I've wrote 1 treeView/model system last year that had it all re-implemented from scratch, a few days ago I decided to rewrite it to increase performance and relay more on native functions of QT. Which I did and works great but the margin thing is a sticking point. I remember how much work it was before to get it to work properly... took a while as I had to write quite a few functions before I got to the point of margin detection. I'm very surprised there is no

margin = (qt code ) + userMarginOverrideVal

in the canDrop() function... this would have allowed us to just say, hey add +5 pixels to detection or something like that...

Perhaps I should add it to feature request, would posting it in a bug report tracker be a good place? I don't see feature requests and I have no idea how to get the source to compile (Just didn't spend enough time on it) and add it myself & post to qt as commit...

As far as I can tell we need like 2 functions, setOverrideMargin(int val) and getOverrideMargin(), skipping setting flag true/false if we want to use it as defaultVal should just be 0...

TIA

@raven-worx said in QTreeView/Model drop action - no expand sign ?:

Also did you emit the dataChanged() signal on the parent index?

I've just read the post. I forgot to correct it. This is what I edit:

bool testModel::dropMimeData(const QMimeData *data, Qt::DropAction action, int row, int column, const QModelIndex &parent) { ... Q_EMIT dataChanged(itemFirst.index(),itemLast.index()) }

Hmmmm so I tried this, the parent is the object that is in function above.

Q_EMIT dataChanged(parent, parent)

Now this now "magically" works, I get extra ">" when I drop item on another item...
I also learned that I also need to call Q_EMIT layoutChanged({parent}/flag) to update view if I drop items on expanded parent...

Ok all good then, thanks! Weird issue o.o

@SGaist Woah... I have lots to read! Thanks! Gotta dig into proxier more then.

Hi,

One alternative could be to use QRubberBand and draw it on top of the lines/cells you are interested in.

@b2soft

I want to customize BrowserModel ::data() returning colors using QSS

@raven-worx knows much more than I, but I don't get this. role == Qt::TextColorRole will want to return a QBrush (http://doc.qt.io/qt-5/qt.html#ItemDataRole-enum). How is that linked to returning a string to use for a name in a stylesheet? :confused:

Ok so after LOTS of fighting with QTreeView... I started looking at viewport() and wuaila!

void myTree::resizeEvent(QResizeEvent *event) { QAbstractItemView::resizeEvent(event); b->setFixedSize(event->size().width(), 25); b->move(0,height()-25); viewport()->setMaximumHeight(height() - 50); };

Now my viewport gets painted exactly where I want it to, and the extra buttons can be placed in area where I want it to (b is the button) + its all nice all in one widget. I can expose needed controls for b as pert spec but drawing is as it should. Yay!

@Dariusz said in Multithreading conversion of QModelIdnexList to treeItem crash:

Well I have 100k items +/- to deal with

Then you should avoid to dereference the pointers to not copy the data around.

I created the base vectors via QVector<myTreeNode> vec(vecSize) and I access them all via [], so no thread should step over another thread,

It does as your backtrace shows. Since you're using QVector which is implicit shared every write access to the container checks if the container needs to get detached. Maybe try with a non-implicit shared container instead (e.g. a std::vector)
See http://doc.qt.io/qt-5/implicit-sharing.html for details.

I'll update this post once I go anywhere with it

This feedback sounds promising.

I am becoming more curious around another description for a similar use case:
An associative table can be created. QStandardItem objects can be generated then based on table entries. Which ways would you like to recommend for the connection of entities to their relationship data?

@JonB
This code is a small example to find solution.
In my real project i need to show tree with different items and nested trees.

Ok I found what was causing this awful bug.

When I check if the map contains the index, I actually check the raw index not the QPersistantModelIndex so this is not the same object and I think qmap does not check equality using == operator .

So I have to iterate over the map and check using QPersistantModelIndex "==" operators which can compare from a QModelIndex.

Hey

Just thoguh I'd drop a "temporary" solution I'm testing now. I ended up extending qtreeView, and capturing expand/collapse signals. Then I generate a QSet with these ids that later I can store/restore via model()->match(id) idea. Not sure how well it will work in very large data sets, say 100k items and 10k expanded state items, But well see.

