[Talk-GB] Yet more musical chairs updates.

Robert Scott lists at humanleg.org.uk
Wed Aug 4 14:48:48 BST 2010


On Wednesday 04 August 2010, Dave F. wrote:
> What's the different between the circles & rectangles? Is it just to do 
> with the zoom factor?

When there are more than n (currently 1024) results in an area, it shows only the first n results. You can choose which n these are (random sample, most recently updated...). This is a non-authoritative view.

Once the view is zoomed in far enough to show all results in an area, it shows an authoritative view.

Non authoritative views are shown with circles, authoritative views show the actual OS Locator bounding boxes. This is partly to do with making a clear and obvious distinction between views where you're seeing everything and views where there are some thing you're not seeing . It's also to do with the way the two different types of geometry behave at different scales. If I showed the boxes at low zoomlevels, they would just end up being tiny subpixel dots.

> Would it be possible to turn these circles off at lower zoom levels? 
> Personally I like to double click on the map to zoom in at these levels 
> as it centres the city I'm interested in & so I can  then use the bar to 
> zoom accurately to the specific area I'm interested in.

Yeah that annoys me too.

I tend to do the shift-drag-box more though.

Previously you weren't able to select non-authoritative points at all, but last night I changed it so that you can make selections that appear to be persistent across the authoritative-non-authoritative boundary, as I found it stupid that you couldn't see details of a match without first zooming right the way in and possibly losing track of which result you were interested in.

It would be nice if I could maybe hijack the doubleclick event and pass it to the map. I'll have to think about this.

> Are there any differences between what you've done & ITO?

My algorithm does fuzzy matching to find streets with smallish errors and AFAIK theirs doesn't.

I keep a history of match state change events, which will probably be useful for some fun features in the future.

Theirs supports not:name=, I haven't got round to that yet (I'm slightly more interested in being able to tag the actual OSL entry as being incorrect).

They've got tiles which are very good for use in-editor. Mine, you've still got to pan around in a separate window.


robert.

(the first thing I've got to do though is fix a really stupid replication bug of mine)



More information about the Talk-GB mailing list