dgilland / alchy Goto Github PK

http://docs.sqlalchemy.org/en/rel_0_9/orm/events.html

But allow an event decorator to function on class methods.

Something like:

class MyModel(Model):
    @event('before_insert', 'before_update'):
    def foo(...):
        ...

with automatic event registration.

In ModelBase, rename advanced_search_config and simple_search_config to __advanced_search__ and __simple_search__ respectively.

Hello

Thanks for the work. It makes integrating SqlAlchemy into a custom framework a snap.

I have an existing database with some database views. I want to reflect the database to map the tables and the views in a dictionary.

1. How would I support only reflecting a select set of tables?
2. How would I support reflecting views? I noticed the reflect method only support tables in the metadata?

Would I follow the Flask pattern and subclass the Manager class and override the needed methods?

Thanks
Greg

FYI: pep8 is now pycodestyle

Details here: PyCQA/pycodestyle#466

In setting up a py.test session fixture, I ran into an issue where Alchy does not properly handle arguments passed from SQLAlchemy.

>       session.add(model)

tests/test_models.py:17:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../.envs/platform/lib/python3.5/site-packages/sqlalchemy/orm/scoping.py:157: in do
    return getattr(self.registry(), name)(*args, **kwargs)
../../../.envs/platform/lib/python3.5/site-packages/sqlalchemy/util/_collections.py:1025: in __call__
    val = self.registry.value = self.createfunc()
../../../.envs/platform/lib/python3.5/site-packages/alchy/manager.py:241: in create_session
    return self.session_class(self, **options)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <alchy.session.Session object at 0x10ddaf908>, manager = <alchy.manager.Manager object at 0x10c512f28>, options = {'binds': {}}, bind = <sqlalchemy.engine.base.Connection object at 0x10dda2a20>

    def __init__(self, manager, **options):
        self.manager = manager
        bind = options.pop('bind', manager.engine)
        super(Session, self).__init__(bind=bind,
                                      binds=manager.binds_map,
>                                     **options)
E       TypeError: __init__() got multiple values for keyword argument 'binds'

../../../.envs/platform/lib/python3.5/site-packages/alchy/session.py:22: TypeError

The easy fix here is to update __init__ within the Session class to pop off the extraneous options. That said, I am not sure what the ramifications are of ignoring the binds value. Therefore, my proposed solution is as follows:

    def __init__(self, manager, **options):
        self.manager = manager
        bind = options.pop('bind', manager.engine)
        binds = options.pop('binds', {}) or manager.binds_map
        super(Session, self).__init__(bind=bind,
                                      binds=binds,
                                      **options)

Sometimes I use hybrid_property or association_proxy for many-to-many relationships where I will prefix the base relationship with _<name> and then have the hybrid or association defined as <name>. When I call to_dict, I generally want to exclude the _<name> and only have <name>. It may be useful to support a convention where any descriptor that begins with __ (double underscore) would be ignored by default when calling __to_dict__/to_dict.

Hi, I'm finding great use in this package to share code base between server and CLI!

However, I ran into an issue where I ended up in an endless loop of calls to to_dict.

This is the affected code snippet:

    def to_dict(self):
        """Return dict representation of model by filtering fields using
        :attr:`__to_dict__`.
        """
        data = {}
        relationships = self.relationships()

        for field in self.__to_dict__:
            value = getattr(self, field)

            # Nest calls to `to_dict`. Try to find method on base value,
            # sequence values, or dict values.
            if hasattr(value, 'to_dict'):
                value = value.to_dict()
            elif is_sequence(value):
                value = [v.to_dict() if hasattr(v, 'to_dict') else v
                         for v in value]
            elif isinstance(value, dict):
                value = dict([(k, v.to_dict() if hasattr(v, 'to_dict') else v)
                              for k, v in iteritems(value)])
            elif field in relationships and value is None:
                value = {}

            data[field] = value

        return data

What I'm thinking might cause this to happen is the fact that if I have a relationship, say "parent" with a backref, "children" this loop will simply never end, no?

See:

Should be:

self.Model = Model or model.make_declarative_base()

If a model has a relationship to a model with a backref, __to_dict__ will recursively try to display the relationship cycle.

In my own project I solved the problem by using a base inherited from ModelBase that overrides __to_dict__ to skip relationships in recursing.

class Base(alchy.ModelBase):
    # Alchy seems broken in its default to_dict implementation
    # It appears to traverse relationships with multiple values, possibly causing
    #  an infinite recursion.  E.g., users have questions and questions have users.
    @property
    def __to_dict__(self):
        keys = super().__to_dict__
        relationships = self.relationships()
        return set(k for k in keys if k not in relationships)

I have some issue with handling postgresql restart.
I get the following error:

sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) terminating connection due to administrator command
server closed the connection unexpectedly
    This probably means the server terminated abnormally
    before or while processing the request.

and my www app return 500.

when I repeat the same request I get:

Traceback (most recent call last):
  File "/usr/lib/python3.3/site-packages/sqlalchemy/engine/base.py", line 1069, in _execute_context
    conn = self.__connection
AttributeError: 'Connection' object has no attribute '_Connection__connection'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/lib/python3.3/site-packages/sqlalchemy/engine/base.py", line 1071, in _execute_context
    conn = self._revalidate_connection()
  File "/usr/lib/python3.3/site-packages/sqlalchemy/engine/base.py", line 391, in _revalidate_connection
    "Can't reconnect until invalid "
sqlalchemy.exc.InvalidRequestError: Can't reconnect until invalid transaction is rolled back

@dgilland thanks for the very excellent blog post!

I have a question about this comment:

Just a heads up, in Flask-SQLAlchemy v3.0.0 (unreleased), they've added a "model_class" argument that will essentially make the recipe here unnecessary: https://github.com/mitsuhiko/flask-sqlalchemy/blob/7d65e2dc1f4e798f3234aaef01e615e3db319b05/flask_sqlalchemy/__init__.py#L728

You said that pallets-eco/flask-sqlalchemy#328 basically makes alchy and Flask-Alchy unnecessary, but after reading the long discussion at pallets-eco/flask-sqlalchemy#250, it sounds like there is more work to be done. Can you comment on how each of those pull requests relates to your blog post?

Thanks and sorry for the ignorance.

P.S. I also added a comment on the blog - feel free to respond to one or the other instead of both.

In ModelBase.update() there is an argument to restrict updatable fields to Model.strict_update_fields. By default, strict update is disabled when calling update() and requires update(..., strict=True) to enable. A configuration property should be available to make strict=True the default. Something like ModelBase.__strict_update__ = False and update(..., strict=None).

Add an upsert() function which

• takes a dict
• queries database using filter_by()
• inserts if record not found
• otherwise fetches first
• returns inserted or fetched record