Ah, OK. On May 6, 2010, at 8:21 AM, Noel Grandin wrote:
> > Yeah, it's more the improved memory usage that SWT.Virtual provides, > mostly by doing more work so that it's memory usage is roughly > O(no rows displayed) > instead of > O (no rows in model) > > We could do that, but it has both a code complexity cost, and a feature > cost - some features become impossible to support because they would > require scanning through the entire model. > > -- Noel > > > Greg Brown wrote: >>>> data sets in TableView ( but throwing huge globs of data at it is a data >>>> control issue not a UI issue ). By making sure I didn't load data using >>>> the UI >>>> thread the app is nice to use even when under allot of stress. >>>> >>>> >>>> >>> The Eclipse JFace/SWT project has a nice solution here, one that >>> may be worth copying - their table has an SWT.Virtual mode where it >>> treats the row-set-data like a virtual memory resource, only loading in >>> the bits it needs when they are needed for display. >>> >> You could also write a table model (i.e. List implementation) that behaves >> this way (fetches pages only as needed). >> >> However, overall TableView performance should be pretty good even with large >> data sets. See this demo app: >> >> >> http://svn.apache.org/repos/asf/pivot/trunk/demos/src/org/apache/pivot/demos/million/ >> >> You can get sample data files here: >> >> http://svn.apache.org/repos/asf/pivot/site/trunk/www/assets/ >> >> Copy them to a local directory and pass the base path to the app; e.g.: >> >> --basePath=/Users/greg/assets >> >> On my machine, it loads a million rows in less than 10 seconds, and scrolls >> smoothly. Make sure you allocate enough heap space, though - I needed 1.5GB >> to load the largest data file: >> >> -Xms128m -Xmx1536m >> >> >
