Deprecation Notice

Cliquet has been folded into the Kinto package under the new name kinto.core. You should update your code to import kinto.core instead of cliquet, and all cliquet.* settings are now prefixed with kinto instead. See https://mail.mozilla.org/pipermail/kinto/2016-May/000119.html for more about the background for this decision.

Cliquet

pypi travis Coverage Documentation Status

Cliquet is a toolkit to ease the implementation of HTTP microservices, such as data-driven REST APIs.

A cliquet, or ratchet, is a mechanical device that allows continuous linear or rotary motion in only one direction while preventing motion in the opposite direction.

A cliquet provides some basic but essential functionality — efficient in a variety of contexts, from bikes rear wheels to most advanced clockmaking!

Table of content

Rationale

Cliquet is a toolkit to ease the implementation of HTTP microservices. It is mainly focused on data-driven REST APIs (aka CRUD).

Philosophy

  • KISS;
  • No magic;
  • Works with defaults;
  • Easy customization;
  • Straightforward component substitution.

Cliquet doesn’t try to be a framework: any project built with Cliquet will expose a well defined HTTP protocol for:

  • Collection and records manipulation;
  • HTTP status and headers handling;
  • API versioning and deprecation;
  • Errors formatting.

This protocol is an implementation of a series of good practices (followed at Mozilla Services and elsewhere).

The goal is to produce standardized APIs, which follow some well known patterns, encouraging genericity in clients code [1].

Of course, Cliquet can be extended and customized in many ways. It can also be used in any kind of project, for its tooling, utilities and helpers.

Features

It is built around the notion of resources: resources are defined by sub-classing, and Cliquet brings up the HTTP endpoints automatically.

Records and synchronization

  • Collection of records by user;
  • Optional validation from schema;
  • Sorting and filtering;
  • Pagination using continuation tokens;
  • Polling for collection changes;
  • Record race conditions handling using preconditions headers;
  • Notifications channel (e.g. run asynchronous tasks on changes).

Generic endpoints

  • Hello view at root url;
  • Heartbeat for monitoring;
  • Batch operations;
  • API versioning and deprecation;
  • Errors formatting;
  • Backoff and Retry-After headers.

Toolkit

_images/cliquet-base.png

Cliquet brings a set of simple but essential features to build APIs.

  • Configuration through INI files or environment variables;
  • Pluggable storage and cache backends;
  • Pluggable authentication and user groups management;
  • Pluggable authorization and permissions management;
  • Structured logging;
  • Monitoring tools;
  • Profiling tools.

Pluggable components can be replaced by another one via configuration.

Dependencies

Cliquet is built on the shoulders of giants:

Everything else is meant to be pluggable and optional.

_images/cliquet-mozilla.png

Examples of configuration for a Cliquet application in production.

  • Basic Auth, FxA OAuth2 or any other source of authentication;
  • Default or custom class for authorization logics;
  • PostgreSQL for storage;
  • Redis for key-value cache with expiration;
  • StatsD metrics;
  • Sentry reporting via logging;
  • NewRelic database profiling (for development);
  • Werkzeug Python code profiling (for development).

A Cliquet application can change or force default values for any setting.

Built with Cliquet

Some applications in the wild built with Cliquet:

  • Reading List, a service to synchronize articles between devices;
  • Kinto, a service to store and synchronize schema-less data.
  • Syncto, a service to access Firefox Sync using kinto.js.
  • Please contact us to add yours.

Context

(to be done)

  • Cloud Services team at Mozilla
  • ReadingList project story
  • Firefox Sync
  • Cloud storage
  • Firefox OS User Data synchronization and backup

Vision

General

Any application built with Cliquet:

  • follows the same conventions regarding the HTTP API;
  • takes advantage of its component pluggability;
  • can be extended using custom code or Pyramid external packages;

Let’s build a sane ecosystem for microservices in Python!

Roadmap

The future features we plan to implement in Cliquet are currently driven by the use-cases we meet internally at Mozilla. Most notable are:

  • Attachments on records (e.g. Remote Storage compatibility);
  • Records generic indexing (e.g. streaming records to ElasticSearch).
  • ... come and discuss enhancements in the issue tracker!

Similar projects

  • Python Eve, built on Flask and MongoDB;
  • Please contact us to add more if any.

Since the protocol is language independant and follows good HTTP/REST principles, in the long term Cliquet should become only one among several server implementations.

Note

We encourage you to implement a clone of this project — using Node.js, Asyncio, Go, Twisted, Django or anything else — following the same protocol!

[1]Switch from custom protocol to JSON-API spec is being discussed.
[2]Currently, the clients code was not extracted from the client projects, such as RL Web client (React.js), Android RL sync (Java) or Firefox RL client (asm.js).
[3]See https://unhosted.org.

Getting started

Installation

$ pip install cliquet

More details about installation and storage backend is provided in a dedicated section.

Start a Pyramid project

As detailed in Pyramid documentation, create a minimal application, or use its scaffolding tool:

$ pcreate -s starter MyProject

Include Cliquet

In the application main file (e.g. MyProject/myproject/__init__.py), just add some extra initialization code:

import pkg_resources

import cliquet
from pyramid.config import Configurator

# Module version, as defined in PEP-0396.
__version__ = pkg_resources.get_distribution(__package__).version


def main(global_config, **settings):
    config = Configurator(settings=settings)

    cliquet.initialize(config, __version__)
    return config.make_wsgi_app()

By doing that, basic features like authentication, monitoring, error formatting, deprecation indicators are now available, and rely on configuration present in development.ini.

Run!

With some backends, like PostgreSQL, some tables and indices have to be created. A generic command is provided to accomplish this:

$ cliquet --ini development.ini migrate

Like any Pyramid application, it can be served locally with:

$ pserve development.ini --reload

A hello view is now available at http://localhost:6543/v0/ (As well as basic endpoints like the utilities).

The next steps will consist in building a custom application using Cornice or the Pyramid ecosystem.

But most likely, it will consist in defining REST resources using Cliquet python API !

Authentication

Currently, if no authentication is set in settings, Cliquet relies on Basic Auth. It will associate a unique user id for every user/password combination.

Using HTTPie, it is as easy as:

$ http -v http://localhost:6543/v0/ --auth user:pass

Note

In the case of Basic Auth, there is no need of registering a user/password. Pick any combination, and include them in each request.

Define resources

In order to define a resource, inherit from cliquet.resource.UserResource, in a subclass, in myproject/views.py for example:

from cliquet import resource

@resource.register()
class Mushroom(resource.UserResource):
    # No schema yet.
    pass

In application initialization, make Pyramid aware of it:

def main(global_config, **settings):
    config = Configurator(settings=settings)

    cliquet.initialize(config, __version__)
    config.scan("myproject.views")
    return config.make_wsgi_app()

In order to bypass the installation and configuration of Redis or PostgreSQL, specify the «in-memory» backends in development.ini:

# development.ini
cliquet.cache_backend = cliquet.cache.memory
cliquet.storage_backend = cliquet.storage.memory
cliquet.permission_backend = cliquet.permission.memory

A Mushroom resource API is now available at the /mushrooms/ URL.

It will accept a bunch of REST operations, as defined in the API section.

Warning

Without schema, a resource will not store any field at all!

The next step consists in attaching a schema to the resource, to control what fields are accepted and stored.

Schema validation

It is possible to validate records against a predefined schema, associated to the resource.

Currently, only Colander is supported, and it looks like this:

import colander
from cliquet import resource


class MushroomSchema(resource.ResourceSchema):
    name = colander.SchemaNode(colander.String())


@resource.register()
class Mushroom(resource.UserResource):
    mapping = MushroomSchema()

What’s next ?

Configuration

See Configuration to customize the application settings, such as authentication, storage or cache backends.

Resource customization

See the resource documentation to specify custom URLs, schemaless resources, read-only fields, unicity constraints, record pre-processing...

Advanced initialization

cliquet.initialize(config, version=None, project_name='', default_settings=None)

Initialize Cliquet with the given configuration, version and project name.

This will basically include cliquet in Pyramid and set route prefix based on the specified version.

Parameters:
  • config (Configurator) – Pyramid configuration
  • version (str) – Current project version (e.g. ‘0.0.1’) if not defined in application settings.
  • project_name (str) – Project name if not defined in application settings.
  • default_settings (dict) – Override cliquet default settings values.

Beyond Cliquet

Cliquet is just a component! The application can still be built and extended using the full Pyramid ecosystem.

See the dedicated section for examples of Cliquet extensions.

HTTP Protocol

API Versioning

The HTTP API exposed by the service will be consumed by clients, like a Javascript client.

The HTTP API is subject to changes. It follows the Cliquet Protocol.

When the HTTP API is changed, its version is incremented. The HTTP API version follows a Semantic Versioning pattern and uses this rule to be incremented:

  1. any change to the HTTP API that is backward compatible increments the MINOR number, and the modification in the documentation should reflect this with a header like “Added in 1.x”.
  2. any change to the HTTP API that is backward incompatible increments the MAJOR number, and the differences are summarized at the begining of the documentation, a new document for that MAJOR version is created.

Note

We’re not using the PATCH level of Semantic Versioning, since bug fixes have no impact on the exposed HTTP API; if they do MINOR or MAJOR should be incremented.

We want to avoid MAJOR changes as much as possible in the future, and stick with 1.x as long as we can.

A client that interacts with the service can query the server to know what is its HTTP API version. This is done with a query on the root view, as described in the root API description.

If a client relies on a feature that was introduced at a particular version, it should check that the server implements the minimal required version.

The JSON response body contains an http_api_version key which value is the MAJOR.MINOR version.

Authentication

Depending on the authentication policies initialized in the application, the HTTP method to authenticate requests may differ.

A policy based on OAuth2 bearer tokens is recommended, but not mandatory. See configuration for further information.

In the current implementation, when multiple policies are configured, user identifiers are isolated by policy. In other words, there is no way to access the same set of records using different authentication methods.

By default, a relatively secure Basic Auth is enabled.

Basic Auth

If enabled in configuration, using a Basic Auth token will associate a unique user identifier to an username/password combination.

Authorization: Basic <basic_token>

The token shall be built using this formula base64("username:password").

Empty passwords are accepted, and usernames can be anything (custom, UUID, etc.)

If the token has an invalid format, or if Basic Auth is not enabled, this will result in a 401 error response.

Warning

Since user id is derived from username and password, there is no way to change the password without loosing access to existing records.

OAuth Bearer token

If the configured authentication policy uses OAuth2 bearer tokens, authentication shall be done using this header:

Authorization: Bearer <oauth_token>

The policy will verify the provided OAuth2 bearer token on a remote server.

notes:If the token is not valid, this will result in a 401 error response.

Firefox Accounts

In order to enable authentication with Firefox Accounts, install and configure mozilla-services/cliquet-fxa.

Resource endpoints

All endpoints URLs are prefixed by the major version of the HTTP API (e.g /v1 for 1.4).

e.g. the URL for all the endpoints is structured as follows::

https://<server name>/<api MAJOR version>/<further instruction>

The full URL prefix will be implied throughout the rest of this document and it will only describe the <further instruction> part.

GET /{collection}

Requires authentication

Returns all records of the current user for this collection.

The returned value is a JSON mapping containing:

  • data: the list of records, with exhaustive fields;

A Total-Records response header indicates the total number of records of the collection.

A Last-Modified response header provides a human-readable (rounded to second) of the current collection timestamp.

For cache and concurrency control, an ETag response header gives the value that consumers can provide in subsequent requests using If-Match and If-None-Match headers (see section about timestamps).

Request:

GET /articles HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Host: localhost:8000

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length, ETag, Next-Page, Total-Records, Last-Modified
Content-Length: 436
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:08:11 GMT
Last-Modified: Mon, 12 Apr 2015 11:12:07 GMT
ETag: "1430222877724"
Total-Records: 2

{
    "data": [
        {
            "id": "dc86afa9-a839-4ce1-ae02-3d538b75496f",
            "last_modified": 1430222877724,
            "title": "MoCo",
            "url": "https://mozilla.com",
        },
        {
            "id": "23160c47-27a5-41f6-9164-21d46141804d",
            "last_modified": 1430140411480,
            "title": "MoFo",
            "url": "https://mozilla.org",
        }
    ]
}
Filtering

Single value

  • /collection?field=value

Minimum and maximum

Prefix field name with min_ or max_:

  • /collection?min_field=4000

Note

The lower and upper bounds are inclusive (i.e equivalent to greater or equal).

Note

lt_ and gt_ can also be used to exclude the bound.

Multiple values

Prefix field with in_ and provide comma-separated values.

  • /collection?in_status=1,2,3

Exclude

Prefix field name with not_:

  • /collection?not_field=0

Exclude multiple values

Prefix field name with exclude_:

  • /collection?exclude_field=0,1

Note

Will return an error if a field is unknown.

Note

The ETag and Last-Modified response headers will always be the same as the unfiltered collection.

Sorting
  • /collection?_sort=-last_modified,field

Note

Ordering on a boolean field gives true values first.

Note

Will return an error if a field is unknown.

Counting

In order to count the number of records, for a specific field value for example, without fetching the actual collection, a HEAD request can be used. The Total-Records response header will then provide the total number of records.

See batch endpoint to count several collections in one request.

Polling for changes

The _since parameter is provided as an alias for gt_last_modified.

  • /collection?_since=1437035923844

When filtering on last_modified every deleted records will appear in the list with a deleted flag and a last_modified value that corresponds to the deletion event.

If the If-None-Match: "<timestamp>" request header is provided as described in the section about timestamps and if the collection was not changed, a 304 Not Modified response is returned.

Note

The _before parameter is also available, and is an alias for lt_last_modified (strictly inferior).

Note

_since and _before also accept a value between quotes (") as it would be returned in the ETag response header (see response timestamps).

Request:

GET /articles?_since=1437035923844 HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Host: localhost:8000

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length, ETag, Next-Page, Total-Records, Last-Modified
Content-Length: 436
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:08:11 GMT
Last-Modified: Mon, 12 Apr 2015 11:12:07 GMT
ETag: "1430222877724"
Total-Records: 2

{
    "data": [
        {
            "id": "dc86afa9-a839-4ce1-ae02-3d538b75496f",
            "last_modified": 1430222877724,
            "title": "MoCo",
            "url": "https://mozilla.com",
        },
        {
            "id": "23160c47-27a5-41f6-9164-21d46141804d",
            "last_modified": 1430140411480,
            "title": "MoFo",
            "url": "https://mozilla.org",
        },
        {
            "id": "11130c47-37a5-41f6-9112-32d46141804f",
            "deleted": true,
            "last_modified": 1430140411480
        }
    ]
}
Paginate

If the _limit parameter is provided, the number of records returned is limited.

If there are more records for this collection than the limit, the response will provide a Next-Page header with the URL for the Next-Page.

When there is no more Next-Page response header, there is nothing more to fetch.

Pagination works with sorting, filtering and polling.

Note

The Next-Page URL will contain a continuation token (_token).

It is recommended to add precondition headers (If-Match or If-None-Match), in order to detect changes on collection while iterating through the pages.

Partial response

If the _fields parameter is provided, only the fields specified are returned. Fields are separated with a comma.

This is vital in mobile contexts where bandwidth usage must be optimized.

