} statement, and call Query. Both this and ReadUsingIndex require the handling of an iterator which you need to process for just a single item, iterator.Done for Not Found etc.

It would be helpful if the APIs for such similar queries weren't so wildly different.

I would like to see ReadRowUsingIndex, which under the hood did the same kind of join as SELECT {FORCE_INDEX} query does, and which returns a single row (technically the first) as a result. I wouldn't even mind if it was enforced that you could only call this on a UNIQUE index. This would dramatically reduce boilerplate, which already has to be quite substantial due to Go's lack of generics.

--

(* meta-point, for those cases where I really do want to do this, I'd love to see a language extension like STORING ALL or something similar. DynamoDB offers this for its Global Indices, for example.)

[jba]

Thanks for your suggestion.

@snehashah16 thoughts?

[snehashah16]

One other way to make this easy, expose the ExecuteQuery (non streaming) version to avoid iterating on resultSets.

[odeke-em]

Thank you for this report! @AlisskaPie and @IlyaFaer are working on this with CL https://code-review.googlesource.com/c/gocloud/+/43190