Re: Writing new unit tests with PostgresNode

On 23 February 2016 at 09:52, Michael Paquier <michael(dot)paquier(at)gmail(dot)com>
wrote:

> On Tue, Feb 23, 2016 at 12:29 AM, Alvaro Herrera
> <alvherre(at)2ndquadrant(dot)com> wrote:
> > Craig Ringer wrote:
> >
> >> > +=pod
> >> > +
> >> > +=head2 Set up a node
> >> > pod format... Do we really want that? Considering that those modules
> >> > are only aimed at being dedicated for in-core testing, I would say no.
> >>
> >> If it's plain comments you have to scan through massive piles of verbose
> >> Perl to find what you want. If it's pod you can just perldoc
> >> /path/to/module it and get a nice summary of the functions etc.
> >>
> >> If these are intended to become usable facilities for people to write
> tests
> >> with then I think it's important that the docs be reasonably accessible.
> >
> > Yes, I think adding POD here is a good idea. I considered doing it
> > myself back when I was messing with PostgresNode ...
>
> OK, withdrawal from here. If there are patches to add that to the
> existing tests, I'll review them, and rebase what I have depending on
> what gets in first. Could a proper patch split be done please?
>

Just finished doing that. Thanks for taking a look at the first patch so
quickly. I hope this one's better.

FWIW I think you were right that using pod for the actual test wasn't the
best choice, I should've just used comments. I do think it's important for
the modules to have structured docs.

I've removed the example suite in favour of adding a SYNOPSIS section to
PostgresNode.pm and describing the rest in the README. It won't be
necessary once your replication tests go in, they'll be a perfectly
adequate example.

I also cut out the changes to the backup method; I'll send a pull to add to
your proposed replication patch instead.

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services