Nested objects fields are specified using dots (e.g. address.street).

Note

The id and last_modified fields are always returned.

Request:

GET /articles?_fields=title,url
Accept: application/json
Authorization: Basic bWF0Og==
Host: localhost:8000

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length, ETag, Next-Page, Total-Records, Last-Modified
Content-Length: 436
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:08:11 GMT
Last-Modified: Mon, 12 Apr 2015 11:12:07 GMT
ETag: "1430222877724"
Total-Records: 2

{
    "data": [
        {
            "id": "dc86afa9-a839-4ce1-ae02-3d538b75496f",
            "last_modified": 1430222877724,
            "title": "MoCo",
            "url": "https://mozilla.com",
        },
        {
            "id": "23160c47-27a5-41f6-9164-21d46141804d",
            "last_modified": 1430140411480,
            "title": "MoFo",
            "url": "https://mozilla.org",
        }
    ]
}
List of available URL parameters
  • <prefix?><field name>: filter by value(s)
  • _since, _before: polling changes
  • _sort: order list
  • _limit: pagination max size
  • _token: pagination token
  • _fields: filter the fields of the records

Filtering, sorting, partial responses and paginating can all be combined together.

  • /collection?_sort=-last_modified&_limit=100&_fields=title
HTTP Status Codes
  • 200 OK: The request was processed
  • 304 Not Modified: Collection did not change since value in If-None-Match header
  • 400 Bad Request: The request querystring is invalid
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type
  • 412 Precondition Failed: Collection changed since value in If-Match header

POST /{collection}

Requires authentication

Used to create a record in the collection. The POST body is a JSON mapping containing:

  • data: the values of the resource schema fields;
  • permissions: optional a json dict containing the permissions for the record to be created.

The POST response body is a JSON mapping containing:

  • data: the newly created record, if all posted values are valid;
  • permissions: optional a json dict containing the permissions for the requested resource.

If the If-Match: "<timestamp>" request header is provided as described in the section about timestamps, and if the collection has changed meanwhile, a 412 Precondition failed error is returned.

If the If-None-Match: * request header is provided, and if the provided data contains an id field, and if there is already an existing record with this id, a 412 Precondition failed error is returned.

Request:

POST /articles HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Content-Type: application/json; charset=utf-8
Host: localhost:8000

{
    "data": {
        "title": "Wikipedia FR",
        "url": "http://fr.wikipedia.org"
    }
}

Response:

HTTP/1.1 201 Created
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length
Content-Length: 422
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:35:02 GMT

{
    "data": {
        "id": "cd30c031-c208-4fb9-ad65-1582d2a7ad5e",
        "last_modified": 1430224502529,
        "title": "Wikipedia FR",
        "url": "http://fr.wikipedia.org"
    }
}
Validation

If the posted values are invalid (e.g. field value is not an integer) an error response is returned with status 400.

See details on error responses.

Conflicts

Since some fields can be defined as unique per collection, some conflicts may appear when creating records.

Note

Empty values are not taken into account for field unicity.

Note

Deleted records are not taken into account for field unicity.

If a conflict occurs, an error response is returned with status 409. A details attribute in the response provides the offending record and field name. See dedicated section about errors.

Timestamp

When a record is created, the timestamp of the collection is incremented.

It is possible to force the timestamp if the specified record has a last_modified field.

If the specified timestamp is in the past, the collection timestamp does not take the value of the created record but is bumped into the future as usual.

HTTP Status Codes
  • 200 OK: This record already exists, the one stored on the database is returned
  • 201 Created: The record was created
  • 400 Bad Request: The request body is invalid
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type
  • 409 Conflict: Unicity constraint on fields is violated
  • 412 Precondition Failed: Collection changed since value in If-Match header
  • 415 Unsupported Media Type: The client request was not sent with a correct Content-Type

DELETE /{collection}

Requires authentication

Delete multiple records. Disabled by default, see Configuration.

The DELETE response is a JSON mapping containing:

  • data: list of records that were deleted, without schema fields.

It supports the same filtering capabilities as GET.

If the If-Match: "<timestamp>" request header is provided, and if the collection has changed meanwhile, a 412 Precondition failed error is returned.

Request:

DELETE /articles HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Host: localhost:8000

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length
Content-Length: 193
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:38:36 GMT

{
    "data": [
        {
            "deleted": true,
            "id": "cd30c031-c208-4fb9-ad65-1582d2a7ad5e",
            "last_modified": 1430224716097
        },
        {
            "deleted": true,
            "id": "dc86afa9-a839-4ce1-ae02-3d538b75496f",
            "last_modified": 1430224716098
        }
    ]
}
HTTP Status Codes
  • 200 OK: The records were deleted
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 405 Method Not Allowed: This endpoint is not available
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type
  • 412 Precondition Failed: Collection changed since value in If-Match header

GET /{collection}/<id>

Requires authentication

Returns a specific record by its id. The GET response body is a JSON mapping containing:

  • data: the record with exhaustive schema fields;
  • permissions: optional a json dict containing the permissions for the requested record.

If the If-None-Match: "<timestamp>" request header is provided, and if the record has not changed meanwhile, a 304 Not Modified is returned.

Request:

GET /articles/d10405bf-8161-46a1-ac93-a1893d160e62 HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Host: localhost:8000

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length, ETag, Last-Modified
Content-Length: 438
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:42:42 GMT
ETag: "1430224945242"

{
    "data": {
        "id": "d10405bf-8161-46a1-ac93-a1893d160e62",
        "last_modified": 1430224945242,
        "title": "No backend",
        "url": "http://nobackend.org"
    }
}
HTTP Status Code
  • 200 OK: The request was processed
  • 304 Not Modified: Record did not change since value in If-None-Match header
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type
  • 412 Precondition Failed: Record changed since value in If-Match header

DELETE /{collection}/<id>

Requires authentication

Delete a specific record by its id.

The DELETE response is the record that was deleted. The DELETE response is a JSON mapping containing:

  • data: the record that was deleted, without schema fields.

If the record is missing (or already deleted), a 404 Not Found is returned. The consumer might decide to ignore it.

If the If-Match request header is provided, and if the record has changed meanwhile, a 412 Precondition failed error is returned.

Note

Once deleted, a record will appear in the collection when polling for changes, with a deleted status (delete=true) and will have most of its fields empty.

Timestamp

When a record is deleted, the timestamp of the collection is incremented.

It is possible to force the timestamp by passing it in the querystring with ?last_modified=<value>.

If the specified timestamp is in the past, the collection timestamp does not take the value of the deleted record but is bumped into the future as usual.

HTTP Status Code
  • 200 OK: The record was deleted
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type.
  • 412 Precondition Failed: Record changed since value in If-Match header

PUT /{collection}/<id>

Requires authentication

Create or replace a record with its id. The PUT body is a JSON mapping containing:

  • data: the values of the resource schema fields;
  • permissions: optional a json dict containing the permissions for the record to be created/replaced.

The PUT response body is a JSON mapping containing:

  • data: the newly created/updated record, if all posted values are valid;
  • permissions: optional the newly created permissions dict, containing the permissions for the created record.

Validation and conflicts behaviour is similar to creating records (POST).

If the If-Match: "<timestamp>" request header is provided as described in the section about timestamps, and if the record has changed meanwhile, a 412 Precondition failed error is returned.

If the If-None-Match: * request header is provided and if there is already an existing record with this id, a 412 Precondition failed error is returned.

Request:

PUT /articles/d10405bf-8161-46a1-ac93-a1893d160e62 HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Content-Type: application/json; charset=utf-8
Host: localhost:8000

{
    "data": {
        "title": "Static apps",
        "url": "http://www.staticapps.org"
    }
}

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length
Content-Length: 439
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:46:36 GMT
ETag: "1430225196396"

{
    "data": {
        "id": "d10405bf-8161-46a1-ac93-a1893d160e62",
        "last_modified": 1430225196396,
        "title": "Static apps",
        "url": "http://www.staticapps.org"
    }
}
Timestamp

When a record is created or replaced, the timestamp of the collection is incremented.

It is possible to force the timestamp if the specified record has a last_modified field.

For replace, if the specified timestamp is less or equal than the existing record, the value is simply ignored and the timestamp is bumped into the future as usual.

For creation, if the specified timestamp is in the past, the collection timestamp does not take the value of the created/updated record but is bumped into the future as usual.

HTTP Status Code
  • 201 Created: The record was created
  • 200 OK: The record was replaced
  • 400 Bad Request: The record is invalid
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type.
  • 409 Conflict: If replacing this record violates a field unicity constraint
  • 412 Precondition Failed: Record was changed or deleted since value in If-Match header.
  • 415 Unsupported Media Type: The client request was not sent with a correct Content-Type.

PATCH /{collection}/<id>

Requires authentication

Modify a specific record by its id. The PATCH body is a JSON mapping containing:

  • data: a subset of the resource schema fields (key-value replace);
  • permissions: optional a json dict containing the permissions for the record to be modified.

The PATCH response body is a JSON mapping containing:

  • data: the modified record (full by default);
  • permissions: optional the modified permissions dict, containing the permissions for the modified record.

If a Response-Behavior request header is set to light, only the fields whose value was changed are returned. If set to diff, only the fields whose value became different than the one provided are returned.

Request:

PATCH /articles/d10405bf-8161-46a1-ac93-a1893d160e62 HTTP/1.1
Accept: application/json
Authorization: Basic bWF0Og==
Content-Type: application/json; charset=utf-8
Host: localhost:8000

{
    "data": {
        "title": "No Backend"
    }
}

Response:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: *
Access-Control-Expose-Headers: Backoff, Retry-After, Alert, Content-Length
Content-Length: 439
Content-Type: application/json; charset=UTF-8
Date: Tue, 28 Apr 2015 12:46:36 GMT
ETag: "1430225196396"

{
    "data": {
        "id": "d10405bf-8161-46a1-ac93-a1893d160e62",
        "last_modified": 1430225196396,
        "title": "No Backend",
        "url": "http://nobackend.org"
    }
}

If the record is missing (or already deleted), a 404 Not Found error is returned. The consumer might decide to ignore it.

If the If-Match: "<timestamp>" request header is provided as described in the section about timestamps, and if the record has changed meanwhile, a 412 Precondition failed error is returned.

Note

last_modified is updated to the current server timestamp, only if a field value was changed.

Note

JSON-Patch is currently not supported. Any help is welcomed though!

Read-only fields

If a read-only field is modified, a 400 Bad request error is returned.

Conflicts

If changing a record field violates a field unicity constraint, a 409 Conflict error response is returned (see error channel).

Timestamp

When a record is modified, the timestamp of the collection is incremented.

It is possible to force the timestamp if the specified record has a last_modified field.

If the specified timestamp is less or equal than the existing record, the value is simply ignored and the timestamp is bumped into the future as usual.

HTTP Status Code
  • 200 OK: The record was modified
  • 400 Bad Request: The request body is invalid, or a read-only field was modified
  • 401 Unauthorized: The request is missing authentication headers
  • 403 Forbidden: The user is not allowed to perform the operation, or the resource is not accessible
  • 406 Not Acceptable: The client doesn’t accept supported responses Content-Type.
  • 409 Conflict: If modifying this record violates a field unicity constraint
  • 412 Precondition Failed: Record changed since value in If-Match header
  • 415 Unsupported Media Type: The client request was not sent with a correct Content-Type.

Notes on permissions attribute

Shareable resources allow permissions management via the permissions attribute in the JSON payloads, along the data attribute. Permissions can be replaced or modified independently from data.

On a request, permissions is a JSON dict with the following structure:

"permissions": {<permission>: [<list_of_principals>]}

Where <permission> is the permission name (e.g. read, write) and <list_of_principals> should be replaced by an actual list of principals.

Example:

{
    "data": {
        "title": "No Backend"
    },
    "permissions": {
        "write": ["twitter:leplatrem", "group:ldap:42"],
        "read": ["system.Authenticated"]
    }
}

In a response, permissions contains the current permissions of the record (i.e. the modified version in case of a creation/modification).

Note

When a record is created or modified, the current user id is always added among the write principals.

Read more about leveraging resource permissions.

Batch operations

POST /batch

Requires authentication

The POST body is a mapping, with the following attributes:

  • requests: the list of requests
  • defaults: (optional) default requests values in common for all requests
Each request is a JSON mapping, with the following attribute:
  • method: HTTP verb
  • path: URI
  • body: a mapping
  • headers: (optional), otherwise take those of batch request
POST /batch HTTP/1.1
Accept: application/json
Accept-Encoding: gzip, deflate
Content-Length: 728
Host: localhost:8888
User-Agent: HTTPie/0.9.2

{
  "defaults": {
    "method" : "POST",
    "path" : "/articles",
  },
  "requests": [
    {
      "body" : {
        "data" : {
          "title": "MoFo",
          "url" : "http://mozilla.org",
          "added_by": "FxOS",
        },
        "permissions": {
          "read": ["system.Everyone"]
        }
      }
    },
    {
      "body" : {
        "data" : {
          "title": "MoCo",
          "url" : "http://mozilla.com"
          "added_by": "FxOS",
        }
      }
    },
    {
      "method" : "PATCH",
      "path" : "/articles/409",
      "body" : {
        "data" : {
          "read_position" : 3477
        }
      }
      "headers" : {
        "Response-Behavior": "light"
      }
    }
  ]
}

The response body is a list of all responses:

HTTP/1.1 200 OK
Access-Control-Expose-Headers: Retry-After, Content-Length, Alert, Backoff
Content-Length: 1674
Date: Wed, 17 Feb 2016 18:44:39 GMT
Server: waitress

{
  "responses": [
    {
      "status": 201,
      "path" : "/articles",
      "body" : {
        "data" : {
          "id": 411,
          "title": "MoFo",
          "url" : "http://mozilla.org",
          ...
        }
      },
      "headers": {
        ...
      }
    },
    {
      "status": 201,
      "path" : "/articles",
      "body" : {
        "data" : {
          "id": 412,
          "title": "MoCo",
          "url" : "http://mozilla.com",
          ...
        }
      },
      "headers": {
        ...
      }
    },
    {
      "status": 200,
      "path" : "/articles/409",
      "body" : {
        "data" : {
          "id": 409,
          "url": "...",
          ...
          "read_position" : 3477
        }
      },
      "headers": {
        ...
      }
    }
  ]
}
HTTP Status Codes
  • 200 OK: The request has been processed
  • 400 Bad Request: The request body is invalid
  • 50X: One of the sub-request has failed with a 50X status

Warning

Since the requests bodies are necessarily mappings, posting arbitrary data (like raw text or binary) is not supported.

Note

Responses are executed and provided in the same order than requests.

About transactions

The whole batch of requests is executed under one transaction only.

In order words, if one of the sub-request fails with a 503 status for example, then every previous operation is rolled back.

Important

With the current implementation, if a sub-request fails with a 4XX status (eg. 412 Precondition failed or 403 Unauthorized for example) the transaction is not rolled back.

Utility endpoints for OPS and Devs

GET /

The returned value is a JSON mapping containing:

Changed in version 2.12.

  • project_name: the name of the service (e.g. "reading list")

  • project_docs: The URL to the service documentation. (this document!)

