]> git.ipfire.org Git - thirdparty/sqlalchemy/sqlalchemy.git/commit
- additional adjustment to the fix made in 8ad968f33100baeb3b13c7e0b724b6b79ab4277f
authorMike Bayer <mike_mp@zzzcomputing.com>
Mon, 22 Feb 2016 03:46:31 +0000 (22:46 -0500)
committerMike Bayer <mike_mp@zzzcomputing.com>
Mon, 22 Feb 2016 03:46:31 +0000 (22:46 -0500)
commitaa9ce3f521f254da9879ede011e520ec35b8270e
treefa2601f7b1f11e60479a104e783d055c95e06eab
parentc87d0bff3ada6368512462211973980e52591f6f
- additional adjustment to the fix made in 8ad968f33100baeb3b13c7e0b724b6b79ab4277f
for ref #3657.  The Oracle dialect makes more use of the "select_wraps_for"
feature than SQL server because Oracle doesn't have "TOP" for a limit-only
select, so tests are showing more happening here.  In the case where
the select() has some dupe columns, these are deduped from the .c collection
so a positional match between the wrapper and original can't use .inner_columns,
because these collections wont match.  Using _columns_plus_names
instead which is the deduped collection that determines the SELECT display,
which definitely have to match up.
lib/sqlalchemy/sql/compiler.py
test/sql/test_compiler.py