Let's say I have a model like this:
class User(db.Model): id = db.Column(db.Integer, primary_key=True) hometown = db.Column(db.String(140)) university = db.Column(db.String(140))
To get a list of users from New York, this is my query:
User.query.filter_by(hometown='New York').all()
To get a list of users who go to USC, this is my query:
User.query.filter_by(university='USC').all()
And to get a list of users from New York, and who go to USC, this is my query:
User.query.filter_by(hometown='New York').filter_by(university='USC').all()
Now, I would like to dynamically generate these queries based on the value of a variable.
For example, my variable might look like this:
{'hometown': 'New York'}
Or like this:
{'university': 'USC'}
... Or even like this:
[{'hometown': 'New York'}, {'university': 'USC'}]
Can you help me out with writing a function which takes a dictionary (or list of dictionaries) as an input, and then dynamically builds the correct sqlalchemy query?
If I try to use a variable for the keyword, I get this err:
key = 'university' User.query.filter_by(key='USC').all() InvalidRequestError: Entity '' has no property 'key'
Secondly, I am not sure how to chain multiple filter_by expressions together dynamically.
I can explicitly, call out a filter_by expression, but how do I chain several together based on a variable?
Hope this makes more sense.
Thanks!
SQLAlchemy's filter_by takes keyword arguments:
filter_by(**kwargs)
In other words, the function will allow you to give it any keyword parameter. This is why you can use any keyword that you want in your code: SQLAlchemy basically sees the arguments a dictionary of values. See the Python tutorial for more information on keyword arguments.
So that allows the developers of SQLAlchemy to receive an arbitrary bunch of keyword arguments in a dictionary form. But you're asking for the opposite: can you pass an arbitrary bunch of keyword arguments to a function?
It turns out that in Python you can, using a feature called unpacking. Simply create the dictionary of arguments and pass it to the function preceded by **, like so:
kwargs = {'hometown': 'New York', 'university' : 'USC'} User.query.filter_by(**kwargs) # This above line is equivalent to saying... User.query.filter_by(hometown='New York', university='USC')
filter_by(**request.args) doesn't work well if you have non-model query parameters, like page for pagination, otherwise you get errors like these:
InvalidRequestError: Entity '' has no property 'page'
I use something like this which ignores query parameters not in the model:
builder = MyModel.query for key in request.args: if hasattr(MyModel, key): vals = request.args.getlist(key) # one or many builder = builder.filter(getattr(MyModel, key).in_(vals)) if not 'page' in request.args: resources = builder.all() else: resources = builder.paginate( int(request.args['page'])).items
Considering a model with a column called valid, something like this will work:
curl -XGET "http://0.0.0.0/mymodel_endpoint?page=1&valid=2&invalid=whatever&valid=1"
invalid will be ignored, and page is available for pagination and best of all, the following SQL will be generated: WHERE mymodel.valid in (1,2)
(get the above snippet for free if you use this boilerplate-saving module)
