[QGIS-Developer] Get rid of Processing scripts in favour of plain alogrithms

[Nyall Dawson]

On 31 January 2018 at 18:24, G. Allegri <giohappy at gmail.com> wrote:
> I understand your point Victor, and I agree that scripts was a clever idea.
> But:
>
>  - with the current shape of Processing (QGIS 3.0) I think the "syntactic
> sugar" provided by scripts has less relevance. As I can see from the
> refactoring, there's less automation in parameters conversions and
> management, and a few new "magic" context variables have been introduced. I
> think scripts now are too similar to plain geoalgorithms, and consequently
> the differences can become misleading and not easily understood.
>
>  - syntactic sugar requires maintanance: if a new parameter is introduced, i
> parameters are added or changed, the corresponding translation method for
> scripts must be updated.
>
>  - syntactic sugar requires doc maintanance, while Processing APIs
> documentation can be mostly automated.
>
> Anyway, this is a proposal to be discussed. Meanwhile I will try to estimate
> the work needed to drop (and adapt) the current implementations.
>
> PS: @Victor, it's nice to follow Processing's history! C++ (SAGA) -> Java
> (Sextante) -> Python (Processing) -> C++ (QGIS 3.0)  :D

I'm a bit late to this discussion, but my 2c:

- I actually like the idea of single file script algorithms, and don't
want to see them go. I can see lots of valid use cases for why they
are needed. (In particular - I'd love to extend models in future to
allow embedded script algorithms which are contained just within a
single model, and not exposed elsewhere).

- It's just the syntax of scripts which I'm not a fan of (for reasons
well covered above)

Alex has a WIP PR already at https://github.com/qgis/QGIS/pull/6225  -
stakeholder comments would be welcome here!

Nyall