What model are you using? QStandardItemModel had the roles argument working from Qt 5.11 onward. An empty role vector implies "all roles changed"

@Athius
Sure it works, here my code for a bookmark tree:

bool dropMimeData(const QMimeData *data, Qt::DropAction action, int row, int column, const QModelIndex &parent) { Q_UNUSED(column) if(action EQ Qt::IgnoreAction) return false; if(action EQ Qt::MoveAction) { QByteArray bytes=data->data(MIME_Bookmark); QDataStream stream(&bytes,QIODevice::QIODevice::ReadOnly); qintptr i; int r,c; stream>>i>>r>>c; QModelIndex index=createIndex(r,c,i); moveRow(index,index.row(),parent,row); } return true; }

Of course you need to create and return your custom data in mimeData () as well.

@raven-worx I don't use any custom item delegate.

My model doesn't update, its items are fixed.

@VRonin: Ok, updated to 5.10.1

Then you have to propagate the information when your underlying data changes.

I'm sorry but I still don't get how you would like it to look

@MrBolton
Sorry, I really don't understand your comment. Precisely the point is that with "bigger data sets" you would have to fetch all the data, which is incredibly slow and may well exceed available memory. (OTOH, if the dataset is "small" the overhead of re-query compared to holding all the previous rows locally is also "small".) As a rule of thumb, one always wants as much work to be done at the server side as possible, not client side.

Let's say the table has a million rows and the query I want to pose is:

SELECT * FROM table ORDER BY column LIMIT 100

Very approximately, pushing the ORDER BY to the server instead of into a QSortFilterProxyModel will run 10 thousand times faster and use one ten-thousandth of the client memory. That's why I would be very careful before advising people to use QSortFilterProxyModel without understanding the ramifications!

rather than the delegate I'd reimplement QTreeView:
protected:

virtual void drawBranches(QPainter *painter, const QRect &rect, const QModelIndex &index) const virtual void drawRow(QPainter *painter, const QStyleOptionViewItem &option, const QModelIndex &index) const

and slots

void collapse(const QModelIndex &index) void expand(const QModelIndex &index)

But agree is not straightforward and will require some work

Thank you for your suggestion. I think it's what I'll eventually settle for, fairly easy to indeed, using just one SQL query.
Although, I'm starting to think ... wouldn't my original aim be easier to fulfil using TreeWidget and not TreeView?

Hello guys. Thanks for your answerS.
The "error" was simple, stupid. In fact the _manageOrderedTable is an instance of a class called ManageOrderedTable which inherits from QTableWidget.
In this class, there's a method called fillAll and I called setColumnCount() from there. Bad idea.
When I call this method from upon the for loop, everything is fine. I don't understand why.
I could understand if setColumnCount() were called too late, but in this case I should have a segfault, trying to call a cell that doesn't exist.
Well, I don't really understand why but now it works...

Tank you for your help, Patrick.

@michalos I'm not sure if I can help you but seems like model test code works for me.

What I did:

Added modeltest.h and cpp into my project Added "testlib" into "QT" section of project file Added this to modeltest.h: #include <QtTest/qtestcase.h> #define qVariantCanConvert(_type_, _var_) _var_.canConvert(QMetaType::_type_) Changed expressions like "qVariantCanConvert<QString> ( variant)" to "qVariantCanConvert(QString, variant)" in modeltest.h

And... my sample model failed to pass the test of course :D

Ok. I've found the answer.

I subclass the QTreeView and override the void resizeEvent(QResizeEvent* e) as shown:

void MyTreeView::resizeEvent(QResizeEvent *e) { if(model() != Q_NULLPTR){ model()->layoutChanged(); } QTreeView::resizeEvent(e); }

And it did the trick.

Had help from the Voidrealms Facebook group.

@Yohdu said:

My second and main question is then : how to implement correct drag and drop operations with custom QStandardItem which needs to own custom data ?

With the standard item widgets/models you can't drag-n-drop custom item roles.
This makes sense, since the model doesn't know which UserRole+ was set.
It would need to iterate all possible item-role values when creating the drop-data.

The only solution i see is to create your own custom model and reimplement the needed drag-n-drop methods. This also give you full control over your data structure storing the data inside the model.
This is more work, but you will learn a lot and gain some performance (by avoiding the standard-item classes)

@Joel-Bodenmann said:

How can I make this work with my own delegates?

each index can max only have 1 delegate. Thus you need to merge the code into your own delegate.
Since you also receive a QModelIndex in the most methods of the delegate you need to check the column there.

