replace QgsMapRenderer #3809

pblottiere · 2016-11-28T15:04:16Z

The aim of this PR is to remove the QgsMapRenderer by using QgsMapRendererCustomPainterJob for qgis/QGIS-Enhancement-Proposals#74.

Moreover, some unit tests have been added.

My starting point was this PR made by @mhugent : #3129

Note that I wanted to use

renderJob.start();
renderJob.waitForFinished()

But there's some timeout in unit tests... So I used renderJob.renderSynchronously(); instead.

Let me know if you have any comments!

nyalldawson

Fantastic work @mhugent and @pblottiere -- looks good to me and great to see the extra tests added too

nyalldawson · 2016-11-28T22:20:33Z

python/server/qgsmsutils.sip

+%End
+
+//! Some utility functions that may be included from everywhere in the code
+namespace QgsMSUtils


With QGIS 3.0 we are trying to follow Qt's name capatilization conventions, so this should be QgsMsUtils. But we're also trying to remove all abbreviated names so QgsMapServerUtils would be best.

nyalldawson · 2016-11-28T22:22:05Z

python/server/qgsmsutils.sip

+
+  struct QgsLayerCoordinateTransform
+  {
+    QString srcAuthId;


As above, should be sourceAuthId and destinationAuthId. (As explanation we are doing this to make the api more predictable - in the past we mix up things like dst/dest/destination which means you're never sure which to use without having to check the docs.

nyalldawson · 2016-11-28T22:24:33Z

src/core/qgsmaprendererjob.h

+    //! @note Added in QGIS 3.0
+    //! Set the feature filter provider used by the QgsRenderContext of
+    //! each LayerRenderJob.
+    void setFeatureFilterProvider( const QgsFeatureFilterProvider *f ) { mFeatureFilterProvider = f; };


Would be good to note in the docs about the lifetime of the provider here - something like ownership is not transferred and the provider must not be deleted before the render job.

nyalldawson · 2016-11-28T22:27:00Z

src/server/qgsserver.cpp

@@ -61,7 +62,7 @@

 QString* QgsServer::sConfigFilePath = nullptr;
 QgsCapabilitiesCache* QgsServer::sCapabilitiesCache = nullptr;
-QgsMapRenderer* QgsServer::sMapRenderer = nullptr;
+QgsMapSettings* QgsServer::sMapSettings = nullptr;


What's the motivation behind a static QgsMapSettings object? Why not just create one per map render?

nyalldawson · 2016-11-28T22:27:49Z

src/server/qgssldconfigparser.cpp

@@ -724,20 +724,11 @@ bool QgsSLDConfigParser::wmsInspireActivated() const
  return false;
 }

-QgsComposition* QgsSLDConfigParser::createPrintComposition( const QString& composerTemplate, QgsMapRenderer* mapRenderer, const QMap< QString, QString >& parameterMap, QStringList& highlightLayers ) const
+QgsComposition* QgsSLDConfigParser::initComposition( const QString& composerTemplate, QgsMapSettings* mapSettings, QList< QgsComposerMap*>& mapList, QList< QgsComposerLegend* >& legendList, QList< QgsComposerLabel* >& labelList, QList<const QgsComposerHtml *>& htmlFrameList ) const


I'd make the mapSettings argument a reference here - it's required and composer will crash if a nullptr is passed.

nyalldawson · 2016-11-28T22:31:00Z

src/server/qgswmsserver.cpp

+  context.setScaleFactor( mConfigParser->outputUnits() == QgsUnitTypes::RenderUnit::RenderMillimeters ? mMapSettings->outputDpi() / 25.4 : 1.0 );
+  context.setRendererScale( mMapSettings->scale() );
+  context.setMapToPixel( mMapSettings->mapToPixel() );
+  context.setExtent( mMapSettings->extent() );


I think this should use QgsRenderContext::fromMapSettings initially, and then override the things it needs for server. It's less maintenance/more future proof that way.

nyalldawson · 2016-11-28T22:32:09Z

src/server/qgswmsserver.cpp

  {
    QgsVectorLayer* vl = qobject_cast<QgsVectorLayer*>( QgsMapLayerRegistry::instance()->mapLayer( layerID ) );
    if ( !vl || !vl->renderer() )
      continue;

