Attempts to import TupleType on SQLAlchemy 1.3

[aabpxdb]

I know 1.3 is an older version, but it is within the range of supported versions stated in the README. What would you recommend for fixing this issue?

[bchoudhary6415]

Hi @aabpxdb,

I verified this behavior with minimal reproducible example.

Test script

from sqlalchemy.sql.sqltypes import TupleType, Integer, String
coltype = TupleType(Integer(), String())
print(coltype)

Observed Behaviour

With SQLAlchemy 1.3.24

Traceback (most recent call last):
  File "C:\issue_ibmdbsa191.py", line 1, in <module>
    from sqlalchemy.sql.sqltypes import TupleType, Integer, String
ImportError: cannot import name 'TupleType' from 'sqlalchemy.sql.sqltypes' (C:\python3.12\Lib\site-packages\sqlalchemy\sql\sqltypes.py). Did you mean: 'NullType'?

With SQLAlchemy 1.4.x/2.0
TupleType(Integer(), String())

Conclusion

TupleType is not available in SQLAlchemy 1.3.
Support for TupleType was introduced starting with SQLAlchemy 1.4 and continues in 2.0+.

This means:

• The error observed with SQLAlchemy 1.3 is expected behavior.
• it is not a bug in ibm_db_sa
• TupleType cannot be used when running against SQLAlchemy 1.3

To use TupleType, please upgrade SQLAlchemy to 1.4 or later.

Closing this issue, as the behavior is consistent with SQLAlchemy’s version support.
If this issue is encountered again after upgrading to SQLAlchemy 1.4+, feel free to reopen.

Thank you!

[aabpxdb]

I know that TupleType is not available. I am trying to run the library without it, and the problem is that it tries to import it even on versions where it doesn't exist.

[bchoudhary6415]

Hi @aabpxdb,

Thanks for the clarification.

To better understand and resolve this, could you please confirm which library is attempting to import TupleType unconditionally?

From our side, this behavior does not originate from python ibm_db_sa itself.
The inability to import TupleType on SQLAlchemy 1.3 is a known and expected limitation of SQLAlchemy, as TupleType was introduced only in SQLAlchemy 1.4.
If the import is occurring inside SQLAlchemy internals or another third-party dependency, this would be an issue in that library rather than in python ibm_db_sa.

I would recommend:

• Identifying the exact library/module that is importing TupleType
• Raising the issue with the SQLAlchemy project (or the relevant upstream library), as they are best positioned to address version-specific import behavior

At this point, there is no actionable change required in python ibm_db_sa.

Thank you!

[aabpxdb]

Yes, I can identify the module. It is this file in ibm_db_sa: https://github.com/ibmdb/python-ibmdbsa/blob/master/ibm_db_sa/base.py

Line 63.

On line 67, it is also attempting to import ExternalType from sqlalchemy.sql.type_api, which is another thing introduced in SQLAlchemy 1.4.

[bchoudhary6415]

Hi @aabpxdb,

Thanks for pointing this out, you’re absolutely right, and thank you for identifying the exact location.
You are correct that in ibm_db_sa/base.py we are currently importing TupleType and ExternalType unconditionally for all SQLAlchemy

versions < 2.0 as well, even though:
• TupleType was introduced in SQLAlchemy 1.4
• ExternalType was also introduced in SQLAlchemy 1.4
• Neither exists in SQLAlchemy 1.3

This means that importing them on SQLAlchemy 1.3 is incorrect and will fail at import time, even if the code path never uses them.
So yes — this is a real compatibility bug in base.py, and thanks for catching it.

We will fix this by importing these symbols conditionally based on the SQLAlchemy version, so that:
• SQLAlchemy 1.3 continues to work without TupleType / ExternalType
• SQLAlchemy 1.4+ uses them where available
• SQLAlchemy 2.x remains fully supported

I’ll raise a PR shortly with the fix.

Thank you!

[bchoudhary6415]

Hi @aabpxdb,

Thanks for reporting this issue.

The TupleType import problem has been fixed by adding proper SQLAlchemy version checks and cleaning up version-specific type imports. The fix works correctly with SQLAlchemy 1.3, 1.4, and 2.0.

The fix is included in this PR #192 and will be available in the next release.

Thank you!