[jan-glx]

Closes #1226.

#1266 fixed the left join when the element and the table share no instance. The other join types still broke in that case:

• points with an inner or right join (match_rows="no" or "left"): KeyError: None, because _get_masked_element passed None to .loc for dask frames
• inner join with match_rows="left": no error, but the element kept all its rows next to a 0-row table
• inner or right join with match_rows="right": IndexError from .iloc[0] on the empty merge in _match_rows

_get_masked_element now uses an empty array instead of None when nothing matches, so the existing ordering code and .loc handle that case for shapes and points alike. _match_rows returns empty indices for an empty merge.

The new test runs every how × match_rows combination for shapes and points against a table that shares no instance with the element; 9 of the 30 cases fail on main. The results now match the docstring of join_spatialelement_table: left keeps the element and returns a 0-row table, inner returns an empty element and a 0-row table, right returns an empty element and the full table.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.43%. Comparing base (602aec0) to head (d7dbad6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1312   +/-   ##
=======================================
  Coverage   91.43%   91.43%           
=======================================
  Files          53       53           
  Lines        8395     8395           
=======================================
  Hits         7676     7676           
  Misses        719      719           

Files with missing lines	Coverage Δ
src/spatialdata/_core/query/relational_query.py	93.34% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.