Table not set for removed columns passed to include_object

[sqlalchemy-bot]

Migrated issue, originally created by Janne Vanhala (@jpvanhal)

I would like some columns, which are not in the model, not to be removed from the database when I run autogenerate. I tried using include_object function like this:

def include_object(object, name, type_, reflected, compare_to):
    if (
        type_ == 'column' and 
        object.table.name == 'some_table' and 
        object.name == 'some_column'
    ):
        return False
    return True

context.configure(
    # ...
    include_object = include_object
)

However, this fails because object.table is None for removed columns. Thus there is no way to know what table removed columns belong to.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

can you confirm what I'm seeing is that in this case the "reflected" object is None, right? that's not right, "reflected" should be the Column we got from the database and it should have the table. Can you try this patch please:

diff --git a/alembic/autogenerate/compare.py b/alembic/autogenerate/compare.py
index cfa110a..2c590f3 100644
--- a/alembic/autogenerate/compare.py
+++ b/alembic/autogenerate/compare.py
@@ -133,8 +133,9 @@ def _compare_columns(schema, tname, object_filters, conn_table, metadata_table,
                     nullable=conn_table.c[cname].nullable,
                     server_default=conn_table.c[cname].server_default
                 )
+        conn_col = conn_table.c[cname]
         if _run_filters(rem_col, cname,
-                                "column", True, None, object_filters):
+                                "column", True, conn_col, object_filters):
             diffs.append(
                 ("remove_column", schema, tname, rem_col)
             )

[sqlalchemy-bot]

Changes by Michael Bayer (@zzzeek):

• added labels: high priority

[sqlalchemy-bot]

Janne Vanhala (@jpvanhal) wrote:

In my case "reflected" is True and "compare_to" is None.

After your patch, "reflected" is still True and "compare_to" is the column from the database and it has the table.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

• Fixed bug where the include_object() filter would not receive
  the original :class:.Column object when evaluating a database-only
  column to be dropped; the object would not include the parent
  :class:.Table nor other aspects of the column that are important
  for generating the "downgrade" case where the column is recreated.
  fixes #200

→ bc6971a

[sqlalchemy-bot]

Changes by Michael Bayer (@zzzeek):

• changed status to closed

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

OK so fix is different, the original way you're trying to use it is correct. the "reflected" flag is True, meaning the object you're getting is supposed to be the actual reflected object. I think at some point that "rem_col" was there to recreate a Column object that might not have been present but now it's not needed.