Description
Migrated issue, originally created by jahs (@jahs)
There appears to be an inconsistency within SQL Alchemy which autogenerated ExcludeConstraint migrations trigger, and an easy workaround. This is only seen for column names that need escaping, such as those containing upper case characters.
I'm using: alembic==0.9.7, SQLAlchemy==1.2.1
Adding an ExcludeConstraint to a Table results in a migration having a line such as:
postgresql.ExcludeConstraint(('"JB_NAM_JB_ID"', '='),
('"JB_NAM_During"', '&&'),
using='gist', name='hkJB_NAM_Job_Name')which appears to be correct following the ExcludeConstraint documentation, namely that the column names should be raw, escaped SQL strings.
However, at sqlalchemy/sql/schema.py(2668)_set_parent() we have
2665 def _set_parent(self, table):
2666 for col in self._pending_colargs:
2667 if isinstance(col, util.string_types):
2668 -> col = table.c[col]
2669 self.columns.add(col)
and so we get a KeyError as col is assumed not to be escaped.
Following the suggestion in the ExcludeConstraint docs, wrapping the column names in Column() rather than double quotes fixes it:
postgresql.ExcludeConstraint((sa.Column('JB_NAM_JB_ID'), '='),
(sa.Column('JB_NAM_During'), '&&'),
using='gist', name='hkJB_NAM_Job_Name')This seems to be a better fix than trying to unescape in SQL Alchemy.
Many thanks, and thank you for Alembic.