Changes between Version 1 and Version 2 of Ticket #18375, comment 22


Ignore:
Timestamp:
May 30, 2013, 7:44:20 AM (11 years ago)
Author:
German M. Bravo

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #18375, comment 22

    v1 v2  
    11I would definitely go for an `annotate()`, for the akaariai's example:
    2 `People.objects.filter(friends__age__gt=F('friends__age') * 2)`
     2{{{
     3People.objects.filter(friends__age__gt=F('friends__age') * 2)
     4}}}
    35This, after the fix, makes just one join.
    46
    57If one, however, wants the friends with 2x the age (two separate joins), it'd need `annotate()` , like so:
    6 `People.objects.annotate(twice_age=F('friends__age') * 2).filter(friends__age__gt=F('twice_age'))`
     8{{{
     9People.objects.annotate(twice_age=F('friends__age') * 2).filter(friends__age__gt=F('twice_age'))
     10}}}
    711
    812How far is django from such behavior? ...and also, it doesn't still feel entirely good (annotate always creating new joins for such cases), since one might actually want to get the reused join... I would probably suggest the order of the calling of annotate to change things (if putting annotate after the filter, it'd reuse the joins from the previous filter, otherwise it would start afresh). This way, the next query would be equivalent to the first one:
    9 `People.objects.filter(friends__age__gt=F('twice_age')).annotate(twice_age=F('friends__age') * 2)`
     13{{{
     14People.objects.filter(friends__age__gt=F('twice_age')).annotate(twice_age=F('friends__age') * 2)
     15}}}
    1016Provided one could use `F()` of not yet "annotated" things.
    1117
Back to Top