Wired my first xamPivotGrid into a user control that is consumed by a winforms app. Borrowed a method from these forums that takes a DataTable and dynamically all columns and rows to an Ilist, which is set to the pivotDataSelector and pivotGrid's DataSource.
My application loads at a cost of 90,512k. I load the data into the pivot grid and it hit 96,180. The DataTable contains 1,216 rows and 8 columns. Dragging just 3 columns to the Rows box drives the application memory cost to ~1,650,227k before the application throws a System.OutOfMemory exception.
As my first venture, I am obviously missing something critical here, but not finding much and not really sure what to look for. Feeling additional paing because I am under the gun to get something working and performing by next Monday. I have seen code examples that set hierarchies and a whole bunch of other things that do not seem very appealing because of hard-coded column names, etc., and not even sure these would help me?
In short, I guess what I need is something like "Jaime, if your clients will be working with large (is my example above large???) data tables, then you need to do this..."
I will seriously cry if I can't get this to perform well. We purchased the IG Ultimate just for the pivot grid and time is running out for me... perhaps literally at that!
Many thanks!!! -Jaime
Hello Jaime,
Could you provide us at least your initialization code? Does your application perform well if you feed it using another items source? I mean you can try to create a dummy list populated with just few manually generated items and pass that list to FlatDataSource.ItemsSource. Also which version of the control you have?
Thanks.
PPilev.
Hello PPilev,
Using the latest 2011.2 controls. Here is the method in the user control to convert the DataTable to an Ilist for the data selector and pivot grid:
public void SetFlatDatasource(DataTable dt) { DynamicTypeBuilder typeBuilder = new DynamicTypeBuilder { DynamicAssemblyName = "MyAssembly", DynamicTypeName = "Pane" }; IList properties = new List(); foreach (DataColumn column in dt.Columns) { DynamicTypePropertyInfo propertyInfo = new DynamicTypePropertyInfo { PropertyName = column.ColumnName, PropertyType = column.DataType }; properties.Add(propertyInfo); } Type dynamicType = typeBuilder.GenerateType(properties); Type listType = typeof(List<>); Type genericListType = listType.MakeGenericType(dynamicType); IList list = (IList)Activator.CreateInstance(genericListType); foreach (DataRow dataRow in dt.Rows) { object myDynamicInstance = Activator.CreateInstance(dynamicType); foreach (DataColumn column in dt.Columns) { PropertyInfo propertyVal = dynamicType.GetProperty(column.ColumnName); if (dataRow[column] != DBNull.Value) { propertyVal.SetValue(myDynamicInstance, dataRow[column], null); } } list.Add(myDynamicInstance); } FlatDataSource flatDataSource = new FlatDataSource() { ItemsSource = list, Cube = DataSourceBase.GenerateInitialCube("Pane"), // if you know the names of demensions you want to be in rows and columns you can define them here Columns = DataSourceBase.GenerateInitialItems("[Columns]"), Rows = DataSourceBase.GenerateInitialItems("[Row]"), Measures = DataSourceBase.GenerateInitialItems("Value") }; dataSelector.DataSource = flatDataSource; pivotGrid.DataSource = flatDataSource; }
When you say initialization, do you mean the xaml? If yes, how can we post xaml in this forum? I will say it does not contain anything special. AllowCompactLayout is set to True, Height and Width are set to Auto...there is really nothing noteworthy.
Simple question, should this control perform well with a row count of say 2,000 with 8 columns?
Hi JaimeZ,
I’m preparing a WPF sample now and I’m going send it to you soon.
Plamen.
Hi again,I have attached the WPF version. In the sample all root nodes are expanded by default. Also flat data source now uses a non-generic hierarchy and level descriptors appropriate for dynamic classes and excel scenarios as well.
Regards.Plamen.
Hi Plamen,
You are the best. REALLY appreciate your support! I was able to use your example to finally see my data pre-loaded in the pivot grid, however staying on topic with memory issues, I need to have all of these levels expanded by default. Users will need to see this view by default else they need to click hundreds if not thousands of expanders...it is not going to happen.
Ideally, what they want is the first row group level expanded by default, but the expansion of each of these to expand the all of the related remaining sub groups. In your example this would mean having All Products expanded to see the names and totals of the Categories. The expansion of one Category should auto-expand of of that Category's Brands and Products.
For Columns, all of the Regions and Distributors should be visible when loaded.
How did you generate this event method? I copied the method into my code but it is never hit because I do not know how to wire it. In winforms I am used to auto generating these events from each control's properties window.
private void AreaItems_CollectionChanging
Disregard my question about the event method. The answer was right in front of my face. Bad morning... Computer troubles delayed me for over two hours. :)
Hello JaimeZ,Please look at the updated sample. I've modified it to work as you've described. Now there are two different event handles listening for CollectionChanging events - one for the rows and one for the columns. There is a note about when there is more than one hierarchy in the area. The hierarchies next to the first one might have more than one node for given member and expanding one node will result in expanding all nodes related to the same member. Look at the images bellow:
No matter which Category1 node is clicked will result in that all Category1 members will be expanded
Look at the sample and if it covers the scenario you have described we can continue to work on the memory issue. Regards.Plamen.
I should try to think before posting. :)
All child nodes should be expanded. What I would love is to not have totals for Category, Brand and Product where there is only one. Please tell me this is possible?
Hello Plamen,
Your sample works great. The only enhancement that would make it perfect would be to not expand child nodes if there is only 1. This way you do not end up with a bunch of "total" rows for a single Category, Brand, and product. Product should only be expanded if there is more than 1. Same for Brand and Category. Is this an easy tweak?
The good news is after expanding all nodes the memory only hit a maximum of 200,000 kb, so the issue is only with dragging dyamic rows into the grid with no pre-defined data class or heirarchies.
Thanks!