Changes between Initial Version and Version 1 of Ticket #20425, comment 2


Ignore:
Timestamp:
2024-02-11T09:13:29+01:00 (2 years ago)
Author:
GerdP

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #20425, comment 2

    initial v1  
    11Looking at this again. According to #5462 the test was written to find duplicated multipolygons in imports. My thoughts:
    2 1. The test should be changed to produce only one error for a given pair of relations, either code 1901 "Duplicated relations" or code 1902 "Relations with same members" or maybe another new message, see below.
    3 2. The test that produces the informational message "Relations with same members" should ignore the order of the members and also the number of occurences. It should only compare roles and member ids and not there geometry. This allows to compare incomplete relations.
    4 3. Relations with different tags but identical member lists should have a different text, maybe "Relations with equal member lists". In this case order and number of occurences of members should matter.
     21. The test should be changed to produce only one error for a given ~~pair~~ set of relations, either code 1901 "Duplicated relations" or code 1902 "Relations with same members" or maybe another new message, see below. Edit: This can be tricky, so I consider it less important.
     32. The test that produces the informational message "Relations with same members" should ignore the order of the members ~~and also the number of occurences~~. It should only compare roles and member types and unique ids and not there geometry. This allows to compare incomplete relations.
     43. Relations with different tags but identical member lists should have a different text, maybe "Relations with identical member lists". In this case order and number of occurences of members should matter.
    554. A message "Duplicated relations" requires knowledge about the meaning of the order of the members. If the order matters it should be compared, else ignored, but different numbers of the same member should never be ignored. I think tags like source=* or fixme=* should not be ignored here, but those which we call "discardable" can be ignored.
    665. The current test tries to compare the geometry, this requires a pair of complete relations. It seems reasonable to ignore the "uninteresting" tags like source in this case and also to offer an autofix which removes the newer relation. I am just not happy with the message "Duplicated relations". I'd prefer to have something like "Relations with equal meaning describe identical geometry". I wonder if this geometry test should better be done in `MultipolygonTest`?