Add Repo.select_all

[dkuku]

{count, result} = Repo.select_all(query)

I proposed this on the Google Group but didn't receive any replies. This has been sitting on my disk, so I'll try to at least get it upstream.

I think this is a missing piece — we have insert_all, delete_all, and update_all, all of which return the count, but there's no equivalent for select. Currently, Repo.all has the count available internally but calls elem(1) on the tuple, so when we actually need the count we have two options:

1. Calling Enum.count/1 on the results.
2. Running a separate Repo.aggregate.

I think this fits nicely inside the current api - you have either the long version select_all that gives you the full result with count, or the short version all that gives you only the result.
The count is useful with pagination
Examples:
https://github.com/duffelhq/paginator/blob/main/lib/paginator.ex#L329 uses Enum.count
https://github.com/mojotech/scrivener_ecto/blob/main/lib/scrivener/paginater/ecto/query.ex#L71 uses sql count
https://github.com/ash-project/ash/blob/main/lib/ash/actions/read/relationships.ex#L1617 uses Enum.split(limit) and checks if there is anything left

[josevalim]

Thank you but unfortunately I don't believe it is worth enlarging the Ecto API for this functionality. Doing a length again is not ideal but unlikely to be problematic compared to all of the cost of going over the wire and instantiating all of the data returned by a query.