  • project_version: complete application/project version ("3.14.116")

  • http_api_version: the MAJOR.MINOR version of the exposed HTTP API ("1.1") defined in configuration.

  • cliquet_protocol_version: the cliquet protocol version ("2")

  • url: absolute URI (without a trailing slash) of the API (can be used by client to build URIs)

  • eos: date of end of support in ISO 8601 format ("yyyy-mm-dd", undefined if unknown)

  • settings: a mapping with the values of relevant public settings for clients

    • batch_max_requests: Number of requests that can be made in a batch request.
    • readonly: Only requests with read operations are allowed.
  • capabilities: a mapping used by clients to detect optional features of the API.

    • Example:

      {
        "auth-fxa": {
          "description": "Firefox Account authentication",
          "url": "http://github.com/mozilla-services/cliquet-fxa"
        }
      }
      

Optional

  • user: A mapping with an id field for the currently connected user id. The field is not present when no Authorization header is provided.

Note

The project_version contains the source code version, whereas the http_api_version contains the exposed HTTP API version.

The source code of the service can suffer changes and have its project version incremented, without impacting the publicly exposed HTTP API.

The cliquet_protocol_version is an internal notion tracking the version for some aspects of the API (e.g. synchronization of REST resources, utilities endpoints, etc.). It will differ from the http_api_version since the service will provide additionnal endpoints and conventions.

GET /__heartbeat__

Return the status of each service the application depends on. The returned value is a JSON mapping containing:

  • storage true if storage backend is operational
  • cache true if cache backend operational
  • permission true if permission backend operational

If cliquet-fxa is installed, an additional key is present:

  • oauth true if authentication is operational

Return 200 if the connection with each service is working properly and 503 if something doesn’t work.

GET /__lbheartbeat__

Always return 200 with empty body.

Unlike the __heartbeat__ health check endpoint, which return an error when backends and other upstream services are unavailable, this should always return 200.

This endpoint is suitable for a load balancer membership test. It the load balancer cannot obtain a response from this endpoint, it will stop sending traffic to the instance and replace it.

Server timestamps

In order to avoid race conditions, each change is guaranteed to increment the timestamp of the related collection. If two changes happen at the same millisecond, they will still have two different timestamps.

The ETag header with the current timestamp of the collection for the current user will be given on collection endpoints.

ETag: "1432208041618"

On record enpoints, the ETag header value will contain the timestamp of the record.

In order to bypass costly and error-prone HTTP date parsing, ETag headers are not HTTP date values.

A human readable version of the timestamp (rounded to second) is provided though in the Last-Modified response headers:

Last-Modified: Wed May 20 17:22:38 2015 +0200

Changed in version 2.0: In previous versions, cache and concurrency control was handled using If-Modified-Since and If-Unmodified-Since. But since the HTTP date does not include milliseconds, they contained the milliseconds timestamp as integer. The current version using ETag is HTTP compliant (see original discussion.)

Note

The client may send If-Unmodified-Since or If-Modified-Since requests headers, but in the current implementation, they will be ignored.

Important

When collection is empty, its timestamp remains the same until new records are created.

Cache control

In order to check that the client version has not changed, a If-None-Match request header can be used. If the response is 304 Not Modified then the cached version is still good.

  GET
If-None-Match: “<timestamp>”
Changed meanwhile Return response content
Not changed Empty HTTP 304

Concurrency control

In order to prevent race conditions, like overwriting changes occured in the interim for example, a If-Match: "timestamp" request header can be used. If the response is 412 Precondition failed then the resource has changed meanwhile.

Concurrency control also allows to make sure a creation won’t overwrite any record using the If-None-Match: * request header.

The following table gives a summary of the expected behaviour of a resource:

  POST PUT PATCH DELETE
If-Match: “timestamp”
Changed meanwhile HTTP 412 HTTP 412 HTTP 412 HTTP 412
Not changed Create Overwrite Modify Delete
If-None-Match: *
Id exists HTTP 412 HTTP 412 No effect No effect
Id unknown Create Create No effect No effect

When the client receives a 412 Precondition failed, it can then choose to:

  • overwrite by repeating the request without concurrency control;
  • reconcile the resource by fetching, merging and repeating the request.

Replication

In order to replicate the timestamps when importing existing records, it is possible to force the last modified values.

When a record is created (via POST or PUT), the specified timestamp becomes the new collection timestamp if it is in the future (i.e. greater than current one). If it is in the past, the record is created with the timestamp in the past but the collection timestamp is bumped into the future as usual.

When a record is replaced, modified or deleted, if the specified timestamp is less or equal than the existing record, the value is simply ignored and the timestamp is bumped into the future as usual.

See the resource endpoints documentation.

Backoff indicators

Backoff header on heavy load

A Backoff header will be added to the success responses (>=200 and <400) when the server is under heavy load. It provides the client with a number of seconds during which it should avoid doing unnecessary requests.

Backoff: 30

Note

The back-off time is configurable on the server.

Note

In other implementations at Mozilla, there was X-Weave-Backoff and X-Backoff but the X- prefix for header has been deprecated since.

Retry-After indicators

A Retry-After header will be added if response is an error (>=500). See more details about error responses.

Error responses

Protocol description

Every response is JSON.

If the HTTP status is not OK (<200 or >=400), the response contains a JSON mapping, with the following attributes:

  • code: matches the HTTP status code (e.g 400)
  • errno: stable application-level error number (e.g. 109)
  • error: string description of error type (e.g. "Bad request")
  • message: context information (e.g. "Invalid request parameters")
  • info: online resource (e.g. URL to error details)
  • details: additional details (e.g. list of validation errors)

Example response

{
    "code": 412,
    "errno": 114,
    "error": "Precondition Failed",
    "message": "Resource was modified meanwhile",
    "info": "https://server/docs/api.html#errors",
}

Refer yourself to the ref:set of errors codes <errors>.

Retry-After indicators

A Retry-After header will be added to error responses (>=500), telling the client how many seconds it should wait before trying again.

Retry-After: 30

Precondition errors

As detailed in the timestamps section, it is possible to add concurrency control using ETag request headers.

When a concurrency error occurs, a 412 Precondition Failed error response is returned.

Additional information about the record currently stored on the server will be provided in the details field:

{
    "code": 412,
    "errno": 114,
    "error":"Precondition Failed"
    "message": "Resource was modified meanwhile",
    "details": {
        "existing": {
            "last_modified": 1436434441550,
            "id": "00dd028f-16f7-4755-ab0d-e0dc0cb5da92",
            "title": "Original title"
        }
    },
}

Conflict errors

When a record violates unicity constraints, a 409 Conflict error response is returned.

Additional information about conflicting record and field name will be provided in the details field.

{
    "code": 409,
    "errno": 122,
    "error": "Conflict",
    "message": "Conflict of field url on record eyjafjallajokull"
    "info": "https://server/docs/api.html#errors",
    "details": {
        "field": "url",
        "record": {
            "id": "eyjafjallajokull",
            "last_modified": 1430140411480,
            "url": "http://mozilla.org"
        }
    }
}

Validation errors

When multiple validation errors occur on a request, the first one is presented in the message.

The full list of validation errors is provided in the details field.

{
    "code": 400,
    "errno": 109,
    "error": "Bad Request",
    "message": "Invalid posted data",
    "info": "https://server/docs/api.html#errors",
    "details": [
        {
            "description": "42 is not a string: {'name': ''}",
            "location": "body",
            "name": "name"
        }
    ]
}

Deprecation

A track of the client version will be kept to know after which date each old version can be shutdown.

The date of the end of support is provided in the API root URL (e.g. /v0)

Using the Alert response header, the server can communicate any potential warning messages, information, or other alerts.

The value is JSON mapping with the following attributes:

  • code: one of the strings "soft-eol" or "hard-eol";
  • message: a human-readable message (optional);
  • url: a URL at which more information is available (optional).

A 410 Gone error response can be returned if the client version is too old, or the service had been remplaced with a new and better service using a new protocol version.

See details in Configuration to activate deprecation.

Internals

Installation

By default, a Cliquet application persists the records and cache in a local Redis.

Using the application configuration, other backends like « in-memory » or PostgreSQL can be enabled afterwards.

Supported Python versions

Cliquet supports Python 2.7, Python 3.4 and PyPy.

Distribute & Pip

Installing Cliquet with pip:

pip install cliquet

For PostgreSQL and monitoring support:

pip install cliquet[postgresql,monitoring]

Note

When installing cliquet with postgresql support in a virtualenv using the PyPy interpreter, the psycopg2cffi PostgreSQL database adapter will be installed, instead of the traditional psycopg2, as it provides significant performance improvements.

If everything is under control python-wise, jump to the next chapter. Otherwise please find more details below.

Python 3.4

Linux
sudo apt-get install python3.4-dev
OS X
brew install python3

Cryptography libraries

Linux

On Debian / Ubuntu based systems:

apt-get install libffi-dev libssl-dev

On RHEL-derivatives:

yum install libffi-devel openssl-devel
OS X

Assuming brew is installed:

brew install libffi openssl pkg-config

Install Redis

Linux

On debian / ubuntu based systems:

apt-get install redis-server

or:

yum install redis
OS X

Assuming brew is installed, Redis installation becomes:

brew install redis

To restart it (Bug after configuration update):

brew services restart redis

Install PostgreSQL

Client libraries only

Install PostgreSQL client headers:

sudo apt-get install libpq-dev

Install Cliquet with related dependencies:

pip install cliquet[postgresql]
Full server

PostgreSQL version 9.4 (or higher) is required.

To install PostgreSQL on Ubuntu/Debian use:

sudo apt-get install postgresql-9.4

If your Ubuntu/Debian distribution doesn’t include version 9.4 of PostgreSQL look at the PostgreSQL Ubuntu and PostgreSQL Debian pages. The PostgreSQL project provides an Apt Repository that one can use to install recent PostgreSQL versions.

By default, the postgres user has no password and can hence only connect if ran by the postgres system user. The following command will assign it:

sudo -u postgres psql -c "ALTER USER postgres PASSWORD 'postgres';"

Cliquet requires UTC to be used as the database timezone, and UTF-8 as the database encoding. You can for example use the following commands to create a database named testdb with the appropriate timezone and encoding:

sudo -u postgres psql -c "ALTER ROLE postgres SET TIMEZONE TO 'UTC';"
sudo -u postgres psql -c "CREATE DATABASE testdb ENCODING 'UTF-8';"
Server using Docker

Install docker, for example on Ubuntu:

sudo apt-get install docker.io

Run the official PostgreSQL container locally:

postgres=$(sudo docker run -d -p 5432:5432 postgres)

(optional) Create the test database:

psql -h localhost -U postgres -W
#> CREATE DATABASE "testdb";

Tag and save the current state with:

sudo docker commit $postgres cliquet-empty

In the future, run the tagged version of the container

cliquet=$(sudo docker run -d -p 5432:5432 cliquet-empty)

...

sudo docker stop $cliquet

Configuration

See Pyramid settings documentation.

Environment variables

In order to ease deployment or testing strategies, Cliquet reads settings from environment variables, in addition to .ini files.

For example, cliquet.storage_backend is read from environment variable CLIQUET_STORAGE_BACKEND if defined, else from application .ini, else from internal defaults.

Project info

cliquet.project_name = project
cliquet.project_docs = https://project.rtfd.org/
# cliquet.project_version = 1.3-stable
# cliquet.http_api_version = 1.0

It can be useful to set the project_version to a custom string, in order to prevent disclosing information about the currently running version (when there are known vulnerabilities for example).

Feature settings

# Limit number of batch operations per request
# cliquet.batch_max_requests = 25

# Force pagination *(recommended)*
# cliquet.paginate_by = 200

# Custom record id generator class
# cliquet.id_generator = cliquet.storage.generators.UUID4

Disabling endpoints

It is possible to deactivate specific resources operations, directly in the settings.

To do so, a setting key must be defined for the disabled resources endpoints:

'cliquet.{endpoint_type}_{resource_name}_{method}_enabled'

Where: - endpoint_type is either collection or record; - resource_name is the name of the resource (by default, Cliquet uses the name of the class); - method is the http method (in lower case): For instance put.

For instance, to disable the PUT on records for the Mushrooms resource, the following setting should be declared in the .ini file:

# Disable article collection DELETE endpoint
cliquet.collection_article_delete_enabled = false

# Disable mushroom record PATCH endpoint
cliquet.record_mushroom_patch_enabled = false
Setting the service in readonly mode

It is also possible to deploy a Cliquet service in readonly mode.

Instead of having settings to disable every resource endpoint, the readonly setting can be set:

cliquet.readonly = true

This will disable every resources endpoints that are not accessed with one of GET, OPTIONS, or HEAD methods. Requests will receive a 405 Method not allowed error response.

This setting will also activate readonly heartbeat checks for the permission and the storage backend.

Warning

The cache backend will still needs read-write privileges, in order to cache OAuth authentication states and tokens for example.

If you do not need cache at all, set the kinto.cache_backend setting to an empty string to disable it.

Deployment

# cliquet.backoff = 10
cliquet.retry_after_seconds = 30
Scheme, host and port

By default Cliquet does not enforce requests scheme, host and port. It relies on WSGI specification and the related stack configuration. Tuning this becomes necessary when the application runs behind proxies or load balancers.

Most implementations, like uwsgi, provide configuration variables to adjust it properly.

However if, for some reasons, this had to be enforced at the application level, the following settings can be set:

# cliquet.http_scheme = https
# cliquet.http_host = production.server:7777

Check the url value returned in the hello view.

Deprecation

Activate the service deprecation. If the date specified in eos is in the future, an alert will be sent to clients. If it’s in the past, the service will be declared as decomissionned.

# cliquet.eos = 2015-01-22
# cliquet.eos_message = "Client is too old"
# cliquet.eos_url = http://website/info-shutdown.html
Logging with Heka

Mozilla Services standard logging format can be enabled using:

cliquet.logging_renderer = cliquet.logs.MozillaHekaRenderer

With the following configuration, all logs are redirected to standard output (See 12factor app):

[loggers]
keys = root

[handlers]
keys = console

[formatters]
keys = heka

[logger_root]
level = INFO
handlers = console
formatter = heka

[handler_console]
class = StreamHandler
args = (sys.stdout,)
level = NOTSET

[formatter_heka]
format = %(message)s
Handling exceptions with Sentry

Requires the raven package, or Cliquet installed with pip install cliquet[monitoring].

Sentry logging can be enabled, as explained in official documentation.

Note

The application sends an INFO message on startup, mainly for setup check.

Monitoring with StatsD

Requires the statsd package, or Cliquet installed with pip install cliquet[monitoring].

StatsD metrics can be enabled (disabled by default):

cliquet.statsd_url = udp://localhost:8125
# cliquet.statsd_prefix = cliquet.project_name
Monitoring with New Relic

Requires the newrelic package, or Cliquet installed with pip install cliquet[monitoring].

Enable middlewares as described here.

New-Relic can be enabled (disabled by default):

cliquet.newrelic_config = /location/of/newrelic.ini
cliquet.newrelic_env = prod

Storage

cliquet.storage_backend = cliquet.storage.redis
cliquet.storage_url = redis://localhost:6379/1

# Safety limit while fetching from storage
# cliquet.storage_max_fetch_size = 10000

# Control number of pooled connections
# cliquet.storage_pool_size = 50

See storage backend documentation for more details.

