Reconsider how "rethinkdb export" calculates progress

[timmaxw]

Currently rethinkdb export runs r.table(...).count() and then r.table(...). As it reads documents from the stream returned by r.table(...), it counts the number of documents; then it uses this value as the numerator of the progress fraction, with the result of r.table(...).count() as the denominator.

This means that the server must traverse the entire data set twice: once to count the number of documents, and again to stream them to the client. For a data set that doesn't fit into RAM, this could make rethinkdb export take twice as long as it needs to. Until the first traversal is finished, rethinkdb export will report 0% progress.

We should consider using a distribution query to estimate the denominator for the progress fraction instead. We can do this by running r.table(...).info()['doc_count_estimates'].sum(). This won't give exact results, but it will run in constant time instead of reading the entire table.

[Tryneus]

This is up in review 2426.

[Tryneus]

This has been approved and merged into next in commit 92c7390. Will be in release 1.16.