Skip to main content

Home/ GWT - MVP/ Group items tagged development

Rss Feed Group items tagged

Filter: All | Bookmarks | Topics Simple Middle
Esfand S

Large scale application development and MVP - Part II - 0 views

code.google.com/...mvp-architecture-2.html

mvp top best gwt

shared by Esfand S on 13 May 10 - Cached
  • segment the code that declares the UI from the code that drives the UI.
  • we want the our ContactsPresenter to implement a Presenter interface that allows our ContactsView to callback into the presenter when it receives a click, select or other event. The Presenter interface defines the following:   public interface Presenter<T> {    void onAddButtonClicked();    void onDeleteButtonClicked();    void onItemClicked(T clickedItem);    void onItemSelected(T selectedItem);  }
  • The first part of wiring everything up is to have our ContactsPresenter implement the Presenter interface, and then register itself with the underlying view. To register itself, we'll need our ContactsView to expose a setPresenter() method:   private Presenter<T> presenter;  public void setPresenter(Presenter<T> presenter) {    this.presenter = presenter;  }
  • ...8 more annotations...
  • Note that our ContactsView is now ContactsViewImpl<T> and implements ContactsView<T>. This is so that we can pass in a mocked ContactsView instance when testing our ContactsPresenter. Now in our AppController, when we create the ContactsView, we can initialize it with the necessary ColumnDefinition(s).
  • With the ColumnDefinition(s) in place, we will start to see the fruits of our labor. Mainly in the way we pass model data to the view. As mentioned above we were previously dumbing down the model into a list of Strings. With our ColumnDefinition(s) we can pass the model untouched.   public class ContactsPresenter implements Presenter,    ...    private void fetchContactDetails() {      rpcService.getContactDetails(new AsyncCallback<ArrayList<ContactDetails>>() {        public void onSuccess(ArrayList<ContactDetails> result) {            contactDetails = result;            sortContactDetails();            view.setRowData(contactDetails);        }        ...      });    }    ...  } And our ContactsViewImpl has the following setRowData() implementation:   public class ContactsViewImpl<T> extends Composite implements ContactsView<T> {    ...    public void setRowData(List<T> rowData) {      contactsTable.removeAllRows();      this.rowData = rowData;      for (int i = 0; i < rowData.size(); ++i) {        T t = rowData.get(i);        for (int j = 0; j < columnDefinitions.size(); ++j) {          ColumnDefinition<T> columnDefinition = columnDefinitions.get(j);          contactsTable.setWidget(i, j, columnDefinition.render(t));        }      }    }    ...  } A definite improvement; the presenter can pass the model untouched and the view has no rendering code that we would otherwise need to test.
  • Code Splitting is the act of wrapping segmented pieces of your application into "split" points by declaring them within a runAsync() call. As long as the split portion of your code is purely segmented, and not referenced by other parts of the app, it will be downloaded and executed at the point that it needs to run.
  • We've figured out how to create the foundation for complex UIs while sticking to our requirement that the view remain as dumb (and minimally testable) as possible, but that's no reason to stop. While functionality is decoupled, there is still room for optimization. Having the ColumnDefinition create a new widget for each cell is too heavy, and can quickly lead to performance degradation as your application grows. The two leading factors of this degradation are: Inefficiencies related to inserting new elements via DOM manipulation Overhead associated with sinking events per Widget To overcome this we will update our application to do the following (in respective order): Replace our FlexTable implementation with an HTML widget that we'll populate by calling setHTML(), effectively batching all DOM manipulation into a single call. Reduce the event overhead by sinking events on the HTML widget, rather than the individual cells. The changes are encompassed within our ContactsView.ui.xml file, as well as our setRowData() and onTableClicked() methods. First we'll need to update our ContactsView.ui.xml file to use a HTML widget rather than a FlexTable widget.
  • The above code is similar to our original setRowData() method, we iterate through the rowData and for each item ask our column definitions to render accordingly. The main differences being that a) we're expecting each column definition to render itself into the StringBuilder rather than passing back a full-on widget, and b) we're calling setHTML on a HTML widget rather than calling setWidget on a FlexTable. This will decrease your load time, especially as your tables start to grow.
  • To reiterate, we're reducing the overhead of sinking events on per-cell widgets, and instead sinking on a single container, our HTML widget. The ClickEvents are still wired up via our UiHandler annotations, but with this approach, we're going to get the Element that was clicked on and walk the DOM until we find a parent TableCellElement. From there we can determine the row, and thus the corresponding rowData.
  • Our current solution is to have our presenters pass a dumbed down version of the model to our views. In the case of our ContactsView, the presenter takes a list of DTOs (Data Transfer Objects) and constructs a list of Strings that it then passes to the view. public ContactsPresenter implements Presenter {  ...  public void onSuccess(ArrayList<ContactDetails> result) {    contactDetails = result;    sortContactDetails();    List<String> data = new ArrayList<String>();    for (int i = 0; i < result.size(); ++i) {      data.add(contactDetails.get(i).getDisplayName());    }    display.setData(data);  }  ...} The "data" object that is passed to the view is a very (and I mean very) simplistic ViewModel — basically a representation of a more complex data model using primitives. This is fine for a simplistic view, but as soon as you start to do something more complex, you quickly realize that something has to give. Either the presenter needs to know more about the view (making it hard to swap out views for other platforms), or the view needs to know more about the data model (ultimately making your view smarter, thus requiring more GwtTestCases). The solution is to use generics along with a third party that abstracts any knowledge of a cell's data type, as well as how that data type is rendered.
  •         // TODO: Really total hack! There's gotta be a better way...        Element child = cell.getFirstChildElement();        if (child != null) {          Event.sinkEvents(child, Event.ONFOCUS | Event.ONBLUR);        }
‹ Previous 21 - 26 of 26
Showing 20▼ items per page