I've 1-2 million geo-located points, ranked by an importance. Most of these will be of zero importance. But between 100,000 and 200,000 points will be of particular interest. Lets say ranked by population, with the top place being the earth, and then continents, and finally the most populous countries/cities/towns.
I'm developing a Openlayers JS web app to view these feature, the initial version just runs a box query for users current view, and limits the number of features based on a simple calculation of map view size (x*y) divided by twice the size of the icon. Works ok for the demo.
For the next version I plan on using a Tile Feature Strategy so that the client and server can cache the responses (JSON).
The map query returns a 'light' JSON version of the feature, when clicking on the feature the app navigates to a URL within the feature data that is info page about that feature.
Most users to the site are expected to jump into the site at a particular feature info page form a google search. So I have also to provide a mechanism for search engines to index these 1-2 million feature info pages. I have an initial clustering algorithm that bubbles the highest ranked place to the top, and it tries to ensure that from the top down every parent feature has 10 lower ranking children. It's relatively slow, and not yet finalised. I think that I will ditch the bottom up and implement a simpler top down algorithm that select the highest rank feature, then the next 10, assigning these as children to the top feature. Then repeat, the next highest 100 unassigned children, assigning them to the previous 10, nearest first, not assigning more that 10 per node. The child features are links on the parent feature info page.
The Tile Feature Strategy and the google indexing are quite similar, except each tile has links to say 10 feature info pages, and four sub-tile/branches. So 14 links to follow, 10 being leaf nodes, and 4 being tile/branch nodes.
Also important feature will be present in multiple tiles (up to 24 if that's the zoom level I end up supporting). Also If I use TFS for indexing then the parent feature pages will need not to have child links.
Main Question: Will the TFS arrangement of features be less effective (re: google indexing) that the 1 parent feature to 10 children arrangement?
If it is as effective can the latest google spider even index Tile Feature Strategy? If not is there a way of helping spiders crawl tile feature feed?
I will accept an informed no answer, and implement the simple top down algorithm.
أكثر...
I'm developing a Openlayers JS web app to view these feature, the initial version just runs a box query for users current view, and limits the number of features based on a simple calculation of map view size (x*y) divided by twice the size of the icon. Works ok for the demo.
For the next version I plan on using a Tile Feature Strategy so that the client and server can cache the responses (JSON).
The map query returns a 'light' JSON version of the feature, when clicking on the feature the app navigates to a URL within the feature data that is info page about that feature.
Most users to the site are expected to jump into the site at a particular feature info page form a google search. So I have also to provide a mechanism for search engines to index these 1-2 million feature info pages. I have an initial clustering algorithm that bubbles the highest ranked place to the top, and it tries to ensure that from the top down every parent feature has 10 lower ranking children. It's relatively slow, and not yet finalised. I think that I will ditch the bottom up and implement a simpler top down algorithm that select the highest rank feature, then the next 10, assigning these as children to the top feature. Then repeat, the next highest 100 unassigned children, assigning them to the previous 10, nearest first, not assigning more that 10 per node. The child features are links on the parent feature info page.
The Tile Feature Strategy and the google indexing are quite similar, except each tile has links to say 10 feature info pages, and four sub-tile/branches. So 14 links to follow, 10 being leaf nodes, and 4 being tile/branch nodes.
Also important feature will be present in multiple tiles (up to 24 if that's the zoom level I end up supporting). Also If I use TFS for indexing then the parent feature pages will need not to have child links.
Main Question: Will the TFS arrangement of features be less effective (re: google indexing) that the 1 parent feature to 10 children arrangement?
If it is as effective can the latest google spider even index Tile Feature Strategy? If not is there a way of helping spiders crawl tile feature feed?
I will accept an informed no answer, and implement the simple top down algorithm.
أكثر...