Alex Osborne
Alex Osborne
Oh, that's unfortunate. Pity we can't use merge operators from Java - doing a get before each put seems likely to have a significant effect on bulk data load performance....
I guess another solution would be to move the original url into the database key to break the tie. That'd mean all of them would be returned in search results...
Well galgeek's original example had two different URLs (player_id=1 vs player_id=2) that were only the same under fuzzy rules. But yes an exact revisit within the same second does sound...
@anjackson see #78 for roughly what I had in mind (without migration logic). I haven't really tested it much yet.
I think we could implement an online upgrade by roughly this process: 1. Temporarily until the index has a 'v4 migration complete' flag: * PUT should issue a delete for...
@kris-sigur reports that truncated timestamps still cause problems with sorting:  We might need to pad them at indexing time rather than just at query time. This leaves a problem...
@galgeek Any thoughts on this? Kris' PR seems reasonable to me but I don't have any indexes with truncated dates so I'm not really affected by this change.
OutbackCDX does not (yet) support storing arbitrarily named fields.
Note the "Things it doesn't do (yet): CDXJ" in the README. :-) While it can now map CDX11 fields to CDXJ for input/output it doesn't actually support storing arbitrary CDXJ...
Yes if the CDXJ input is limited to just the basic CDX11 fields it works.