Notifications

To activate event listeners, use the event_handlers setting, which takes a list of either:

  • aliases (e.g. journal)
  • python modules (e.g. cliquet.listeners.redis)

Each listener will load load its dedicated settings.

In the example below, the Redis listener is activated and will send data in the queue Redis list.

cliquet.event_listeners = redis

cliquet.event_listeners.redis.use = cliquet.listeners.redis
cliquet.event_listeners.redis.url = redis://localhost:6379/0
cliquet.event_listeners.redis.pool_size = 5
cliquet.event_listeners.redis.listname = queue
Filtering

It is possible to filter events by action and/or resource name. By default actions create, update and delete are notified for every resources.

cliquet.event_listeners.redis.actions = create
cliquet.event_listeners.redis.resources = article comment

Cache

Backend
cliquet.cache_backend = cliquet.cache.redis
cliquet.cache_url = redis://localhost:6379/0
cliquet.cache_prefix = stack1_

# Control number of pooled connections
# cliquet.storage_pool_size = 50

See cache backend documentation for more details.

Headers

It is possible to add cache control headers on a particular resource for anonymous requests. The client (or proxy) will use them to cache the resource responses for a certain amount of time.

By default, Cliquet indicates the clients to invalidate their cache (Cache-Control: no-cache).

cliquet.mushroom_cache_expires_seconds = 3600

Basically, this will add both Cache-Control: max-age=3600 and Expire: <server datetime + 1H> response headers to the GET responses.

If setting is set to 0, then the resource follows the default behaviour.

CORS

By default, CORS headers are cached by clients during 1H (Access-Control-Max-Age).

The duration can be set from settings. If set to empty or to 0, the header is not sent to clients.

cliquet.cors_max_age_seconds = 7200

Authentication

Since user identification is hashed in storage, a secret key is required in configuration:

# cliquet.userid_hmac_secret = b4c96a8692291d88fe5a97dd91846eb4
Authentication setup

Cliquet relies on pyramid multiauth to initialize authentication.

Therefore, any authentication policy can be specified through configuration.

For example, using the following example, Basic Auth, Persona and IP Auth are enabled:

multiauth.policies = basicauth pyramid_persona ipauth

multiauth.policy.ipauth.use = pyramid_ipauth.IPAuthentictionPolicy
multiauth.policy.ipauth.ipaddrs = 192.168.0.*
multiauth.policy.ipauth.userid = LAN-user
multiauth.policy.ipauth.principals = trusted

Similarly, any authorization policies and group finder function can be specified through configuration in order to deeply customize permissions handling and authorizations.

Basic Auth

basicauth is mentioned among multiauth.policies by default.

multiauth.policies = basicauth

By default, it uses an internal Basic Auth policy bundled with Cliquet.

In order to replace it by another one:

multiauth.policies = basicauth
multiauth.policy.basicauth.use = myproject.authn.BasicAuthPolicy
Custom Authentication

Using the various Pyramid authentication packages, it is possible to plug any kind of authentication.

(Github/Twitter example to be done)

Firefox Accounts

Enabling Firefox Accounts consists in including cliquet_fxa in configuration, mentioning fxa among policies and providing appropriate values for OAuth2 client settings.

See mozilla-services/cliquet-fxa.

Permissions

Backend
cliquet.permission_backend = cliquet.permission.redis
cliquet.permission_url = redis://localhost:6379/1

# Control number of pooled connections
# cliquet.permission_pool_size = 50

See permission backend documentation for more details.

Resources

ACEs are usually set on objects using the permission backend.

It is also possible to configure them from settings, and it will bypass the permission backend.

For example, for a resource named “bucket”, the following setting will enable authenticated people to create bucket records:

cliquet.bucket_create_principals = system.Authenticated

The format of these permission settings is <resource_name>_<permission>_principals = comma,separated,principals.

See shareable resource documentation for more details.

Application profiling

It is possible to profile the application while its running. This is especially useful when trying to find slowness in the application.

Enable middlewares as described here.

Update the configuration file with the following values:

cliquet.profiler_enabled = true
cliquet.profiler_dir = /tmp/profiling

Run a load test (for example):

SERVER_URL=http://localhost:8000 make bench -e

Render execution graphs using GraphViz:

sudo apt-get install graphviz
pip install gprof2dot
gprof2dot -f pstats POST.v1.batch.000176ms.1427458675.prof | dot -Tpng -o output.png

Enable middleware

In order to enable Cliquet middleware, wrap the application in the project main function:

def main(global_config, **settings):
    config = Configurator(settings=settings)
    cliquet.initialize(config, __version__)
    app = config.make_wsgi_app()
    return cliquet.install_middlewares(app, settings)

Initialization sequence

In order to control what part of Cliquet should be run during application startup, or add custom initialization steps from configuration, it is possible to change the initialization_sequence setting.

Warning

This is considered as a dangerous zone and should be used with caution.

Later, a better formalism should be introduced to easily allow addition or removal of steps, without repeating the whole list and without relying on internal functions location.

cliquet.initialization_sequence = cliquet.initialization.setup_request_bound_data
                                  cliquet.initialization.setup_json_serializer
                                  cliquet.initialization.setup_logging
                                  cliquet.initialization.setup_storage
                                  cliquet.initialization.setup_permission
                                  cliquet.initialization.setup_cache
                                  cliquet.initialization.setup_requests_scheme
                                  cliquet.initialization.setup_version_redirection
                                  cliquet.initialization.setup_deprecation
                                  cliquet.initialization.setup_authentication
                                  cliquet.initialization.setup_backoff
                                  cliquet.initialization.setup_statsd
                                  cliquet.initialization.setup_listeners
                                  cliquet.events.setup_transaction_hook

Resource

Cliquet provides a basic component to build resource oriented APIs. In most cases, the main customization consists in defining the schema of the records for this resource.

Full example

import colander

from cliquet import resource
from cliquet import utils


class BookmarkSchema(resource.ResourceSchema):
    url = colander.SchemaNode(colander.String(), validator=colander.url)
    title = colander.SchemaNode(colander.String())
    favorite = colander.SchemaNode(colander.Boolean(), missing=False)
    device = colander.SchemaNode(colander.String(), missing='')

    class Options:
        readonly_fields = ('device',)
        unique_fields = ('url',)


@resource.register()
class Bookmark(resource.UserResource):
    mapping = BookmarkSchema()

    def process_record(self, new, old=None):
        new = super(Bookmark, self).process_record(new, old)
        if new['device'] != old['device']:
            new['device'] = self.request.headers.get('User-Agent')

        return new

See the ReadingList and Kinto projects source code for real use cases.

URLs

By default, a resource defines two URLs:

  • /{classname}s for the list of records
  • /{classname}s/{id} for single records

Since adding an s suffix for the plural form might not always be relevant, URLs can be specified during registration:

@resource.register(collection_path='/user/bookmarks',
                   record_path='/user/bookmarks/{{id}}')
class Bookmark(resource.UserResource):
    mapping = BookmarkSchema()

Note

The same resource can be registered with different URLs.

Schema

Override the base schema to add extra fields using the Colander API.

class Movie(ResourceSchema):
    director = colander.SchemaNode(colander.String())
    year = colander.SchemaNode(colander.Int(),
                               validator=colander.Range(min=1850))
    genre = colander.SchemaNode(colander.String(),
                                validator=colander.OneOf(['Sci-Fi', 'Comedy']))

See the resource schema options to define schema-less resources or specify rules for unicity or readonly.

Permissions

Using the cliquet.resource.UserResource, the resource is accessible by any authenticated request, but the records are isolated by user id.

In order to define resources whose records are not isolated, open publicly or controlled with individual fined-permissions, a cliquet.resource.ShareableResource could be used.

But there are other strategies, please refer to dedicated section about permissions.

HTTP methods and options

In order to specify which HTTP verbs (GET, PUT, PATCH, ...) are allowed on the resource, as well as specific custom Pyramid (or cornice) view arguments, refer to the viewset section.

Events

When a record is created/deleted in a resource, an event is sent. See the dedicated section about notifications to plug events in your Pyramid/Cliquet application or plugin.

Model

Plug custom model

In order to customize the interaction of a HTTP resource with its storage, a custom model can be plugged-in:

from cliquet import resource


class TrackedModel(resource.Model):
    def create_record(self, record, parent_id=None, unique_fields=None):
        record = super(TrackedModel, self).create_record(record,
                                                         parent_id,
                                                         unique_fields)
        trackid = index.track(record)
        record['trackid'] = trackid
        return record


class Payment(resource.UserResource):
    default_model = TrackedModel
Relationships

With the default model and storage backend, Cliquet does not support complex relations.

However, it is possible to plug a custom model class, that will take care of saving and retrieving records with relations.

Note

This part deserves more love, please come and discuss!

In Pyramid views

In Pyramid views, a request object is available and allows to use the storage configured in the application:

from cliquet import resource

def view(request):
    registry = request.registry

    flowers = resource.Model(storage=registry.storage,
                             collection_id='app:flowers')

    flowers.create_record({'name': 'Jonquille', 'size': 30})
    flowers.create_record({'name': 'Amapola', 'size': 18})

    min_size = resource.Filter('size', 20, resource.COMPARISON.MIN)
    records, total = flowers.get_records(filters=[min_size])

    flowers.delete_record(records[0])
Outside views

Outside views, an application context has to be built from scratch.

As an example, let’s build a code that will copy a collection into another:

from cliquet import resource, DEFAULT_SETTINGS
from pyramid import Configurator


config = Configurator(settings=DEFAULT_SETTINGS)
config.add_settings({
    'cliquet.storage_backend': 'cliquet.storage.postgresql'
    'cliquet.storage_url': 'postgres://user:pass@db.server.lan:5432/dbname'
})
cliquet.initialize(config, '0.0.1')

local = resource.Model(storage=config.registry.storage,
                       parent_id='browsing',
                       collection_id='history')

remote = resource.Model(storage=config_remote.registry.storage,
                        parent_id='',
                        collection_id='history')

records, total = in remote.get_records():
for record in records:
    local.create_record(record)

Custom record ids

By default, records ids are UUID4.

A custom record ID generator can be set globally in Configuration, or at the resource level:

from cliquet import resource
from cliquet import utils
from cliquet.storage import generators


class MsecId(generators.Generator):
    def __call__(self):
        return '%s' % utils.msec_time()


@resource.register()
class Mushroom(resource.UserResource):
    def __init__(request):
        super(Mushroom, self).__init__(request)
        self.model.id_generator = MsecId()

Python API

Resource
class cliquet.resource.UserResource(request, context=None)

Base resource class providing every endpoint.

default_viewset

Default cliquet.viewset.ViewSet class to use when the resource is registered.

alias of ViewSet

default_model

Default cliquet.resource.model.Model class to use for interacting the cliquet.storage and cliquet.permission backends.

alias of Model

mapping = <cliquet.resource.schema.ResourceSchema object at 140563396976720 (named )>

Schema to validate records.

timestamp

Return the current collection timestamp.

Return type:int
collection

The collection property.

get_parent_id(request)

Return the parent_id of the resource with regards to the current request.

Parameters:request – The request used to create the resource.
Return type:str
is_known_field(field)

Return True if field is defined in the resource mapping.

Parameters:field (str) – Field name
Return type:bool
collection_get()

Model GET endpoint: retrieve multiple records.

Raises:HTTPNotModified if If-None-Match header is provided and collection not modified in the interim.
Raises:HTTPPreconditionFailed if If-Match header is provided and collection modified in the iterim.
Raises:HTTPBadRequest if filters or sorting are invalid.
collection_post()

Model POST endpoint: create a record.

If the new record conflicts against a unique field constraint, the posted record is ignored, and the existing record is returned, with a 200 status.

Raises:HTTPPreconditionFailed if If-Match header is provided and collection modified in the iterim.

See also

Add custom behaviour by overriding cliquet.resource.UserResource.process_record()

collection_delete()

Model DELETE endpoint: delete multiple records.

Raises:HTTPPreconditionFailed if If-Match header is provided and collection modified in the iterim.
Raises:HTTPBadRequest if filters are invalid.
get()

Record GET endpoint: retrieve a record.

Raises:HTTPNotFound if the record is not found.
Raises:HTTPNotModified if If-None-Match header is provided and record not modified in the interim.
Raises:HTTPPreconditionFailed if If-Match header is provided and record modified in the iterim.
put()

Record PUT endpoint: create or replace the provided record and return it.

Raises:HTTPPreconditionFailed if If-Match header is provided and record modified in the iterim.

Note

If If-None-Match: * request header is provided, the PUT will succeed only if no record exists with this id.

See also

Add custom behaviour by overriding cliquet.resource.UserResource.process_record().

patch()

Record PATCH endpoint: modify a record and return its new version.

If a request header Response-Behavior is set to light, only the fields whose value was changed are returned. If set to diff, only the fields whose value became different than the one provided are returned.

Raises:HTTPNotFound if the record is not found.
Raises:HTTPPreconditionFailed if If-Match header is provided and record modified in the iterim.
delete()

Record DELETE endpoint: delete a record and return it.

Raises:HTTPNotFound if the record is not found.
Raises:HTTPPreconditionFailed if If-Match header is provided and record modified in the iterim.
process_record(new, old=None)

Hook for processing records before they reach storage, to introduce specific logics on fields for example.

def process_record(self, new, old=None):
    new = super(MyResource, self).process_record(new, old)
    version = old['version'] if old else 0
    new['version'] = version + 1
    return new

Or add extra validation based on request:

from cliquet.errors import raise_invalid

def process_record(self, new, old=None):
    new = super(MyResource, self).process_record(new, old)
    if new['browser'] not in request.headers['User-Agent']:
        raise_invalid(self.request, name='browser', error='Wrong')
    return new
Parameters:
  • new (dict) – the validated record to be created or updated.
  • old (dict) – the old record to be updated, None for creation endpoints.
Returns:

the processed record.

Return type:

dict

apply_changes(record, changes)

Merge changes into record fields.

Note

This is used in the context of PATCH only.

Override this to control field changes at record level, for example:

def apply_changes(self, record, changes):
    # Ignore value change if inferior
    if record['position'] > changes.get('position', -1):
        changes.pop('position', None)
    return super(MyResource, self).apply_changes(record, changes)
Raises:HTTPBadRequest if result does not comply with resource schema.
Returns:the new record with changes applied.
Return type:dict
Schema
class cliquet.resource.schema.ResourceSchema(*arg, **kw)

Base resource schema, with Cliquet specific built-in options.

class Options

Resource schema options.

This is meant to be overriden for changing values:

class Product(ResourceSchema):
    reference = colander.SchemaNode(colander.String())

    class Options:
        unique_fields = ('reference',)
unique_fields = ()

Fields that must have unique values for the user collection. During records creation and modification, a conflict error will be raised if unicity is about to be violated.

readonly_fields = ()

Fields that cannot be updated. Values for fields will have to be provided either during record creation, through default values using missing attribute or implementing a custom logic in cliquet.resource.UserResource.process_record().

preserve_unknown = False

Define if unknown fields should be preserved or not.

