QE: Batch huge IN queries into smaller batches

[pimeys]

In all our supported databases a big enough IN query will panic and crash due to the database not allowing too large set of variables. It's especially easy to replicate with a big enough database due to the relation queries joining the data with a huge IN statement.

Example:

query A {
  findManyUser { address { street } }
}

If having thousands of users (between 5000-10000 or more), the address query will be including a filter with user_id IN (thousands of ids) and the database will error, causing a panic.

We could instead load the addresses in batches of maximum 5000 ids and concatenating the results in the connector.

Things to be aware of:

• Databases will dedup ids in an IN statement. For batched statements, the deduplicating should be done in the connector
• Ordering will cause trouble. For batched queries, the ordering needs to happen in the connector.