Add a where clause to support soft delete

[ankitml]

Why

Tons of databases in the wild are bloated not only by dead tuples, also by zombie tuples. ie Data that exists in table just because it is hard to cleanup and prune the old data.
Pg_repack is everyone's favourite tool to clean up dead tuples. It can easily clean up data that is not needed with a new flag.
--where-clause="deleted_at IS NOT NULL". The where clause is generic and can reference to foreign tables as well.
To keep data that is updated in last 90 days in mytable.
--table="<mytable>" --where-clause="updated_at > NOW() - Interval '90 days'"

This PR adds Data cleanup and soft-delete support with pg_repack.
pg_repack --dbname="ankitmittal" --table="test_repack" --echo --elevel=DEBUG --where-clause="deleted_at IS NOT NULL"
pg_repack --dbname="ankitmittal" --table="test_repack" --echo --elevel=DEBUG --where-clause="updated_at < NOW() - Interval '90 days'"

This has been discussed before (#279) with a different approach.

Why not

It could cause data loss if used incorrectly.

If used properly this cleans up logical bloat ie data that is thrown in database but not cleaned up. An alternative here is to perform repack-like online table-swap manually.

What it doesnt do

Incoming stream of data while repack is running is left as it is for sake of simplicity.

[ankitml]

@fabriziomello curious to know what are your thoughts on this

[andreasscherbaum]

There is a bin/.idea/workspace.xml file committed in this PR which does not belong there.

[andreasscherbaum]

It could cause data loss if used incorrectly

How can this be prevented?

[ankitml]

Thanks, removed the workspace file

[ankitml]

It could cause data loss if used incorrectly

How can this be prevented?

The arguments to where clause as specified by the user is intended to delete data. Users would need to verify if data being deleted is what they want to delete. Dry run of the pg-repack shows precisely which data is going to be remain. It is recommended to think through the outputs of dry run before running real command.

[ankitml]

We have been running this repack at instacart on few tables weekly to clean up zombie tuples for last 4 months. Without any manual intervention, on a cron.

[andreasscherbaum]

@ankitml Can you add unit tests to ensure this is working as intended?
Please also add tests which have ... let's say unusual where clauses. Like something which is broken, which clearly should not be in a where clause, something with spaces or special characters. Anything which can break the query.