[mandli]

kmltools.make_input_data_kmls read topofiles and dtopofiles only in the legacy list form, taking the path from f[-1] and the type from f[0]. A setrun.py that appends Topography or DTopography objects, which is the current style, failed with

TypeError: 'Topography' object is not subscriptable

That happened outside the try, so no kml files were written at all. For a cropped topofile, the box also showed the whole file rather than the region GeoClaw uses. That matters for the multi-tile setups in #760, where only part of each tile is used.

---

Claude Analysis

Changes

• make_input_data_kmls converts entries with the same TopographyData._normalize_topofiles() / DTopoData._normalize_dtopofiles() that make data uses. Legacy [type, path] lists still work. Their deprecation warning is left to make data instead of being repeated here.
• The topo box shows the region used, chosen by the new _topo_region_used():
  • the crop_extent, if one is set;
  • for a topo_entries() descriptor entry, its crop_bounds moved by its lon_wrap_offset;
  • otherwise the file's extent moved by x_shift/y_shift (the plain file extent when there is no shift, as before).
• topo2kml takes an optional extent= to draw a box other than the file's own extent. Existing calls are unchanged.

Tests

tests/test_kmltools_inputs.py runs make_input_data_kmls and reads the box back from the kml it writes:

• Topography with no crop: the file extent;
• Topography with crop_extent: the crop;
• Topography with x_shift/y_shift: the shifted extent;
• DTopography object: its extent;
• a legacy [3, path] list: still works, with no second DeprecationWarning.

The four object cases fail on master (0ba4db7b) with the TypeError and pass here. Full tests/: 635 passed, 5 skipped, 1 xfailed.

Related

#760. Once the crop-coverage rule is in, every crop_extent is guaranteed to lie within its file, so the box drawn here is exactly the region used. A tile-selection helper (also #760) would then leave only the tiles that contribute in topofiles, which gives the kml output Randy described.