Euler Taveira

The question is: Is it necessary to have a dozen of ssh connection for parallel copy? I have a gut feeling that you saturate network and/or disk with a few...

@abrkn Could you a test case?

@rafalcieslak You should use CREATE OR REPLACE VIEW to avoid breaking other objects that depends on that view. Take a look at the function implementation. Since 9.3 the pg_get_viewdef() changes...

> However, choosing this strategy means that in order to correctly deal with dependencies, one would need to drop such dependent objects, drop&create the view, and then recreate the objects....

@abrkn pgquarrel supports it. The problem is that the function is created **after** the table. I will figure out a way to rearrange them.

@roytan883 this is not a bug. The DDL is correct. You can claim it is a new feature. `pgquarrel` orders table columns by name to compare columns (of course in...

> The problem is that：pgquarrel works great, but it can't handle POLICY. So after use diff.sql, I must manually pg_dump source and target DB, and use other diff tools to...

@roytan883 open another issue. It is a totally different problem. Please provide a small test case.

@barkingfoodog Why don't you use `appendStringLiteralDQ` instead of duplicating `pg_dump` code ?

I pushed the foreign key commit.