[Rebuild] Status of the remaining failing tests

Andy Allan gravitystorm at gmail.com
Fri Jun 29 16:01:38 BST 2012


Hi All,

There's two remaining tests in the test suite that are currently failing.
I'm posting this here to solicit discussion about how big a problem you
think they will be in the real world. On one hand, it's nice to have
everything working perfectly, but on the other hand it might not be worth
holding up the whole process for rare events - we can always fix small
numbers of problems later, and add extra redactions if needs be.

I've written some commentary with the two cases, as well as trying to show
the output clearly from the bot. Apologies if it's destroyed by line
wrapping. An unwrapped version is at https://gist.github.com/3018460

Cheers,
Andy


A) Test Way Name Swap

There is an edge case in deciding significance of names. In Russia, "Foo
Street" and "Street Foo" are the same, and so if a decliner adds a street
name and reversed by an agreer, the name should be removed. Currently the
bot treats "Street Foo" as a distinct name, and so keeps it.

    history = [OSM::Way[[1,2,3], :id=>1, :changeset=>1, :version=>1,
"highway"=>"residental"], # created by agreer
               OSM::Way[[1,2,3], :id=>1, :changeset=>3, :version=>2,
"highway"=>"residental", "name"=>"ул. Гая"], # name added by decliner
               OSM::Way[[1,2,3], :id=>1, :changeset=>2, :version=>3,
"highway"=>"residental", "name"=>"Гая ул."]] # name swapped around by agreer

Expected

 [Edit[obj=OSM::Way[[1, 2,
3],:element_id=>1,:changeset_id=>-1,:timestamp=>,:visible=>true,:version=>3,"highway"=>"residental"]],
 Redact[klass=OSM::Way,element_id=1,version=2,mode=:hidden],
 Redact[klass=OSM::Way,element_id=1,version=3,mode=:visible]],
not
 [Redact[klass=OSM::Way,element_id=1,version=2,mode=:hidden]].

B) Test Way Reversed By Decliner

The geometry processing needed to figure out what state to leave a road in
(after figuring out what nodes need expunging) is complex and doesn't
handle every situation. Matt recently re-wrote it fully, and it now covers
all situations bar one. In this case, the order of the outputted nodes
isn't ideal - the "unreversing" of the way is being done to the nodes that
were in the way at the time it was reversed, ignoring the later node. Matt
reckons fixing this could involve a considerable about of work - adding a
fourth type of geometry processing to the three the bot already recognises.

    history = [OSM::Way[[1,2  ], :id=>1, :changeset=>1, :version=>1,
"oneway"=>"-1"], # created by agreer
               OSM::Way[[2,1  ], :id=>1, :changeset=>3, :version=>2,
"oneway"=>"yes"], # way reversed by decliner
               OSM::Way[[3,2,1], :id=>1, :changeset=>2, :version=>3,
"oneway"=>"yes"]] # node added by agreer

Expected
 [Edit[obj=OSM::Way[[1, 2,
3],:element_id=>1,:changeset_id=>-1,:timestamp=>,:visible=>true,:version=>3,"oneway"=>"-1"]],
 Redact[klass=OSM::Way,element_id=1,version=2,mode=:hidden],
 Redact[klass=OSM::Way,element_id=1,version=3,mode=:visible]],
not
 [Edit[obj=OSM::Way[[3, 1,
2],:element_id=>1,:changeset_id=>-1,:timestamp=>,:visible=>true,:version=>3,"oneway"=>"-1"]],
 Redact[klass=OSM::Way,element_id=1,version=2,mode=:hidden],
 Redact[klass=OSM::Way,element_id=1,version=3,mode=:visible]]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/rebuild/attachments/20120629/8225ad23/attachment.html>


More information about the Rebuild mailing list