I think Jacob is correct, both choices result in the entire result set
being read into memory, however #2 leaves the users code database
agnostic.
On Jul 14, 7:21 pm, "Jacob Kaplan-Moss" <[EMAIL PROTECTED]>
wrote:
> On Mon, Jul 14, 2008 at 7:13 PM, Malcolm Tredinnick
>
> <[EMAIL PROTECTED]> wrote
On Mon, Jul 14, 2008 at 7:13 PM, Malcolm Tredinnick
<[EMAIL PROTECTED]> wrote:
> (2) For SQLite *only*, we go back to the pre-queryset-refactor
> behaviour: all results are read in immediately upon accessing the
> queryset.
I'd prefer this -- ``list(qs)`` essentially ties user code to their
choic
In essence, ticket #7411 says that for SQLite (only, amongst our core
back-ends), having a cursor that contains partially read results will
cause write-commit problems when saving something to the database. In
code, which might be easier to understand, something like this will be
problematic somet
