[QGIS-Developer] only read name and group name of processing algorithms during startup
Rashad Kanavath
mohammedrashadkm at gmail.com
Tue Apr 3 07:06:17 PDT 2018
Hello Nyall,
On Tue, Mar 20, 2018 at 10:55 PM, Nyall Dawson <nyall.dawson at gmail.com>
wrote:
> On 19 March 2018 at 18:39, Rashad Kanavath <rashad.kanavath at c-s.fr> wrote:
> > Nyall,
> >
> > could we have defineCharacteristicsFromFile method in
> > QgsProcessingAlgorithm ?.
> > the diff with canExecute is more of a working draft.
>
>
> Hi Rashad,
>
> Can you open a PR? I'm having trouble following the change as the
> whitespace is messed up in the email.
>
Here we go..
https://github.com/qgis/QGIS/pull/6739
>
> Thanks!
> Nyall
>
> >
> >
> >
> > On Thu, Mar 15, 2018 at 03:27:50PM +0100, Rashad Kanavath wrote:
> >> Hello,
> >>
> >> On Wed, Mar 14, 2018 at 4:38 AM, Nyall Dawson <nyall.dawson at gmail.com>
> >> wrote:
> >>
> >> > On 13 March 2018 at 20:56, Rashad Kanavath <rashad.kanavath at c-s.fr>
> wrote:
> >> > > if a provier is enabled then qgis reads all algorithm's during
> >> > > startup. This involves reading descriptor files and doing all
> parsing
> >> > > required to make gui when user actually open algorithm (click on
> >> > > treeitem).
> >> > > So If I have like three provider enabled it takes more time to start
> >> > qgis than
> >> > > usual. Even though, I might open one or two later at some point.
> >> > >
> >> > > It is not necessary here to parse "all" data during startup. name
> and
> >> > > group name is only needed to fill provider-algorithm tree. This is
> >> > > true for any provider which uses descriptor files (OTB, GRASS, SAGA
> >> > etc..).
> >> >
> >> > Actually it's a bit trickier here -- models make things more complex
> >> > as they self-validate on creation, and need full knowledge of the
> >> > dependent algorithm's parameters and outputs.
> >> >
> >>
> >> Yes. there are different cases like modeler, standard, batch, scripts.
> >> I tried to utilize canExecute() in otb and seems to be working.
> >> You already have a very flexible, strong API in qgis processing such
> >> trickier feature like this one has become too easy!!.
> >>
> >> I had to call alg.canExecute() in two places and that's it.
> >> diff attached below. With this loading time will be hopefully improved
> for
> >> OTB, SAGA, GRASS GIS.
> >>
> >> index f5bdc40dc5..1cca57769a 100644
> >> --- a/python/plugins/processing/modeler/ModelerParametersDialog.py
> >> +++ b/python/plugins/processing/modeler/ModelerParametersDialog.py
> >> @@ -70,6 +70,7 @@ class ModelerParametersDialog(QDialog):
> >> QDialog.__init__(self)
> >> self.setModal(True)
> >> # The algorithm to define in this dialog. It is an instance of
> >> QgsProcessingModelAlgorithm
> >> + alg.canExecute()
> >> self._alg = alg
> >> # The model this algorithm is going to be added to
> >> self.model = model
> >> @@ -80,6 +81,9 @@ class ModelerParametersDialog(QDialog):
> >> settings = QgsSettings()
> >> self.restoreGeometry(settings.value("/Processing/
> >> modelParametersDialogGeometry", QByteArray()))
> >>
> >> + def algorithm(self):
> >> + return self._alg
> >> +
> >> def closeEvent(self, event):
> >> settings = QgsSettings()
> >> settings.setValue("/Processing/modelParametersDialogGeometry",
> >> self.saveGeometry())
> >> @@ -242,7 +246,6 @@ class ModelerParametersDialog(QDialog):
> >> outTypes = []
> >> elif not isinstance(outTypes, (tuple, list)):
> >> outTypes = [outTypes]
> >> -
> >> return self.model.availableSourcesForChild(self.childId,
> >> [p.typeName() for p in paramType if
> >>
> >> issubclass(p, QgsProcessingParameterDefinition)],
> >> [o.typeName() for o
> in
> >> outTypes if
> >> diff --git a/src/core/processing/models/qgsprocessingmodelchildalgorit
> hm.cpp
> >> b/src/core/processing/models/qgsprocessingmodelchildalgorithm.cpp
> >> index f5f8074c88..01906a7f62 100644
> >> --- a/src/core/processing/models/qgsprocessingmodelchildalgorithm.cpp
> >> +++ b/src/core/processing/models/qgsprocessingmodelchildalgorithm.cpp
> >> @@ -58,7 +58,7 @@ QgsProcessingModelChildAlgorithm &
> >> QgsProcessingModelChildAlgorithm::operator=( c
> >>
> >> const QgsProcessingAlgorithm *QgsProcessingModelChildAlgorit
> hm::algorithm()
> >> const
> >> {
> >> - return mAlgorithm.get();
> >> + return mAlgorithm && mAlgorithm->canExecute() ? mAlgorithm.get():
> >> nullptr;
> >> }
> >>
> >> void QgsProcessingModelChildAlgorithm::setModelOutputs( const
> >> QMap<QString, QgsProcessingModelOutput> &modelOutputs )
> >>
> >> test code for otb provider:
> >> https://gitlab.orfeo-toolbox.org/orfeotoolbox/qgis-otb-plugin
> >> I can add modification to SAGA and GRASS if you are okay to move this
> to PR.
> >>
> >>
> >> > > Issue is more about calling defineDescriptorFile() in Algorithm's
> >> > > __init__ method. And ofcourse, one cannot avoid init when adding
> >> > > algorithm and tree need to add an instance of algorithm. what can be
> >> > > done?
> >> >
> >> > What I've been thinking/planning is to take advantage of task manager
> >> > here. So basically:
> >> >
> >> > - on processing startup, there's no providers added
> >> > - when the qgis interface is fully loaded then we fire up a background
> >> > task which loads the providers, allowing them to fully build their
> >> > available algorithms and parameters without blocking the startup
> >> > - after the task completes, the tree is populated
> >> >
> >>
> >> this is still loading all algorithm with parsing but things will be in
> >> background to not block ui for users. correct?
> >> In that case, you could include above diff and then add loading in
> >> background task. IMHO that will be best.
> >>
> >>
> >> > I'd like to see more plugins take this approach, so that qgis can load
> >> > nice and quick and the slow stuff happens without annoying users...
> >> >
> >> > Nyall
> >> > _______________________________________________
> >> > QGIS-Developer mailing list
> >> > QGIS-Developer at lists.osgeo.org
> >> > List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> >> > Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> >> >
> >>
> >>
> >>
> >> --
> >> Regards,
> >> Rashad
> >
> > --
> > Take care,
> > Rashad
>
--
Regards,
Rashad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20180403/3c7e5803/attachment.html>
More information about the QGIS-Developer
mailing list