There are couple of changes in this PR to fix this repo pipeline which is failing from last 3 months:
1) Spanner doesn't supports inserting 2 rows (a row with NULL value and a row with some value for a given datatype) for a table with empty primary key. SqlAlchemy added tests couple of months back to insert 2 such rows for different datatypes columns so those tests started failing for this repo.
Fixing to override those tests so as not to insert a row with NULL value but just a single row for each datatype
There are couple of changes in this PR to fix this repo pipeline which is failing from last 3 months:
1) Spanner doesn't supports inserting 2 rows (a row with NULL value and a row with some value for a given datatype) for a table with empty primary key. SqlAlchemy added tests couple of months back to insert 2 such rows for different datatypes columns so those tests started failing for this repo. Fixing to override those tests so as not to insert a row with NULL value but just a single row for each datatype
2) tests: remove editable install in tests
3) accommodate NULL at the compiler level for literal_render. Please see https://github.com/sqlalchemy/sqlalchemy/commit/1a23c8dee5665ebda75a1ea7d5e7ca355ea1f78b where are copying changes from compiler.py