[Qgis-psc] [qgis/QGIS-Enhancement-Proposals] Support SAP HANA databases in QGIS (#151)

[Nyall Dawson]

On Thu, 5 Sep 2019 at 09:15, Tim Sutton <tim at kartoza.com> wrote:
>
>
> @Nyall we discussed this a bit in Bucharest. There are a few concerns - figuring out how to reasonably manage testing infrastructure as Even touches on below being one, the lack of any clear directive mandating tests (I think we should add something to our contributor guidelines),

> the lack of any precedent test suite (previously we only mandated libqgiscore tests) being another.

This is far from true. EVERY vector data provider, even the infamous
db2 provider, implements the standard provider test suite. The vast
majority of these are tested using travis, including Postgres (sql
server was for a stage, but the docker image of sql server was too
fragile and it's been temporarily disabled, waiting on an updated
image for testing).

Furthermore, there's clearly a strong precedent regarding testing of
contributions. It may not be mandated, but its only very rare
exceptions (or gui heavy contributions) which are merged today without
unit tests. I think it's time we mandated this, AND also mandated the
forced use of pull requests and mandatory reviews for all
contributions. The vast majority of the QGIS developer community just
follow this as standard practice now, so it's unlikely to have any
significant effect on QGIS contributions.

> Personally I wonder if we shouldn’t be thinking more broadly and have a generic test suite that works with all providers to validate general CRUD operations?

We do already ;) It's been around since the first hackfest in Nodebo,
and has grown to many thousands of lines of code. It's also been
single-handedly responsible for the dramatic increase in provider
stability and consistency we've seen over the last few years.

See https://github.com/qgis/QGIS/blob/master/tests/src/python/providertestbase.py
and https://github.com/qgis/QGIS/blob/master/tests/src/python/featuresourcetestbase.py

Nyall