id	summary	reporter	owner	description	type	status	component	version	severity	resolution	keywords	cc	stage	has_patch	needs_docs	needs_tests	needs_better_patch	easy	ui_ux
2061	[patch] PostgreSQL index introspection fails on tables that have dropped columns	Chris Chamberlin <dja@…>	Adrian Holovaty	"When building models from an existing PostgreSQL database,  {{{python manage.py inspectdb}}} fails if the table being inspected has had columns that were created and then dropped before inspection.
{{{
Traceback (most recent call last):
  File ""manage.py"", line 11, in ?
    execute_manager(settings)
  File ""/usr/lib/python2.3/site-packages/django/core/management.py"", line 1250
, in execute_manager
    execute_from_command_line(action_mapping)
  File ""/usr/lib/python2.3/site-packages/django/core/management.py"", line 1179
, in execute_from_command_line
    for line in action_mapping[action]():
  File ""/usr/lib/python2.3/site-packages/django/core/management.py"", line 716,
 in inspectdb
    indexes = introspection_module.get_indexes(cursor, table_name)
  File ""/usr/lib/python2.3/site-packages/django/db/backends/postgresql/introsp
ection.py"", line 67, in get_indexes
    col_name = desc[int(row[0])-1][0]
IndexError: tuple index out of range
}}}

The problem is that when a column is dropped, PostgreSQL doesn't actually remove it from the pg_catalog.pg_attribute table, which the pg_catalog.pg_index.indkey column references; instead, it just renames them to something like ""...pg.dropped.4..."".  The pgsql driver then leaves that column out of its column list, but this means you can't index into an array of columns using the indkey field.

I'll attach a patch that fixes get_indexes() by joining in the indexed column names in the main query.  This is tested with PostgreSQL 7.4.8; I don't have access to 8.x.  Also, I assume {{{django/db/backends/postgresql_psycopg2/introspection.py}}} should get the same patch.

"	defect	closed	Database layer (models, ORM)	dev	normal	fixed			Unreviewed	1	0	0	0	0	0