-    if ( vl->hasScaleBasedVisibility() && ( mMapRenderer->scale() < vl->minimumScale() || mMapRenderer->scale() > vl->maximumScale() ) )
+    if ( vl->hasScaleBasedVisibility() && ( mMapSettings->scale() < vl->minimumScale() || mMapSettings->scale() > vl->maximumScale() ) )


Use QgsMapLayer::isInScaleRange here instead.

nyalldawson · 2016-11-28T22:33:09Z

src/server/qgswmsserver.cpp

@@ -1699,13 +1704,13 @@ int QgsWmsServer::getFeatureInfo( QDomDocument& result, const QString& version )

  //Render context is needed to determine feature visibility for vector layers
  QgsRenderContext renderContext;
-  if ( mMapRenderer )
+  if ( mMapSettings )


This should also use QgsRenderContext::fromMapSettings

m-kuhn · 2016-11-28T22:47:53Z

src/server/qgsmsutils.h

@@ -20,6 +20,13 @@
 //! Some utility functions that may be included from everywhere in the code
 namespace QgsMSUtils
 {
+  struct QgsLayerCoordinateTransform
+  {
+    QString srcAuthId;


Would be great if you could add some doxygen comments to highlight the meaning of these variables.

nyalldawson · 2016-11-30T23:40:22Z

src/server/qgswmsserver.cpp

@@ -992,7 +992,7 @@ void QgsWmsServer::runHitTest( const QgsMapSettings& mapSettings, QPainter* pain
    if ( !vl || !vl->renderer() )
      continue;

-    if ( vl->hasScaleBasedVisibility() && ( mapSettings.scale() < vl->minimumScale() || mapSettings.scale() > vl->maximumScale() ) )
+    if ( vl->hasScaleBasedVisibility() && vl->isInScaleRange( mapSettings.scale() ) )


the first test hasScaleBasedVisibility is included in isInScaleRange and can also be safely removed

nyalldawson · 2016-11-30T23:41:35Z

Looks great to me! +1 from my side, but a review by @wonder-sk would be good too.

wonder-sk

Great stuff - apart from few comments/questions I have the PR looks good to me.

wonder-sk · 2016-12-01T02:10:28Z

src/server/qgswmsserver.cpp

+    QgsMapSettings mapSettings;
+    legendSettings.setMapScale( mapSettings.scale() );
+    double scaleFactor = mConfigParser->outputUnits() == QgsUnitTypes::RenderUnit::RenderMillimeters ? mapSettings.outputDpi() / 25.4 : 1.0;
+    legendSettings.setMmPerMapUnit( 1 / ( mapSettings.mapUnitsPerPixel() * scaleFactor ) );


Does this bit of code work? It looks like a temporary map settings instance is constructed to get configuration from it (which will be just some default values)

wonder-sk · 2016-12-01T02:21:14Z

src/server/qgswmsserver.cpp

-  context.setRendererScale( mMapRenderer->scale() );
-  context.setMapToPixel( *mMapRenderer->coordinateTransform() );
-  context.setExtent( mMapRenderer->extent() );
+  context.setRasterScaleFactor(( thePaintDevice->logicalDpiX() + thePaintDevice->logicalDpiY() ) / 2.0 / mapSettings.outputDpi() );


It should not be necessary to set "raster scale factor" anymore and it should always stay equal to 1. It is a legacy used with QgsMapRenderer.

Next step would be to actually remove it from QgsRenderContext finally (in some other PR).

wonder-sk · 2016-12-01T02:23:09Z

src/server/qgswmsserver.cpp

-  context.setMapToPixel( *mMapRenderer->coordinateTransform() );
-  context.setExtent( mMapRenderer->extent() );
+  context.setRasterScaleFactor(( thePaintDevice->logicalDpiX() + thePaintDevice->logicalDpiY() ) / 2.0 / mapSettings.outputDpi() );
+  context.setScaleFactor( mConfigParser->outputUnits() == QgsUnitTypes::RenderUnit::RenderMillimeters ? mapSettings.outputDpi() / 25.4 : 1.0 );


The scale factor should have been already set to a correct value in QgsRenderContext::fromMapSettings()

wonder-sk · 2016-12-01T02:28:40Z

python/server/server.sip

@@ -8,6 +8,7 @@
 %Import QtXml/QtXmlmod.sip

 %Import core/core.sip
+%Include core/qgsdatumtransformstore.sip


Why do we include an extra sip file from "core"?