For example, in order to define a schema-less resource, in other words a resource that will accept any form of record, the following schema definition is enough:

class SchemaLess(ResourceSchema):
    class Options:
        preserve_unknown = True
ResourceSchema.is_readonly(field)

Return True if specified field name is read-only.

Parameters:field (str) – the field name in the schema
Returns:True if the specified field is read-only, False otherwise.
Return type:bool
class cliquet.resource.schema.PermissionsSchema(*args, **kwargs)

A permission mapping defines ACEs.

It has permission names as keys and principals as values.

{
    "write": ["fxa:af3e077eb9f5444a949ad65aa86e82ff"],
    "groups:create": ["fxa:70a9335eecfe440fa445ba752a750f3d"]
}
class cliquet.resource.schema.TimeStamp(*arg, **kw)

Basic integer schema field that can be set to current server timestamp in milliseconds if no value is provided.

class Book(ResourceSchema):
    added_on = TimeStamp()
    read_on = TimeStamp(auto_now=False, missing=-1)
schema_type

alias of Integer

title = 'Epoch timestamp'

Default field title.

auto_now = True

Set to current server timestamp (milliseconds) if not provided.

missing = None

Default field value if not provided in record.

class cliquet.resource.schema.URL(*arg, **kw)

String field representing a URL, with max length of 2048. This is basically a shortcut for string field with ~colander:colander.url.

class BookmarkSchema(ResourceSchema):
    url = URL()
schema_type

alias of String

Model
class cliquet.resource.Model(storage, id_generator=None, collection_id='', parent_id='', auth=None)

A collection stores and manipulate records in its attached storage.

It is not aware of HTTP environment nor protocol.

Records are isolated according to the provided name and parent_id.

Those notions have no particular semantic and can represent anything. For example, the collection name can be the type of objects stored, and parent_id can be the current user id or a group where the collection belongs. If left empty, the collection records are not isolated.

id_field = 'id'

Name of id field in records

modified_field = 'last_modified'

Name of last modified field in records

deleted_field = 'deleted'

Name of deleted field in deleted records

timestamp(parent_id=None)

Fetch the collection current timestamp.

Parameters:parent_id (str) – optional filter for parent id
Return type:integer
get_records(filters=None, sorting=None, pagination_rules=None, limit=None, include_deleted=False, parent_id=None)

Fetch the collection records.

Override to post-process records after feching them from storage.

Parameters:
  • filters (list of cliquet.storage.Filter) – Optionally filter the records by their attribute. Each filter in this list is a tuple of a field, a value and a comparison (see cliquet.utils.COMPARISON). All filters are combined using AND.
  • sorting (list of cliquet.storage.Sort) – Optionnally sort the records by attribute. Each sort instruction in this list refers to a field and a direction (negative means descending). All sort instructions are cumulative.
  • pagination_rules (list of list of cliquet.storage.Filter) – Optionnally paginate the list of records. This list of rules aims to reduce the set of records to the current page. A rule is a list of filters (see filters parameter), and all rules are combined using OR.
  • limit (int) – Optionnally limit the number of records to be retrieved.
  • include_deleted (bool) – Optionnally include the deleted records that match the filters.
  • parent_id (str) – optional filter for parent id
Returns:

A tuple with the list of records in the current page, the total number of records in the result set.

Return type:

tuple

delete_records(filters=None, parent_id=None)

Delete multiple collection records.

Override to post-process records after their deletion from storage.

Parameters:
  • filters (list of cliquet.storage.Filter) – Optionally filter the records by their attribute. Each filter in this list is a tuple of a field, a value and a comparison (see cliquet.utils.COMPARISON). All filters are combined using AND.
  • parent_id (str) – optional filter for parent id
Returns:

The list of deleted records from storage.

get_record(record_id, parent_id=None)

Fetch current view related record, and raise 404 if missing.

Parameters:
  • record_id (str) – record identifier
  • parent_id (str) – optional filter for parent id
Returns:

the record from storage

Return type:

dict

create_record(record, parent_id=None, unique_fields=None)

Create a record in the collection.

Override to perform actions or post-process records after their creation in storage.

def create_record(self, record):
    record = super(MyModel, self).create_record(record)
    idx = index.store(record)
    record['index'] = idx
    return record
Parameters:
  • record (dict) – record to store
  • parent_id (str) – optional filter for parent id
  • unique_fields (tuple) – list of fields that should remain unique
Returns:

the newly created record.

Return type:

dict

update_record(record, parent_id=None, unique_fields=None)

Update a record in the collection.

Override to perform actions or post-process records after their modification in storage.

def update_record(self, record, parent_id=None,unique_fields=None):
    record = super(MyModel, self).update_record(record,
                                                parent_id,
                                                unique_fields)
    subject = 'Record {} was changed'.format(record[self.id_field])
    send_email(subject)
    return record
Parameters:
  • record (dict) – record to store
  • parent_id (str) – optional filter for parent id
  • unique_fields (tuple) – list of fields that should remain unique
Returns:

the updated record.

Return type:

dict

delete_record(record, parent_id=None, last_modified=None)

Delete a record in the collection.

Override to perform actions or post-process records after deletion from storage for example:

def delete_record(self, record):
    deleted = super(MyModel, self).delete_record(record)
    erase_media(record)
    deleted['media'] = 0
    return deleted
Parameters:
  • record (dict) – the record to delete
  • record – record to store
  • parent_id (str) – optional filter for parent id
Returns:

the deleted record.

Return type:

dict

Generators
class cliquet.storage.generators.Generator(config=None)

Base generator for records ids.

Id generators are used by storage backend during record creation, and at resource level to validate record id in requests paths.

regexp = '^[a-zA-Z0-9][a-zA-Z0-9_-]*$'

Default record id pattern. Can be changed to comply with custom ids.

match(record_id)

Validate that record ids match the generator. This is used mainly when a record id is picked arbitrarily (e.g with PUT requests).

Returns:True if the specified record id matches expected format.
Return type:bool
class cliquet.storage.generators.UUID4(config=None)

UUID4 record id generator.

UUID block are separated with -. (example: '472be9ec-26fe-461b-8282-9c4e4b207ab3')

UUIDs are very safe in term of unicity. If 1 billion of UUIDs are generated every second for the next 100 years, the probability of creating just one duplicate would be about 50% (source).

regexp = '^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$'

UUID4 accurate pattern.

Viewsets

Cliquet maps URLs, endpoints and permissions to resources using ViewSets.

Since a resource defines two URLs with several HTTP methods, a view set can be considered as a set of rules for registring the resource views into the routing mechanism of Pyramid.

To use Cliquet in a basic fashion, there is no need to understand how viewsets work in full detail.

Override defaults

Viewsets defaults can be overriden by passing arguments to the cliquet.resource.register() class decorator:

from cliquet import resource


@resource.register(collection_methods=('GET',))
class Resource(resource.UserResource):
    mapping = BookmarkSchema()

Subclassing

In case this isn’t enough, the cliquet.resource.viewset.ViewSet class can be subclassed and specified during registration:

from cliquet import resource


class NoSchemaViewSet(resource.ViewSet):

    def get_record_schema(self, resource_cls, method):
        simple_mapping = colander.MappingSchema(unknown='preserve')
        return simple_mapping


@resource.register(viewset=NoSchemaViewSet())
class Resource(resource.UserResource):
    mapping = BookmarkSchema()

ViewSet class

class cliquet.resource.ViewSet(**kwargs)

The default ViewSet object.

A viewset contains all the information needed to register any resource in the Cornice registry.

It provides the same features as cornice.resource(), except that it is much more flexible and extensible.

update(**kwargs)

Update viewset attributes with provided values.

get_view_arguments(endpoint_type, resource_cls, method)

Return the Pyramid/Cornice view arguments for the given endpoint type and method.

Parameters:
  • endpoint_type (str) – either “collection” or “record”.
  • resource_cls – the resource class.
  • method (str) – the HTTP method.
get_record_schema(resource_cls, method)

Return the Cornice schema for the given method.

get_view(endpoint_type, method)

Return the view method name located on the resource object, for the given type and method.

  • For collections, this will be “collection_{method|lower}
  • For records, this will be “{method|lower}.
get_name(resource_cls)

Returns the name of the resource.

get_service_name(endpoint_type, resource_cls)

Returns the name of the service, depending a given type and resource.

is_endpoint_enabled(endpoint_type, resource_name, method, settings)

Returns if the given endpoint is enabled or not.

Uses the settings to tell so.

Storage

Backends

PostgreSQL
class cliquet.storage.postgresql.Storage(client, max_fetch_size, *args, **kwargs)

Storage backend using PostgreSQL.

Recommended in production (requires PostgreSQL 9.4 or higher).

Enable in configuration:

cliquet.storage_backend = cliquet.storage.postgresql

Database location URI can be customized:

cliquet.storage_url = postgres://user:pass@db.server.lan:5432/dbname

Alternatively, username and password could also rely on system user ident or even specified in ~/.pgpass (see PostgreSQL documentation).

Note

Some tables and indices are created when cliquet migrate is run. This requires some privileges on the database, or some error will be raised.

Alternatively, the schema can be initialized outside the python application, using the SQL file located in cliquet/storage/postgresql/schema.sql. This allows to distinguish schema manipulation privileges from schema usage.

A connection pool is enabled by default:

cliquet.storage_pool_size = 10
cliquet.storage_maxoverflow = 10
cliquet.storage_max_backlog = -1
cliquet.storage_pool_recycle = -1
cliquet.storage_pool_timeout = 30
cliquet.cache_poolclass = cliquet.storage.postgresql.pool.QueuePoolWithMaxBacklog

The max_backlog limits the number of threads that can be in the queue waiting for a connection. Once this limit has been reached, any further attempts to acquire a connection will be rejected immediately, instead of locking up all threads by keeping them waiting in the queue.

See dedicated section in SQLAlchemy documentation for default values and behaviour.

Note

Using a dedicated connection pool is still recommended to allow load balancing, replication or limit the number of connections used in a multi-process deployment.

Redis
class cliquet.storage.redis.Storage(client, *args, **kwargs)

Storage backend implementation using Redis.

Warning

Useful for very low server load, but won’t scale since records sorting and filtering are performed in memory.

Enable in configuration:

cliquet.storage_backend = cliquet.storage.redis

(Optional) Instance location URI can be customized:

cliquet.storage_url = redis://localhost:6379/0

A threaded connection pool is enabled by default:

cliquet.storage_pool_size = 50
Memory
class cliquet.storage.memory.Storage(*args, **kwargs)

Storage backend implementation in memory.

Useful for development or testing purposes, but records are lost after each server restart.

Enable in configuration:

cliquet.storage_backend = cliquet.storage.memory

API

Implementing a custom storage backend consists in implementating the following interface:

class cliquet.storage.Filter(field, value, operator)

Filtering properties.

field

Alias for field number 0

operator

Alias for field number 2

value

Alias for field number 1

class cliquet.storage.Sort(field, direction)

Sorting properties.

direction

Alias for field number 1

field

Alias for field number 0

class cliquet.storage.StorageBase

Storage abstraction used by resource views.

It is meant to be instantiated at application startup. Any operation may raise a HTTPServiceUnavailable error if an error occurs with the underlying service.

Configuration can be changed to choose which storage backend will persist the objects.

Raises:HTTPServiceUnavailable
initialize_schema()

Create every necessary objects (like tables or indices) in the backend.

This is excuted when the cliquet migrate command is ran.

flush(auth=None)

Remove every object from this storage.

collection_timestamp(collection_id, parent_id, auth=None)

Get the highest timestamp of every objects in this collection_id for this parent_id.

Note

This should take deleted objects into account.

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
Returns:

the latest timestamp of the collection.

Return type:

int

create(collection_id, parent_id, object, id_generator=None, unique_fields=None, id_field='id', modified_field='last_modified', auth=None)

Create the specified object in this collection_id for this parent_id. Assign the id to the object, using the attribute cliquet.resource.Model.id_field.

Note

This will update the collection timestamp.

Raises:

cliquet.storage.exceptions.UnicityError

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • object (dict) – the object to create.
Returns:

the newly created object.

Return type:

dict

get(collection_id, parent_id, object_id, id_field='id', modified_field='last_modified', auth=None)

Retrieve the object with specified object_id, or raise error if not found.

Raises:

cliquet.storage.exceptions.RecordNotFoundError

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • object_id (str) – unique identifier of the object
Returns:

the object object.

Return type:

dict

update(collection_id, parent_id, object_id, object, unique_fields=None, id_field='id', modified_field='last_modified', auth=None)

Overwrite the object with the specified object_id.

If the specified id is not found, the object is created with the specified id.

Note

This will update the collection timestamp.

Raises:

cliquet.storage.exceptions.UnicityError

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • object_id (str) – unique identifier of the object
  • object (dict) – the object to update or create.
Returns:

the updated object.

Return type:

dict

delete(collection_id, parent_id, object_id, with_deleted=True, id_field='id', modified_field='last_modified', deleted_field='deleted', auth=None)

Delete the object with specified object_id, and raise error if not found.

Deleted objects must be removed from the database, but their ids and timestamps of deletion must be tracked for synchronization purposes. (See cliquet.storage.StorageBase.get_all())

Note

This will update the collection timestamp.

Raises:

cliquet.storage.exceptions.RecordNotFoundError

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • object_id (str) – unique identifier of the object
  • with_deleted (bool) – track deleted record with a tombstone
Returns:

the deleted object, with minimal set of attributes.

Return type:

dict

delete_all(collection_id, parent_id, filters=None, with_deleted=True, id_field='id', modified_field='last_modified', deleted_field='deleted', auth=None)

Delete all objects in this collection_id for this parent_id.

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • filters (list of cliquet.storage.Filter) – Optionnally filter the objects to delete.
  • with_deleted (bool) – track deleted records with a tombstone
Returns:

the list of deleted objects, with minimal set of attributes.

Return type:

list of dict

purge_deleted(collection_id, parent_id, before=None, id_field='id', modified_field='last_modified', auth=None)

Delete all deleted object tombstones in this collection_id for this parent_id.

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • before (int) – Optionnal timestamp to limit deletion (exclusive)
Returns:

The number of deleted objects.

Return type:

int

get_all(collection_id, parent_id, filters=None, sorting=None, pagination_rules=None, limit=None, include_deleted=False, id_field='id', modified_field='last_modified', deleted_field='deleted', auth=None)

Retrieve all objects in this collection_id for this parent_id.

Parameters:
  • collection_id (str) – the collection id.
  • parent_id (str) – the collection parent.
  • filters (list of cliquet.storage.Filter) – Optionally filter the objects by their attribute. Each filter in this list is a tuple of a field, a value and a comparison (see cliquet.utils.COMPARISON). All filters are combined using AND.
  • sorting (list of cliquet.storage.Sort) – Optionnally sort the objects by attribute. Each sort instruction in this list refers to a field and a direction (negative means descending). All sort instructions are cumulative.
  • pagination_rules (list of list of cliquet.storage.Filter) – Optionnally paginate the list of objects. This list of rules aims to reduce the set of objects to the current page. A rule is a list of filters (see filters parameter), and all rules are combined using OR.
  • limit (int) – Optionnally limit the number of objects to be retrieved.
  • include_deleted (bool) – Optionnally include the deleted objects that match the filters.
Returns:

the limited list of objects, and the total number of matching objects in the collection (deleted ones excluded).

