feat: SQLAlchemy 1.4 support

[uranusjr]

Description

An alternative way to fix #1710. Flask-SQLAlchemy 2 broke the clause=None parameter, but since mapper=None, clause=None is simply the default since SQLAlchemy 0.7.2 (released more than ten years ago), we can simply fix the problem by not passing those parameters at all.

ADDITIONAL INFORMATION

• Has associated issue: Relax SQLAlchemy version dependency #1710
• Is CRUD MVC related.
• Is Auth, RBAC security related.
• Changes the security db schema.
• Introduces new feature
• Removes existing feature

[codecov]

Codecov Report

Merging #1783 (13e7fd5) into master (e2b744c) will not change coverage.
The diff coverage is 50.00%.

❗ Current head 13e7fd5 differs from pull request most recent head b3b53ca. Consider uploading reports for the commit b3b53ca to get more accurate results

@@           Coverage Diff           @@
##           master    #1783   +/-   ##
=======================================
  Coverage   77.05%   77.05%           
=======================================
  Files          56       56           
  Lines        8175     8175           
=======================================
  Hits         6299     6299           
  Misses       1876     1876           

Flag	Coverage Δ
python	77.05% <50.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
flask_appbuilder/cli.py	57.00% <0.00%> (ø)
flask_appbuilder/security/sqla/manager.py	71.38% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e2b744c...b3b53ca. Read the comment docs.

[dpgaspar]

Hi, this would be great, but while testing it I got:

DB Creation and initialization failed: get_bind() got an unexpected keyword argument 'bind'

I think that get_bind from flask-sqlalchemy is still incompatible with SQLAlchemy 4.X.

Bump the Flask-SQLAlchemy on requirements.txt also.

[uranusjr]

I regenerated the entire requirements.txt with pip-compile instead. I think this should work?

[uranusjr]

Oh, since all test jobs share the same requirements.txt, the 3.6 test does not run anymore since some of the dependencies have dropped 3.6 (since it already reached EOL). What is a good approach to this? IMO the test jobs should not share requirements.txt because the file depends on Python version. They should instead directly install the package from setup.py. Alternatively, if a fully-pinned requirements file is desired, each Python version should independently generate one file, instead of all versions sharing one file.

[dpgaspar]

Oh, since all test jobs share the same requirements.txt, the 3.6 test does not run anymore since some of the dependencies have dropped 3.6 (since it already reached EOL). What is a good approach to this? IMO the test jobs should not share requirements.txt because the file depends on Python version. They should instead directly install the package from setup.py. Alternatively, if a fully-pinned requirements file is desired, each Python version should independently generate one file, instead of all versions sharing one file.

Right but besides that, try your change locally with just SQLAlchemy>=1.4.
I've started this: #1786, and have surpassed the get_bind issue on flask-sqlalchemy. But bumped on another problem this time with FAB vs SQLAlchemy

[dpgaspar]

@uranusjr thank you to giving it a shot on fixing this issue. Going to close it in favour of #1786