@NIXIN

class MySortFilterProxyModel : public QSortFilterProxyModel { Q_OBJECT Q_PROPERTY(bool FilterEnabled READ isFilterEnabled WRITE setFilterEnabled ) public: virtual bool filterAcceptsRow(int source_row, const QModelIndex & source_parent) const { if( !m_IsFilterEnabled ) return true; // your filtering code here } bool isFilterEnabled() const { return m_IsFilterEnabled }; void setFilterEnabled( bool enabled ) { if( m_IsFilterEnabled != enabled ) { m_IsFilterEnabled = enabled; this->invalidate(); } } private: bool m_IsFilterEnabled;

Problem Solved.

CODE

QModelIndex index2=proxymodel->mapToSource(index); //Returns the source model index (In my case it was QFileSystemModel) QString selectedrow=filemodel->fileInfo(index2).absoluteFilePath(); //filemodel is My QFileSystem Model.

Hi,

Apply the same filter you are using for your single folder version except that this time rather than using a QString you should keep a QStringList of the folders you want to show and use that.

That wasn't really clear from your original description.

In that case, use a filter. QSortFilterProxyModel for example

Someone on stackoverflow was kind enough to point out my error. My parent method() should have been:

QModelIndex MyModel::parent(const QModelIndex &index) const { if (!index.isValid()) { return QModelIndex(); } QObject *item = (QObject *)index.internalPointer(); Directory *parentItem = qobject_cast<Directory *>(item->parent()); if (!parentItem || parentItem == invisibleTopLevelDirectory()) { return QModelIndex(); } Directory *parentOfParentItem = qobject_cast<Directory *>(parentItem->parent()); if (!parentOfParentItem) { return QModelIndex(); } int listIndex = parentOfParentItem->indexOfItem(parentItem); if (listIndex < 0) { return QModelIndex(); } else { return createIndex(listIndex, index.column(), parentItem); } }

Hi,

Can you provide a minimal compilable example that shows that behavior ?

One possibility might be to use directoryLoaded.

I don't know if there are any Qt models that support SQL for trees. I think they were all built for tables so you might have to build one yourself.
Have you checked out the Simple Tree Model example? I found it pretty helpful: http://doc.qt.io/qt-5/qtwidgets-itemviews-simpletreemodel-example.html . Also, if you find it easier, you can use a QStandardItemModel instead of a QAbstractItemModel.

@mmoerdijk
since Qt 5.10 there is a recursiveFilteringEnabled property on a QSortFilterProxyModel, which does exactly this.

This issue is now resolved. It turned out to be a problem in the model: I forgot to tell the root (invalid parent) to accept drops too. This is done in the QAbstractItemModel::flags() implementation.
The incorrect version was:

Qt::ItemFlags MyModel::flags(const QModelIndex& index) const { if (!index.isValid()) { return 0; } return Qt::ItemIsEnabled | Qt::ItemIsSelectable | Qt::ItemIsDropEnabled; }

While this is how it should look like to allow drops in the root (to add new items without a parent):

Qt::ItemFlags MyModel::flags(const QModelIndex& index) const { if (!index.isValid()) { return Qt::ItemIsDropEnabled; // Allow drops in the top-level (no parent) } return Qt::ItemIsEnabled | Qt::ItemIsSelectable | Qt::ItemIsDropEnabled; }

Cool, Thanks for all the insite: I have solved the problem with a little more testing.
For the sake of completeness and for others Ive included my code:

void DialogStructureXML::on_btn_add_toroot_released() { QModelIndex currentIndex = ui->tree->selectionModel()->currentIndex(); QModelIndex rootIndex = ui->tree->rootIndex(); QModelIndex traceIndex = currentIndex; QModelIndex child = currentIndex; tree_model *model = (tree_model*) ui->tree->model(); int rowCount = model->rowCount(); int currentRow = 0; int prevRow = 0; //some string cleanup QString ItemText = ui->edit_textvaluefortree->text() ; ItemText.simplified(); //If its a valid index (ie: valid something in the GUI tree) trace back to the parent if ( traceIndex.isValid() ) { while ( traceIndex.isValid() ) { currentRow = traceIndex.row()+1; traceIndex = model->parent(traceIndex); } } if (!model->insertRow(currentRow, rootIndex)) return; child = model->index(currentRow, 0, rootIndex); //Column index 0 model->setData(child, QVariant( ItemText ), Qt::EditRole); return; }

@kshegunov

Since I already worked with the graphics framework, I will stick with it!
Thank you both for your posts. It saved me a lot of time trying ;)

solved

Sounds pretty crazy...