jfinkels / flask-restless Goto Github PK

Create a base class for database backends and refactor code to use this base class.

Create subclass implementations for SQLAlchemy, FlaskSQLAlchemy, and Elixir, then move on to other database abstraction libraries, like PeeWee, etc.

To be an approved Flask extension:

1. Relicense all significant contributions as 3-clause BSD.
2. Ensure that the code works for Python versions 2.5, 2.6, and 2.7.

See http://flask.pocoo.org/docs/extensiondev/#approved-extensions for more information.

Use the Request.json property to get parsed JSON instead of parsing it ourselves. Possible issue: how does that handle invalid JSON?

Automatically lowercase model names when creating APIs for models with uppercase characters.

For example, in docs/customizing.rst.

Figure out how to get the name of the table from an Elixir entity, then use that as the endpoint name for each specific model.

This is example how to do this
coleifer/flask-peewee@eaf9059

Incorporate Flask-Mimerender to allow for responses of multiple mimetypes?

Add location of examples/ directory to the documentation.

Depends on the resolution of this request to the maintainer of sphinxcontrib-httpdomain.

Use setupClass() to create the database file, and each setUp() to create the database tables and tearDown() to drop the database tables.

This should save on hard disk usage a bit.

May need to use dataType, contentType.

Support for Flask-SQLAlchemy is on the master branch.

Initial support for SQLAlchemy can be found on the sqlalchemy branch.

Support for Elixir can be found on the elixir branch.

Is there a simple way to merge these three so that APIManager can be configured to (possibly transparently) choose among the appropriate options, depending on how the user has defined his or her models?

The problem in definining views.py comes from the fact that each of the three has a slightly different way of doing things like querying, adding, deleting, and committing. One idea is to have an abstract class which represents this interface (that is, it might provide "add", "delete", "get_by" methods), and then implement a new class for each backend.

Currently search.evaluate_functions is run on all instances of a model instead of only the ones matching the query parameters. Change this so that the functions are evaluated only on the SQLALchemy query which corresponds to the requested search.

If, for example, person 1 is accessible at URL /api/person/1, allow accessing for example, /api/person/1/pets to get all pets related to person 1.

Remove allow_patch_many argument from flask_restless.views.API class. Replace that logic with simpler logic in flask_restless.manager.APIManager.create_api, which can just not expose PATCH method on the collection endpoint (e.g. /api/people).

Don't manually convert date objects to strings in to_dict. Instead use the default Flask jsonify behavior introduced in pallets/flask#471, which serializes dates automatically. This will require Flask version 0.9.

Python 2.5 fails to install Flask in Travis-CI.

The flask_restless.views.API class must define the put method to delegate to the patch method. Add tests which ensure that PUT and PATCH requests do the same thing.

create_api() should return some information about the created endpoint or blueprint so that users know where the endpoints are.

Add a include_fields=[...] keyword argument to APIManager.create_api which allows the user to specify which columns will be exposed when a GET request is issued.

Possibly add two: read_fields and write_fields.

Hey all,

Trying to bundle a set of requirements for distribution inside a network behind a firewall that doesn't allow easy access for pip to get at github, etc.

This eventually results in this error for me on Mac OSX 10.7.2

pip bundle app.pybundle Flask-Restless 
...
OSError: [Errno 2] No such file or directory: '/path_to_application/build-bundle3/Flask-Restless/README'

It seems to happen when installing different packages but always says that the README file doesn't exist. Can anyone replicate this?

pallets/flask#430 provides functionality that is the same as the custom jsonify_status_code function we have written. When Flask 0.9 is released, we should 1. mark that as the minimum required version, and 2. change the code to use the new Flask functionality.

APIManager should accept either a Session class or a Session instance. If it is a class, instantiate the class using scoped_session.

Capture errors occurring due to validation and pass that on as an error message to the client.

Document the PATCH /api/people?q=<searchjson> functionality in requestformat.rst.

For example, as in views._get_by.

https://github.com/flosch/simpleapi

Hi,

I'm wondering if this can be used with vanilla sqlalchemy?

And if not, what features of flask-sqlalchemy are being used?

Thankyou.

This will be an issue when we merge branch use-request-json into master.

Requests with the incorrect Content-Type header should respond with an error message suggesting that the client should set the Content-Type to application/json.

Improve the score of this project according to cheesecake_index to 100%.

Allow users to specify which methods will require authentication. For example, a user should not be allowed to POST unless logged in.

exception.message is deprecated in Python 2.6. Possibly solution, give a custom message for each exception.

Update the license headers in each file to use the correct header, stating "This file is a part of Flask-Restless" instead of "This program", since Flask-Restless is a multi-file program.

See the bottom of http://www.gnu.org/licenses/agpl.html for the correct text.

We want our APIs to have a uniform application/json content-type for all requests and responses, so we should use jsonify for all responses.

For example, Travis-CI: http://about.travis-ci.org/docs/user/languages/python/

Due to 2e4d1e7, we are using whatever json module Flask picks up on the system. If it is simplejson, it will decode strings within a JSON string to objects of type str, not objects of type unicode. This is an issue if the database model has a field of type Unicode; SQLAlchemy raises a warning when writing a str to a unicode field. This occurs in methods which write to the database, like patch or post.

Provide SQLAlchemy support in addition to Elixir support. By doing this, existing code which uses the Flask-SQLAlchemy extension can easily add our extension for use with their models.

Create an example (and a test case) in which a single model is affected by different URL endpoints. For example, /add/person for POST requests, /delete/person/:id for DELETE requests, /update/person/:id for PATCH requests, etc.

Allow user to specify different endpoint names in create_api.

Once pallets/flask#471 is pulled into Flask (hopefully in Flask version 0.9), replace _to_dict and _to_dict_include in views.py with a custom subclass of Flask.JSONEncoder which checks which columns to include. This might be problematic because of the deep and exclude arguments of _to_dict.

Two tests about validation is skipped silently. For more details see pallets/flask#485
The goal is to make setup.py test use unittest2 instead of unittest.
The problem of existing implementation here aed13d5 is that setup.py test does not install neither test requirements, nor the one needed for flask-restless.

Flask-Peewee seems to implement much of what we do, but for the PeeWee ORM instead of for SQLAlchemy. Can we adapt their code?