Return type:

tuple (list, integer)

Exceptions

Exceptions raised by storage backend.

exception cliquet.storage.exceptions.BackendError(original=None, message=None, *args, **kwargs)

A generic exception raised by storage on error.

Parameters:original (Exception) – the wrapped exception raised by underlying library.
exception cliquet.storage.exceptions.RecordNotFoundError

An exception raised when a specific record could not be found.

exception cliquet.storage.exceptions.UnicityError(field, record, *args, **kwargs)

An exception raised on unicity constraint violation.

Raised by storage backend when the creation or the modification of a record violates the unicity constraints defined by the resource.

Store custom data

Storage can be used to store arbitrary data.

data = {'subscribed': datetime.now()}
user_id = request.authenticated_userid

storage = request.registry.storage
storage.create(collection_id='__custom', parent_id='', record=data)

See the Model class to manipulate collections of records.

Cache

PostgreSQL

class cliquet.cache.postgresql.Cache(client, *args, **kwargs)

Cache backend using PostgreSQL.

Enable in configuration:

cliquet.cache_backend = cliquet.cache.postgresql

Database location URI can be customized:

cliquet.cache_url = postgres://user:pass@db.server.lan:5432/dbname

Alternatively, username and password could also rely on system user ident or even specified in ~/.pgpass (see PostgreSQL documentation).

Note

Some tables and indices are created when cliquet migrate is run. This requires some privileges on the database, or some error will be raised.

Alternatively, the schema can be initialized outside the python application, using the SQL file located in cliquet/cache/postgresql/schema.sql. This allows to distinguish schema manipulation privileges from schema usage.

A connection pool is enabled by default:

cliquet.cache_pool_size = 10
cliquet.cache_maxoverflow = 10
cliquet.cache_max_backlog = -1
cliquet.cache_pool_recycle = -1
cliquet.cache_pool_timeout = 30
cliquet.cache_poolclass = cliquet.storage.postgresql.pool.QueuePoolWithMaxBacklog

The max_backlog limits the number of threads that can be in the queue waiting for a connection. Once this limit has been reached, any further attempts to acquire a connection will be rejected immediately, instead of locking up all threads by keeping them waiting in the queue.

See dedicated section in SQLAlchemy documentation for default values and behaviour.

Note

Using a dedicated connection pool is still recommended to allow load balancing, replication or limit the number of connections used in a multi-process deployment.

Noindex:

Redis

class cliquet.cache.redis.Cache(client, *args, **kwargs)

Cache backend implementation using Redis.

Enable in configuration:

cliquet.cache_backend = cliquet.cache.redis

(Optional) Instance location URI can be customized:

cliquet.cache_url = redis://localhost:6379/1

A threaded connection pool is enabled by default:

cliquet.cache_pool_size = 50

If the database is used for multiple Kinto deployement cache, you may want to add a prefix to every key to avoid collision:

cliquet.cache_prefix = stack1_
Noindex:

Memory

class cliquet.cache.memory.Cache(*args, **kwargs)

Cache backend implementation in local thread memory.

Enable in configuration:

cliquet.cache_backend = cliquet.cache.memory
Noindex:

API

Implementing a custom cache backend consists on implementing the following interface:

class cliquet.cache.CacheBase(*args, **kwargs)
initialize_schema()

Create every necessary objects (like tables or indices) in the backend.

This is excuted when the cliquet migrate command is ran.

flush()

Delete every values.

ttl(key)

Obtain the expiration value of the specified key.

Parameters:key (str) – key
Returns:number of seconds or negative if no TTL.
Return type:float
expire(key, ttl)

Set the expiration value ttl for the specified key.

Parameters:
  • key (str) – key
  • ttl (float) – number of seconds
set(key, value, ttl=None)

Store a value with the specified key. If ttl is provided, set an expiration value.

Parameters:
  • key (str) – key
  • value (str) – value to store
  • ttl (float) – expire after number of seconds
get(key)

Obtain the value of the specified key.

Parameters:key (str) – key
Returns:the stored value or None if missing.
Return type:str
delete(key)

Delete the value of the specified key.

Parameters:key (str) – key

Notifications

Knowing some records have been modified in a resource is very useful to integrate a Cliquet-based application with other services.

For example, a search service that gets notified everytime something has changed, can continuously update its index.

Cliquet leverages Pyramid’s built-in event system and produces the following events:

  • cliquet.events.ResourceRead: a read operation occured on the resource.

  • cliquet.events.ResourceChanged: a resource is being changed. This event occurs synchronously within the transaction and within the request/response cycle. Commit is not yet done and rollback is still possible.

    Subscribers of this event are likely to perform database operations, alter the server response, or cancel the transaction (by raising an HTTP exception for example). Do not subscribe to this event for operations that will not be rolled-back automatically.

  • cliquet.events.AfterResourceChanged: a resource was changed and committed.

    Subscribers of this event can fail, errors are swallowed and logged. The final transaction result (or response) cannot be altered.

    Subscribers of this event are likely to perform irreversible actions that requires data to be committed in database (like sending messages, deleting files on disk, or run asynchronous tasks).

Event subscribers can then pick up those events and act upon them.

from cliquet.events import AfterResourceChanged


def on_resource_changed(event):
    for change in event.impacted_records:
        start_download(change['new']['url'])

config.add_subscriber(on_resource_changed, AfterResourceChanged)

Transactions

Only one event is sent per transaction, per resource and per action.

In other words, if every requests of a batch requests perform the same action on the same resource, only one event will be sent.

The AfterResourceChanged is sent only if the transaction was comitted successfully.

It is possible to cancel the current transaction by raising an HTTP Exception from a ResourceChanged event. For example:

from cliquet.events import ResourceChanged
from pyramid import httpexceptions

def check_quota(event):
     max_quota = event.request.registry.settings['max_quota']
     if check_quota(event, max_quota):
        raise httpexceptions.HTTPInsufficientStorage()

config.add_subscriber(check_quota, ResourceChanged)

Filtering

It is possible to filter events based on its action or the name of the resource where it occured.

For example:

from cliquet.events import ResourceChanged, ACTIONS

config.add_subscriber(on_mushroom_changed, ResourceChanged, for_resources=('mushroom',))
config.add_subscriber(on_record_deleted, ResourceChanged, for_actions=(ACTIONS.DELETE,))

Payload

The cliquet.events.ResourceChanged and cliquet.events.AfterResourceChanged events contain a payload attribute with the following information:

  • timestamp: the time of the event
  • action: what happened. ‘create’, ‘update’ or ‘delete’
  • uri: the uri of the impacted resource
  • user_id: the authenticated user id
  • resource_name: the name of the impacted resouce (e.g. ‘article’, ‘bookmark’, bucket’, ‘group’ etc.)
  • <resource_name>_id: id of the impacted record
  • <matchdict value>: every value matched by each URL pattern name (see Pyramid request matchdict)

And provides the list of affected records in the impacted_records attribute. This list contains dictionnaries with new and old keys. For creation events, only new is provided. For deletion events, only old is provided. This also allows listeners to react on particular field change or handle diff between versions.

Example, when deleting a collection with two records:

>>> event.impacted_records
[{'old': {'deleted': True, 'last_modified': 1447240896769, 'id': u'a1f4af60-ddf5-4c49-933f-4cfeff18ad07'}},
 {'old': {'deleted': True, 'last_modified': 1447240896770, 'id': u'7a6916aa-0ea1-42a7-9741-c24fe13cb70b'}}]

Event listeners

It is possible for an application or a plugin to listen to events and execute some code. Triggered code on events is synchronously called when a request is handled.

Cliquet offers custom listeners that can be activated through configuration, so that every Cliquet-based application can benefit from pluggable listeners without using config.add_event_subscriber() explicitely.

Currently, a simple built-in listener is available, that just delivers the events into a Redis queue, allowing asynchronous event handling:

class cliquet.listeners.redis.Listener(client, listname, *args, **kwargs)

A Redis-based event listener that simply pushes the events payloads into the specified Redis list as they happen.

This listener allows actions to be performed asynchronously, using Redis Pub/Sub notifications, or scheduled inspections of the queue.

To activate it, look at the dedicated configuration.

Implementing a custom listener consists on implementing the following interface:

class cliquet.listeners.ListenerBase(*args, **kwargs)
__call__(event)
Parameters:event – Incoming event

Permissions

Cliquet provides a mechanism to handle authorization on the stored objects.

This section gives details about the behaviour of resources in regards to permissions.

User resource

This is the simplest one, as presented in the resource section.

When using a cliquet.resource.UserResource, every authenticated user can manipulate and read their own records. There is no way to restrict this or allow sharing of records.

Method URL permission
GET / HEAD /{collection} Authenticated
POST /{collection} Authenticated
DELETE /{collection} Authenticated
GET / HEAD /{collection}/{id} Authenticated
PUT /{collection}/{id} Authenticated
PATCH /{collection}/{id} Authenticated
DELETE /{collection}/{id} Authenticated

Note

When using only these resource, the permission backend remains unused. Its configuration is not necessary.

Public BasicAuth

If Basic Auth authentication is enabled, private user resources can become semi-private or public if the user:pass is publicly known and shared (for example public: is a valid user:pass combination). That’s how most simple demos of Kinto — a Cliquet-based application — are built by the way!

Shareable resource

Warning

When using this kind of resource, the permission_backend setting must be set, as described in the configuration section.

To introduce more flexibility, the cliquet.resource.ShareableResource can be used instead.

from cliquet import resource

@resource.register()
class Toadstool(resource.ShareableResource):
    mapping = MushroomSchema()

With this alternative resource class, Cliquet will register the endpoints with a specific route factory, that will take care of checking the appropriate permission for each action.

Method URL permission Comments
GET / HEAD /{collection} read If not allowed by setting cliquet.{collection}_read_principals, will return list of records where user has read permission.
POST /{collection} create Allowed by setting cliquet.{collection}_create_principals
DELETE /{collection} write If not allowed by setting cliquet.{collection}_write_principals, will delete the list of records where user has write permission.
GET / HEAD /{collection}/{id} read If not allowed by setting cliquet.{collection}_read_principals, will check record permissions
PUT /{collection}/{id} create if record doesn’t exist, write otherwise Allowed by setting cliquet.{collection}_create_principals, or cliquet.{collection}_create_principals or existing record permissions
PATCH /{collection}/{id} write If not allowed by setting cliquet.{collection}_write_principals, will check record permissions
DELETE /{collection}/{id} write If not allowed by setting cliquet.{collection}_write_principals, will check record permissions

The record permissions can be manipulated via the permissions attribute in the JSON payload, aside the data attribute. It allows to specify the list of principals allowed for each permission, as detailed in the API section.

Important

When defining permissions, there are two specific principals:

  • system.Authenticated: any authenticated user
  • system.Everyone: any user

The write permission is required to be able to modify the permissions of an existing record.

When a record is created or modified, the current user is added to list of principals for the write permission on this object. That means that a user is always able to replace or delete the records she created.

Note

Don’t hesitate to submit a contribution to introduce a way to control the current behaviour instead of always granting write on current user!

BasicAuth trickery

Like for user resources, if Basic Auth authentication is enabled, the predictable user id can be used to define semi-private or public if the user:pass is known and shared (for example public: is a valid user:pass combination).

For example, get the user id obtained in the hello root view with a user:pass combination and use it in the permissions JSON payloads, or settings:

cliquet.{collection}_read_principals = basicauth:631c2d625ee5726172cf67c6750de10a3e1a04bcd603bc9ad6d6b196fa8257a6
Manipulate permissions

One way of achieving dynamic permissions is to manipulate the permission backend manually.

For example, in some imaginary admin view:

def admin_view(request):
    # Custom Pyramid view.
    permission = request.registry.permission

    # Give `create` permission to `user_id` in POST
    some_user_id = request.POST['user_id']
    permission_object_id = '/articles'
    permission = 'create'
    permission.add_principal_to_ace(permission_object_id,
                                    permission,
                                    some_user_id)

Or during application init (or scripts):

def main(global_config, **settings):
    # ...
    cliquet.initialize(config, __version__)
    # ...

    some_user_id = 'basicauth:ut082jghnrgnjnj'
    permission_object_id = '/articles'
    permission = 'create'
    config.registry.permission.add_principal_to_ace(permission_object_id,
                                                    permission,
                                                    some_user_id)

Since principals can be anything, it is also possible to use them to define groups:

def add_to_admins(request):
    # Custom Pyramid view.
    permission = request.registry.permission

    some_user_id = request.POST['user_id']
    group_name = 'group:admins'
    permission.add_user_principal(some_user_id, group_name)

And then refer as group:admins in the list of allowed principals.

Custom permission checking

The permissions verification in Cliquet is done with usual Pyramid authorization abstractions. Most notably using an implementation of a RootFactory in conjonction with an Authorization policy.

In order to completely override (or mimic) the defaults, a custom RootFactory and a custom Authorization policy can be plugged on the resource during registration.

from cliquet import resource

class MyViewSet(resource.ViewSet):

    def get_view_arguments(self, endpoint_type, resource_cls, method):
        args = super(MyViewSet, self).get_view_arguments(endpoint_type,
                                                         resource_cls,
                                                         method)
        if method.lower() not in ('get', 'head'):
            args['permission'] = 'publish'
        return args

    def get_service_arguments(self):
        args = super(MyViewSet, self).get_service_arguments()
        args['factory'] = myapp.MyRootFactory
        return args


@resource.register(viewset=MyViewSet())
class Resource(resource.UserResource):
    mapping = BookmarkSchema()

See more details about available customization in the viewset section.

A custom RootFactory and AuthorizationPolicy should implement the permission checking using Pyramid mecanisms.

For example, a simplistic example with the previous resource viewset:

from pyramid.security import IAuthorizationPolicy

class MyRootFactory(object):
    def __init__(self, request):
        self.current_resource = None
        service = request.current_service
        if service and hasattr(service, 'resource'):
            self.current_resource = service.resource


@implementer(IAuthorizationPolicy)
class AuthorizationPolicy(object):
    def permits(self, context, principals, permission):
        if context.current_resource == BlogArticle:
            if permission == 'publish':
                return ('group:publishers' in principals)
        return False

Backends

The ACLs are stored in a permission backend. Like for Storage and Cache, it is pluggable from configuration.

PostgreSQL
class cliquet.permission.postgresql.Permission(client, *args, **kwargs)

Permission backend using PostgreSQL.

Enable in configuration:

cliquet.permission_backend = cliquet.permission.postgresql

Database location URI can be customized:

cliquet.permission_url = postgres://user:pass@db.server.lan:5432/dbname

Alternatively, username and password could also rely on system user ident or even specified in ~/.pgpass (see PostgreSQL documentation).

Note

Some tables and indices are created when cliquet migrate is run. This requires some privileges on the database, or some error will be raised.

Alternatively, the schema can be initialized outside the python application, using the SQL file located in cliquet/permission/postgresql/schema.sql. This allows to distinguish schema manipulation privileges from schema usage.

