Open jerstlouis opened 1 year ago
with the clipping algorithm marking artificial segments at the edges (e.g., as 'hidden' flag for From and To of vertices of (Multi)Polygon contours, or a vertex of a cut (Multi)LineString)
I don't understand what this means: is it something the writer should do ? If so, are there examples of such tiles ?
@rouault Re-clipping and marking artificial segments is something that the "reader" of MVTs which contain an extra border would do to facilitate re-assembling them as features. Some formats of tiled vector data may be encoded in a way where this is already part of the data model (e.g., MIL-STD-2407 Vector Product Format or our GNOSIS Map Tiles). Without either of these (extra border or artificial segments markers) options, a reader would need to look at adjacent tiles and look for touching features with identical IDs to identify artificial segments.
Our server also generates GeoJSON with a "line::hidden"
property that follows this idea:
See OGC Testbed 13 - Vector Tiles ER and OGC Change Request 515 for some additional details on the idea.
Feature request
Support automatically and efficiently recombining vector tiles as features in OGR/GDAL.
As initially brought up in #7934
When marking artificial segments, an efficient "union" algorithm can be implemented. For MVTs, this can be done by clipping the geometry spilling in the borders to a 2DTMS tile's axis-aligned bounding box (AABB) again, with the clipping algorithm marking artificial segments at the edges (e.g., as 'hidden' flag for From and To of vertices of (Multi)Polygon contours, or a vertex of a cut (Multi)LineString). Then an efficient algorithm can recombine geometry for shared tile boundary edges based on those few points along the edge, for which a correspondance can easily be established from both sides of the tile boundary.
This would apply to the OGC API, MVT, MBTiles and PMTiles driver, and is a much more useful resolution than the proposed: