fix: AttributeError in single-output mode with XML/o5m output

[psca]

Summary

• Bug: AttributeError: 'list' object has no attribute 'min_lat' crash when running in single-output mode (--maxNodesPerTile=0) with XML or o5m output.
• Root cause: typing.cast() is a runtime no-op — it returns its second argument unchanged. The old code passed a plain list[float] where a BBox NamedTuple was expected, which fails when downstream code accesses named fields like .min_lat. The fix constructs an actual BBox, which also required moving the import out of TYPE_CHECKING.
• Regression test: Added test_process_files_single_output_bbox_type — the existing test_process_files_single_output E2E test uses PBF output, which unpacks bbox as a tuple and silently accepts a plain list; this new test targets the XML/o5m path by asserting get_osm_output receives a proper BBox instance, catching the exact failure mode.

Traceback

File "pyhgtmap/hgt/processor.py", line 277, in process_files
    self.get_osm_output(...)
  File "pyhgtmap/output/factory.py", line 87, in get_osm_output
    output = osmUtil.Output(...)
  File "pyhgtmap/output/osmUtil.py", line 72, in __init__
    self.boundsTag = make_bounds_tag(bbox)
  File "pyhgtmap/output/osmUtil.py", line 29, in make_bounds_tag
    return f'<bounds minlat="{bbox.min_lat:.7f}" ...'
AttributeError: 'list' object has no attribute 'min_lat'

Test plan

• hatch run fmt — formatting passes
• hatch run test — 296 passed, 5 skipped
• New regression test test_process_files_single_output_bbox_type covers the exact failure path
• Verified fix end-to-end against a real tile set covering low-elevation SEA region

🤖 Generated with Claude Code

[agrenott]

Thanks for the proposal.
At a first glance, I'm not convinced the proposed fix is fixing the root cause, but rather its consequences.
Could you please share the overall command/files used to reproduce the original issue?

[psca]

Thanks for the review!

Also updated the PR summary, the original description incorrectly framed this as specific to areas with no elevation data. The crash happens in single-output mode with XML/o5m output regardless, as the repro below shows.

Reproduction:

pyhgtmap -a 6:43:7:44 --max-nodes-per-tile=0 tests/data/N43E006.hgt

Root cause: typing.cast() is a runtime no-op, it returns its second argument unchanged. The old code passed a plain list[float] where a BBox NamedTuple was expected, which fails when downstream code accesses named fields like .min_lat. The fix constructs an actual BBox, which also required moving the import out of TYPE_CHECKING.

Longer-term, storing options.area as a BBox on Configuration would avoid re-parsing. Happy to follow up separately.