A connection pool is enabled by default:

cliquet.permission_pool_size = 10
cliquet.permission_maxoverflow = 10
cliquet.permission_max_backlog = -1
cliquet.permission_pool_recycle = -1
cliquet.permission_pool_timeout = 30
cliquet.cache_poolclass = cliquet.storage.postgresql.pool.QueuePoolWithMaxBacklog

The max_backlog limits the number of threads that can be in the queue waiting for a connection. Once this limit has been reached, any further attempts to acquire a connection will be rejected immediately, instead of locking up all threads by keeping them waiting in the queue.

See dedicated section in SQLAlchemy documentation for default values and behaviour.

Note

Using a dedicated connection pool is still recommended to allow load balancing, replication or limit the number of connections used in a multi-process deployment.

Noindex:
Redis
class cliquet.permission.redis.Permission(client, *args, **kwargs)

Permission backend implementation using Redis.

Enable in configuration:

cliquet.permission_backend = cliquet.permission.redis

(Optional) Instance location URI can be customized:

cliquet.permission_url = redis://localhost:6379/2

A threaded connection pool is enabled by default:

cliquet.permission_pool_size = 50
Noindex:
Memory
class cliquet.permission.memory.Permission(*args, **kwargs)

Permission backend implementation in local thread memory.

Enable in configuration:

cliquet.permission_backend = cliquet.permission.memory
Noindex:

API

Implementing a custom permission backend consists in implementating the following interface:

class cliquet.permission.PermissionBase(*args, **kwargs)
initialize_schema()

Create every necessary objects (like tables or indices) in the backend.

This is excuted with the cliquet migrate command.

flush()

Delete all data stored in the permission backend.

add_user_principal(user_id, principal)

Add an additional principal to a user.

Parameters:
  • user_id (str) – The user_id to add the principal to.
  • principal (str) – The principal to add.
remove_user_principal(user_id, principal)

Remove an additional principal from a user.

Parameters:
  • user_id (str) – The user_id to remove the principal to.
  • principal (str) – The principal to remove.
remove_principal(principal)

Remove a principal from every user.

Parameters:principal (str) – The principal to remove.
user_principals(user_id)

Return the set of additionnal principals given to a user.

Parameters:user_id (str) – The user_id to get the list of groups for.
Returns:The list of group principals the user is in.
Return type:set
add_principal_to_ace(object_id, permission, principal)

Add a principal to an Access Control Entry.

Parameters:
  • object_id (str) – The object to add the permission principal to.
  • permission (str) – The permission to add the principal to.
  • principal (str) – The principal to add to the ACE.
remove_principal_from_ace(object_id, permission, principal)

Remove a principal to an Access Control Entry.

Parameters:
  • object_id (str) – The object to remove the permission principal to.
  • permission (str) – The permission that should be removed.
  • principal (str) – The principal to remove to the ACE.
object_permission_principals(object_id, permission)

Return the set of principals of a bound permission (unbound permission + object id).

Parameters:
  • object_id (str) – The object_id the permission is set to.
  • permission (str) – The permission to query.
Returns:

The list of user principals

Return type:

set

principals_accessible_objects(principals, permission, object_id_match=None, get_bound_permissions=None)

Return the list of objects id where the specified principals have the specified permission.

Parameters:
  • principal (list) – List of user principals
  • permission (str) – The permission to query.
  • object_id_match (str) – Filter object ids based on a pattern (e.g. '*articles*').
  • get_bound_permissions (function) – The methods to call in order to generate the list of permission to verify against. (ie: if you can write, you can read)
Returns:

The list of object ids

Return type:

set

object_permission_authorized_principals(object_id, permission, get_bound_permissions=None)

Return the full set of authorized principals for a given permission + object (bound permission).

Parameters:
  • object_id (str) – The object_id the permission is set to.
  • permission (str) – The permission to query.
  • get_bound_permissions (function) – The methods to call in order to generate the list of permission to verify against. (ie: if you can write, you can read)
Returns:

The list of user principals

Return type:

set

check_permission(object_id, permission, principals, get_bound_permissions=None)

Test if a principal set have got a permission on an object.

Parameters:
  • object_id (str) – The identifier of the object concerned by the permission.
  • permission (str) – The permission to test.
  • principals (set) – A set of user principals to test the permission against.
  • get_bound_permissions (function) – The method to call in order to generate the set of permission to verify against. (ie: if you can write, you can read)
object_permissions(object_id, permissions=None)

Return the set of principals for each object permission.

Parameters:
  • object_id (str) – The object_id the permission is set to.
  • permissions (list) – List of permissions to retrieve. If not define will try to find them all.
Returns:

The dictionnary with the list of user principals for each object permissions

Return type:

dict

replace_object_permissions(object_id, permissions)

Replace given object permissions.

Parameters:
  • object_id (str) – The object to replace permissions to.
  • permissions (str) – The permissions dict to replace.
delete_object_permissions(*object_id_list)

Delete all listed object permissions.

Parameters:object_id (str) – Remove given objects permissions.

Errors

cliquet.errors.http_error(httpexception, errno=None, code=None, error=None, message=None, info=None, details=None)

Return a JSON formated response matching the error protocol.

Parameters:
  • httpexception – Instance of httpexceptions
  • errno – stable application-level error number (e.g. 109)
  • code – matches the HTTP status code (e.g 400)
  • error – string description of error type (e.g. “Bad request”)
  • message – context information (e.g. “Invalid request parameters”)
  • info – information about error (e.g. URL to troubleshooting)
  • details – additional structured details (conflicting record)
Returns:

the formatted response object

Return type:

pyramid.httpexceptions.HTTPException

cliquet.errors.json_error_handler(errors)

Cornice JSON error handler, returning consistant JSON formatted errors from schema validation errors.

This is meant to be used is custom services in your applications.

upload = Service(name="upload", path='/upload',
                 error_handler=errors.json_error_handler)

Warning

Only the first error of the list is formatted in the response. (c.f. protocol).

cliquet.errors.raise_invalid(request, location='body', name=None, description=None, **kwargs)

Helper to raise a validation error.

Parameters:
  • location – location in request (e.g. 'querystring')
  • name – field name
  • description – detailed description of validation error
Raises:

HTTPBadRequest

cliquet.errors.send_alert(request, message=None, url=None, code='soft-eol')

Helper to add an Alert header to the response.

Parameters:
  • code – The type of error ‘soft-eol’, ‘hard-eol’
  • message – The description message.
  • url – The URL for more information, default to the documentation url.

Utils

cliquet.utils.strip_whitespace(v)

Remove whitespace, newlines, and tabs from the beginning/end of a string.

Parameters:v (str) – the string to strip.
Return type:str
cliquet.utils.msec_time()

Return current epoch time in milliseconds.

Return type:int
cliquet.utils.classname(obj)

Get a classname from an object.

Return type:str
cliquet.utils.merge_dicts(a, b)

Merge b into a recursively, without overwriting values.

Parameters:a (dict) – the dict that will be altered with values of b.
Return type:None
cliquet.utils.random_bytes_hex(bytes_length)

Return a hexstring of bytes_length cryptographic-friendly random bytes.

Parameters:bytes_length (integer) – number of random bytes.
Return type:str
cliquet.utils.native_value(value)

Convert string value to native python values.

Parameters:value (str) – value to interprete.
Returns:the value coerced to python type
cliquet.utils.read_env(key, value)

Read the setting key from environment variables.

Parameters:
  • key – the setting name
  • value – default value if undefined in environment
Returns:

the value from environment, coerced to python type

cliquet.utils.encode64(content, encoding='utf-8')

Encode some content in base64.

Return type:str
cliquet.utils.decode64(encoded_content, encoding='utf-8')

Decode some base64 encoded content.

Return type:str
cliquet.utils.hmac_digest(secret, message, encoding='utf-8')

Return hex digest of a message HMAC using secret

cliquet.utils.dict_subset(d, keys)

Return a dict with the specified keys

cliquet.utils.reapply_cors(request, response)

Reapply cors headers to the new response with regards to the request.

We need to re-apply the CORS checks done by Cornice, in case we’re recreating the response from scratch.

cliquet.utils.current_service(request)

Return the Cornice service matching the specified request.

Returns:the service or None if unmatched.
Return type:cornice.Service
cliquet.utils.current_resource_name(request)

Return the name used when the Cliquet resource was registered along its viewset.

Returns:the resource identifier.
Return type:str
cliquet.utils.build_request(original, dict_obj)

Transform a dict object into a pyramid.request.Request object.

It sets a parent attribute on the resulting request assigned with the original request specified.

Parameters:
  • original – the original request.
  • dict_obj – a dict object with the sub-request specifications.
cliquet.utils.build_response(response, request)

Transform a pyramid.response.Response object into a serializable dict.

Parameters:
  • response – a response object, returned by Pyramid.
  • request – the request that was used to get the response.
cliquet.utils.follow_subrequest(request, subrequest, **kwargs)

Run a subrequest (e.g. batch), and follow the redirection if any.

Return type:tuple
Returns:the reponse and the redirection request (or subrequest if no redirection happened.)
cliquet.utils.encode_header(value, encoding='utf-8')

Make sure the value is of type str in both PY2 and PY3.

cliquet.utils.decode_header(value, encoding='utf-8')

Make sure the header is an unicode string.

cliquet.utils.strip_uri_prefix(path)

Remove potential version prefix in URI.

class cliquet.utils.DeprecatedMeta

A metaclass to be set on deprecated classes.

Warning will happen when class is inherited.

Glossary

Cliquet Protocol
A system of rules that explains the way to interact with the HTTP API endpoints (utilities, synchronization, headers etc.), and how data is organized (JSON responses etc.).
CRUD
Acronym for Create, Read, Update, Delete
endpoint
endpoints
An endpoint handles a particular HTTP verb at a particular URL.
extensible
«Extensible» means that the component behaviour can be overriden via lines of code. It differs from «pluggable».
Firefox Accounts
Account account system run by Mozilla (https://accounts.firefox.com).
HTTP API
Multiple publicly exposed endpoints that accept HTTP requests and respond with the requested data, in the form of JSON.
KISS
«Keep it simple, stupid» is a design priciple which states that most systems work best if they are kept simple rather than made complicated.
pluggable
«Pluggable» means that the component can be replaced via configuration. It differs from «extensible».
resource
A resource is a collection of records. A resource has two URLs, one for the collection and one for individual records.
user id
user identifier
user identifiers

A string that identifies a user. When using the built-in Basic Auth authentication, Cliquet uses cryptography (HMAC) to generate an identification string.

See Pyramid authentication.

object
objects
Also refered as «records», objects are stored by Cliquet resources.
tombstone
tombstones
When a record is deleted in a resource, a tombstone is created to keep track of the deletion when polling for changes. A tombstone only contains the id and last_modified fields, everything else is really deleted.
principal
principals
An entity that can be authenticated. Principals can be individual people, computers, services, or any group of such things.
permission
permissions
An action that can be authorized or denied. read, write, create are permissions.
Semantic Versioning
A standard MAJOR.MINOR.PATCH versioning scheme. See http://semver.org/.
ACE
ACEs
Access Control Entity
An association of a principal, an object and a permission. For instance, (Alexis, article, write).
ACL
ACLs
Access Control List
A list of Access Control Entities (ACE).

Ecosystem

This section gathers information about extending Cliquet, and third-party packages.

Packages

Note

If you build a package that you would like to see listed here, just get in touch with us!

Extending Cliquet

Pluggable components

Pluggable components can be substituted from configuration files, as long as the replacement follows the original component API.

# myproject.ini
cliquet.logging_renderer = cliquet_fluent.FluentRenderer

This is the simplest way to extend Cliquet, but will be limited to its existing components (cache, storage, log renderer, ...).

In order to add extra features, including external packages is the way to go!

Include external packages

Appart from usual python «import and use», Pyramid can include external packages, which can bring views, event listeners etc.

import cliquet
from pyramid.config import Configurator


def main(global_config, **settings):
    config = Configurator(settings=settings)

    cliquet.initialize(config, '0.0.1')
    config.scan("myproject.views")

    config.include('cliquet_elasticsearch')

    return config.make_wsgi_app()

Alternatively, packages can also be included via configuration:

# myproject.ini
cliquet.includes = cliquet_elasticsearch
                   pyramid_debugtoolbar

There are many available packages, and it is straightforward to build one.

Include me

In order to be included, a package must define an includeme(config) function.

For example, in cliquet_elasticsearch/init.py:

def includeme(config):
    settings = config.get_settings()

    config.add_view(...)

Configuration

In order to ease the management of settings, Cliquet provides a helper that reads values from environment variables and uses default application values.

import cliquet
from pyramid.settings import asbool


DEFAULT_SETTINGS = {
    'cliquet_elasticsearch.refresh_enabled': False
}


def includeme(config):
    cliquet.load_default_settings(config, DEFAULT_SETTINGS)
    settings = config.get_settings()

    refresh_enabled = settings['cliquet_elasticsearch.refresh_enabled']
    if asbool(refresh_enabled):
        ...

    config.add_view(...)

In this example, if the environment variable CLIQUET_ELASTICSEARCH_REFRESH_ENABLED is set to true, the value present in configuration file is ignored.

Declare API capabilities

Arbitrary capabilities can be declared and exposed in the root URL.

Clients can rely on this to detect optional features on the server. For example, features brought by plugins.

def main(global_config, **settings):
    config = Configurator(settings=settings)

    cliquet.initialize(config, __version__)
    config.scan("myproject.views")

    settings = config.get_settings()
    if settings['flush_enabled']:
        config.add_api_capability("flush",
                                  description="Flush server using endpoint",
                                  url="http://kinto.readthedocs.io/en/latest/configuration/settings.html#activating-the-flush-endpoint")

    return config.make_wsgi_app()

Note

Any argument passed to config.add_api_capability() will be exposed in the root URL.

Custom backend

As a simple example, let’s add add another kind of cache backend to Cliquet.

cliquet_riak/cache.py:

from cliquet.cache import CacheBase
from riak import RiakClient


class Riak(CacheBase):
    def __init__(self, **kwargs):
        self._client = RiakClient(**kwargs)
        self._bucket = self._client.bucket('cache')

    def set(self, key, value, ttl=None):
        key = self._bucket.new(key, data=value)
        key.store()
        if ttl is not None:
            # ...

    def get(self, key):
        fetched = self._bucked.get(key)
        return fetched.data

    #
    # ...see cache documentation for a complete API description.
    #


def load_from_config(config):
    settings = config.get_settings()
    uri = settings['cliquet.cache_url']
    uri = urlparse.urlparse(uri)

    return Riak(pb_port=uri.port or 8087)

Once its package installed and available in Python path, this new backend type can be specified in application configuration:

# myproject.ini
cliquet.cache_backend = cliquet_riak.cache

Adding features

Another use-case would be to add extra-features, like indexing for example.

  • Initialize an indexer on startup;
  • Add a /search/{collection}/ end-point;
  • Index records manipulated by resources.