wonder-sk · 2016-12-01T02:33:59Z

src/server/qgsserver.h

@@ -117,7 +117,7 @@ class SERVER_EXPORT QgsServer
    // Status
    static QString* sConfigFilePath;
    static QgsCapabilitiesCache* sCapabilitiesCache;
-    static QgsMapRenderer* sMapRenderer;
+    static QgsMapSettings* sMapSettings;


Is this still used anywhere? Not sure if we need a static map settings instance...

pblottiere · 2016-12-02T17:42:09Z

@nyalldawson @m-kuhn @wonder-sk

Thank you for your comments! I've updated the code accordingly.

pblottiere · 2016-12-02T17:43:42Z

src/server/qgswmsserver.cpp

+#endif
+    renderJob.renderSynchronously();
+    //renderJob.start();
+    //renderJob.waitForFinished();
  }


A timeout is raised during unit tests when the start/waitForFinished functions are used. Do you know why?

@pblottiere just guessing: maybe you miss an event loop? (A local event loop in the test might help if that's the case, connect to https://qgis.org/api/classQgsMapRendererJob.html#a8dbfbe2cc381f0eb6e721733838d3baf)

pblottiere · 2016-12-02T17:48:07Z

tests/src/python/test_qgsserver.py

+
+    #    r, h = self._result(self.server.handleRequest(qs))
+    #    self._img_diff_error(r, h, "WMS_GetLegendGraphic_BBox2")
+


Since the rebase, I didn't manage to use the getLegendGraphic request with a BBOX parameter. Each time, I have an empty response. So i disabled my tests for now.

What am i missing?

pblottiere force-pushed the render branch from 1f392b1 to 4ffed96 Compare November 28, 2016 15:18

nyalldawson reviewed Nov 28, 2016

View reviewed changes

nyalldawson mentioned this pull request Nov 28, 2016

Server renderjob #3129

Closed

m-kuhn reviewed Nov 28, 2016

View reviewed changes

nyalldawson reviewed Nov 30, 2016

View reviewed changes

wonder-sk reviewed Dec 1, 2016

View reviewed changes

pblottiere force-pushed the render branch from 545c8ee to bba5c95 Compare December 2, 2016 17:32

pblottiere commented Dec 2, 2016

View reviewed changes

pblottiere force-pushed the render branch 2 times, most recently from 8ea9b83 to e8cb233 Compare December 5, 2016 13:49

rldhont mentioned this pull request Dec 6, 2016

QEP 74: QGIS server code refactoring for QGIS 3.0 qgis/QGIS-Enhancement-Proposals#74

Open

9 tasks

pblottiere force-pushed the render branch 3 times, most recently from a46b3e2 to 3dcb3cd Compare December 12, 2016 00:05

remove QgsMapRenderer

f3a5c98

pblottiere force-pushed the render branch from 3dcb3cd to 4d76308 Compare December 12, 2016 06:51

fix getLegendGraphics when BBox parameter is used

8cda1bb

pblottiere force-pushed the render branch from 4d76308 to 8cda1bb Compare December 12, 2016 07:16

rldhont merged commit 7028eee into qgis:master Dec 12, 2016

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

replace QgsMapRenderer #3809

replace QgsMapRenderer #3809

pblottiere commented Nov 28, 2016 •

edited

Loading

nyalldawson left a comment

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

nyalldawson Nov 28, 2016

m-kuhn Nov 28, 2016

nyalldawson Nov 30, 2016

nyalldawson commented Nov 30, 2016

wonder-sk left a comment

wonder-sk Dec 1, 2016

wonder-sk Dec 1, 2016

wonder-sk Dec 1, 2016

wonder-sk Dec 1, 2016

wonder-sk Dec 1, 2016

pblottiere commented Dec 2, 2016 •

edited

Loading

pblottiere Dec 2, 2016

elpaso Dec 4, 2016

pblottiere Dec 2, 2016


		# r, h = self._result(self.server.handleRequest(qs))
		# self._img_diff_error(r, h, "WMS_GetLegendGraphic_BBox2")

replace QgsMapRenderer #3809

replace QgsMapRenderer #3809

Conversation

pblottiere commented Nov 28, 2016 • edited Loading

nyalldawson left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nyalldawson commented Nov 30, 2016

wonder-sk left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pblottiere commented Dec 2, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pblottiere commented Nov 28, 2016 •

edited

Loading

pblottiere commented Dec 2, 2016 •

edited

Loading