Inclusion and startup in cliquet_indexing/__init__.py:

DEFAULT_BACKEND = 'cliquet_indexing.elasticsearch'

def includeme(config):
    settings = config.get_settings()
    backend = settings.get('cliquet.indexing_backend', DEFAULT_BACKEND)
    indexer = config.maybe_dotted(backend)

    # Store indexer instance in registry.
    config.registry.indexer = indexer.load_from_config(config)

    # Activate end-points.
    config.scan('cliquet_indexing.views')

End-point definitions in cliquet_indexing/views.py:

from cornice import Service

search = Service(name="search",
                 path='/search/{collection_id}/',
                 description="Search")

@search.post()
def get_search(request):
    collection_id = request.matchdict['collection_id']
    query = request.body

    # Access indexer from views using registry.
    indexer = request.registry.indexer
    results = indexer.search(collection_id, query)

    return results

Example indexer class in cliquet_indexing/elasticsearch.py:

class Indexer(...):
    def __init__(self, hosts):
        self.client = elasticsearch.Elasticsearch(hosts)

    def search(self, collection_id, query, **kwargs):
        try:
            return self.client.search(index=collection_id,
                                      doc_type=collection_id,
                                      body=query,
                                      **kwargs)
        except ElasticsearchException as e:
            logger.error(e)
            raise

    def index_record(self, collection_id, record, id_field):
        record_id = record[id_field]
        try:
            index = self.client.index(index=collection_id,
                                      doc_type=collection_id,
                                      id=record_id,
                                      body=record,
                                      refresh=True)
            return index
        except ElasticsearchException as e:
            logger.error(e)
            raise

Indexed resource in cliquet_indexing/resource.py:

class IndexedModel(cliquet.resource.Model):
    def create_record(self, record):
        r = super(IndexedModel, self).create_record(self, record)

        self.indexer.index_record(self, record)

        return r

class IndexedResource(cliquet.resource.UserResource):
    def __init__(self, request):
        super(IndexedResource, self).__init__(request)
        self.model.indexer = request.registry.indexer

Note

In this example, IndexedResource must be used explicitly as a base resource class in applications. A nicer pattern would be to trigger Pyramid events in Cliquet and let packages like this one plug listeners. If you’re interested, we started to discuss it!

JavaScript client

One of the main goal of Cliquet is to ease the development of REST microservices, most likely to be used in a JavaScript environment.

A client could look like this:

var client = new cliquet.Client({
    server: 'https://api.server.com',
    store: localforage
});

var articles = client.resource('/articles');

articles.create({title: "Hello world"})
  .then(function (result) {
    // success!
  });

articles.get('id-1234')
  .then(function (record) {
    // Read from local if offline.
  });

articles.filter({
    title: {'$eq': 'Hello'}
  })
  .then(function (results) {
    // List of records.
  });

articles.sync()
  .then(function (result) {
    // Synchronize offline store with server.
  })
  .catch(function (err) {
    // Error happened.
    console.error(err);
  });

Troubleshooting

We are doing the best we can so you do not have to read this section.

That said, we have included solutions (or at least explanations) for some common problems below.

If you do not find a solution to your problem here, please ask for help!

ConnectionError: localhost:6379. nodename nor servname provided, or not known

Make sure /etc/hosts has correct mapping to localhost.

IOError: [Errno 24] Too many open files

Make sure that max number of connections to redis-server and the max number of file handlers in operating system have access to required memory.

To fix this, increase the open file limit for non-root user:

$ ulimit -n 1024

ERROR: InterpreterNotFound: pypy

You need to install Pypy so that it can be found by tox.

Contributing

Thank you for considering to contribute to Cliquet!

Note

No contribution is too small; we welcome fixes about typos and grammar bloopers. Don’t hesitate to send us a pull request!

Note

Open a pull-request even if your contribution is not ready yet! It can be discussed and improved collaboratively, and avoid having you doing a lot of work without getting feedback.

Setup your development environment

To prepare your system with Python 3.4, PostgreSQL and Redis, please refer to the Installation guide.

You might need to install curl, if you don’t have it already.

Prepare your project environment by running:

$ make install-dev
$ pip install tox

OS X

On OSX especially you might get the following error when running tests:

$ ValueError: unknown locale: UTF-8

If this is the case add the following to your ~/.bash_profile:

export LC_ALL=en_US.UTF-8
export LANG=en_US.UTF-8

Then run:

$ source ~/.bash_profile

Run tests

Currently, running the complete test suite implies to run every type of backend.

That means:

  • Run Redis on localhost:6379
  • Run a PostgreSQL 9.4 testdb database on localhost:5432 with user postgres/postgres. The database encoding should be UTF-8, and the database timezone should be UTC.
make tests

Run a single test

For Test-Driven Development, it is a possible to run a single test case, in order to speed-up the execution:

nosetests -s --with-mocha-reporter cliquet.tests.test_views_hello:HelloViewTest.test_returns_info_about_url_and_version

Load tests

A load test is provided in order to run end-to-end tests on Cliquet through a sample application, or prevent regressions in terms of performance.

The following make command will check briefly the overall sanity of the API, by running a server and running a very few random HTTP requests on it.

make loadtest-check-simulation

It is possible to customise this load test, by focusing on a single action, or customise the number of users and hits.

First, run the test application manually in a terminal:

cd loadtests/
pserve loadtests/testapp.ini

And then, run the test suite against it:

SERVER_URL=http://localhost:8888 make test -e

To run a specific action, specify it with:

LOAD_ACTION=batch_create SERVER_URL=http://localhost:8888 make test -e

Or a specific configuration:

cp test.ini custom.ini
CONFIG=custom.ini SERVER_URL=http://localhost:8888 make test -e

Definition of done

In order to have your changes incorporated, you need to respect these rules:

  • Tests pass; Travis-CI will build the tests for you on the branch when you push it.
  • Code added comes with tests; We try to have a 100% coverage on the codebase to avoid surprises. No code should be untested :) If you fail to see how to test your changes, feel welcome to say so in the pull request, we’ll gladly help you to find out.
  • Documentation is up to date;

IRC channel

If you want to discuss with the team behind Cliquet, please come and join us on #storage on irc.mozilla.org.

  • Because of differing time zones, you may not get an immediate response to your question, but please be patient and stay logged into IRC — someone will almost always respond if you wait long enough (it may take a few hours).
  • If you don’t have an IRC client handy, use the webchat for quick feedback.
  • You can direct your IRC client to the channel using this IRC link or you can manually join the #storage IRC channel on the mozilla IRC network.

How to release

In order to prepare a new release, we are following the following steps.

The prerelease and postrelease commands are coming from zest.releaser.

Install zest.releaser with the recommended dependencies. They contain wheel and twine, which are required to release a new version.

$ pip install "zest.releaser[recommended]"

Step 1

  • Merge remaining pull requests
  • Update CHANGELOG.rst
  • Update version in cliquet_docs/conf.py
  • Known good versions of dependencies in requirements.txt
  • Update CONTRIBUTORS.rst using: git shortlog -sne | awk '{$1=""; sub(" ", ""); print}' | awk -F'<' '!x[$1]++' | awk -F'<' '!x[$2]++' | sort
$ git checkout -b prepare-X.Y.Z
$ prerelease
$ vim cliquet_docs/conf.py
$ make build-requirements
$ git commit -a --amend
$ git push origin prepare-X.Y.Z
  • Open a pull-request with to release the version.

Step 2

Once the pull-request is validated, merge it and do a release. Use the release command to invoke the setup.py, which builds and uploads to PyPI

$ git checkout master
$ git merge --no-ff prepare-X.Y.Z
$ release
$ postrelease

Step 3

As a final step:

  • Close the milestone in Github
  • Add entry in Github release page
  • Create next milestone in Github in the case of a major release
  • Configure the version in ReadTheDocs
  • Send mail to ML (If major release)

That’s all folks!

Cleaning your environment

There are three levels of cleaning your environment:

  • make clean will remove *.pyc files and __pycache__ directory.
  • make distclean will also remove *.egg-info files and *.egg, build and dist directories.
  • make maintainer-clean will also remove the .tox and the .venv directories.

Changelog

This document describes changes between each past release.

3.2.0 (unreleased)

Bug fixes

  • Add an explicit message when the server is configured as read-only and the collection timestamp fails to be saved (ref Kinto/kinto#558)

3.1.5 (2016-05-17)

Bug fixes

  • Prevent the browser to cache server responses between two sessions. (#593)

3.1.4 (2016-05-10)

Bug fixes

  • Redirects version prefix to hello page when trailing_slash_redirect is enabled. (#700)
  • Fix crash when setting empty permission list with PostgreSQL permission backend (fixes Kinto/kinto#575)
  • Fix crash when type of values in querystring for exclude/include is wrong (fixes Kinto/kinto#587)
  • Fix crash when providing duplicated principals in permissions with PostgreSQL permission backend (fixes #702)

3.1.3 (2016-04-29)

Bug fixes

  • Fix migration of triggers in PostgreSQL storage backend when upgrading from <3.0.0. Running the migrate command will basically re-create them (fixes Kinto/kinto#559)

Documentation

  • Fix some typos in documentation (#696)
  • Update links to readthedocs (#697)

3.1.2 (2016-04-19)

Bug fixes

  • Fix safe creation (If-None-Match: *) if a record used to exist with the same id (Kinto/kinto#512)

New features

  • A cache_prefix setting was added for cache backends. (#680)

3.1.1 (2016-04-05)

Bug fixes

  • Allow numeric comparison and sorting with the PostgreSQL backend. (#690)

3.1.0 (2016-03-07)

Protocol

  • Add the __lbheartbeat__ endpoint, for load balancer membership test.

New features

  • Default console log renderer now has colours (#671)

Bug fixes

  • Do not always return 412 errors when request header If-None-Match: * is sent on POST /collection (fixes #673)

Internal changes

  • Remove some imports at initialization to speed startup (#674)

3.0.0 (2016-02-26)

Breaking changes

  • Errors are not swallowed anymore during the execution of ResourceChanged events subscribers.

    Subscribers are still executed within the transaction like before.

    Subscribers are still executed even if transaction is eventually rolledback. Every subscriber execution succeeds, or none.

    Thus, subscribers of these events should only perform operations that are reversed on transaction rollback: most likely database storage operations.

    For irreversible operations see the new AfterResourceChanged event.

Protocol

  • Clients are redirected to URLs without trailing slash only if the current URL does not exist (#656)
  • Partial responses can now be specified for nested objects (#660, Kinto/kinto#445) For example, /records?_fields=address.street.
  • List responses are now sorted by to last_modified descending by default (#662, thanks @ayusharma)

New features

  • Resource events are now merged in batch requests. One event per resource and per action is emitted when a transaction is committed (#634)

  • Monitor time of events listeners execution (fixes #503)

  • Add method to remove a principal from every user

  • Validate that the client can accept JSON response. (#667)

  • Validate that the client can only send JSON request body. (#667)

  • Added a new AfterResourceChanged event, that is sent only when the commit in database is done and successful.

    Subscribers of this event can fail, errors are swallowed and logged. The final transaction result (or response) cannot be altered.

    Since commit occured successfully and operations will not be rolledback, subcribers running irreversible actions should subscribe to this event (like sending messages, deleting files, or run asynchronous tasks).

  • Track execution time on StatsD for each authentication sub-policy (#639)

  • Output the cliquet version with –version (#636)

Bug fixes

  • ResourceChanged events are not emitted if a batch subrequest fails (#634) There are still emitted if the whole batch transaction is eventually rolledback.
  • Fix a migration of PostgreSQL schema introduced in #604 that was never executed
  • Fix PostgreSQL backend timestamps when collection is empty (ref Kinto/kinto#433)
  • Fix statsd initialization on storage (#637)
  • Providing bad last modified values on delete now returns 400 (#665)
  • Providing last modified in the past for delete now follows behaviour create/update (#665)

Internal changes

  • Moved utils.current_service(request) to reified request method (ref #631)
  • Optimized (and cleaned) usage of (un)authenticated_userid (#641)
  • Apply request extensions on forged requests (ref Kinto/kinto#438)
  • Get rid of custom Enum (fixes #645)
  • Pyramid config link updated (#654, thanks @ayusharma)
  • Add missing readonly param for read operations in PostgreSQL (#653)
  • Move Retry-After to error responses in documentation (#657)
  • Improve batch endpoint documentation about transactions (ref #629)
  • Mac OS python installation command updated (#661)
  • Added details about forced timestamps (#665)
  • Added troubleshooting section in docs (thanks @ayusharma)

2.15.0 (2016-01-27)

Protocol

  • Forward slashes (/) are not escaped anymore in JSON responses (#537)
  • The API capabilities can be exposed in a capabilities attribute in the root URL (#628). Clients can rely on this to detect optional features on the server (e.g. enabled plugins).

Bug fixes

  • Batch now relies on custom views responses (@viewconfig(context=Error)) (fixes mozilla-services/syncto#78, rel #629)
  • Fix listener name logging during startup (#626)

New features

  • The policy name used to configure authentication in settings is now used for the user id prefix and StatsD authn_type counters.

2.14.0 (2016-01-15)

Protocol

  • Fields can be filtered in GET requests using _fields=f1,f2 in querystring (fixes #384)

New features

  • Add ability to filter fields from querystring (#601, thanks @MrChoclate)
  • Check backends configuration at startup (fixes Kinto/kinto#228)

Bug fixes

  • Do not overwrite original error message and details in batch requests (fixes #617)
  • Fix plugins not being able to use custom requests methods (#618)
  • Do not log batch subrequests twice (fixes #450, fixes Kinto/kinto#264)

Internal changes

  • Refactor filtering of events to use Pyramid subscribers predicates (fixes #609)
  • Refactor logger imports (fixes #102)

2.13.1 (2015-12-15)

Bug fixes

  • Fix hmac digest with python3 (fixes Kinto/kinto#288)

Documentation

  • Fixed typo in RHEL package manager command (#610, thanks @FooBarQuaxx)

2.13.0 (2015-12-01)

Bug fixes

  • Fixes duplicated records in paginated results when limit is forced via settings (fixes #588)

New features

  • Allow to keep the last_modified field of records when updating or creating them. Useful for replication of remote collections (fixes #604)

Internal changes

  • Documentation has been updated to explain better how versionning is handled (fixes #603)

2.12.0 (2015-11-27)

Protocol

Minor changes in the root URL (hello view):

  • Added http_api_version (#600)
  • Renamed hello to project_name
  • Renamed protocol_version to cliquet_protocol_version
  • Renamed documentation to project_docs
  • Renamed version to project_version

Breaking changes

  • When using cliquet-fxa, the setting multiauth.policy.fxa.use must now be explicitly set to cliquet_fxa.authentication.FxAOAuthAuthenticationPolicy
  • Fields in the root view were renamed (#600)

Bug fixes

  • Include plugins after setting up components (like authn/authz) so that plugins can register views with permissions checking
  • Remove __permissions__ from impacted records values in ResourceChanged events (#586)

New features

  • New options in configuration of listeners to specify filtered actions and resource names (#492, #555)
  • Add ability to listen to read action on resource (disabled by default) (#493)
  • Add ability to ask for partial response via _fields parameter on GET endpoints

Internal

  • Fixed a few details in quickstart docs since backends are not Redis by default anymore
  • Replace usage of assert by explicit exceptions since the former can be ignored when python is ran with -O (fixes #592)
  • Improved documentation about permissions (#572, thanks for the feedback @MrChoclate)
  • Fixed docs building under Python 3 (#591)

2.11.0 (2015-11-17)

Protocol

  • _since and _before now accepts an integer value between quotes ", as it would be returned in the ETag response header.
  • A batch request now fails if one of the subrequests fails (#510) (see new feature about transactions)

Breaking changes

  • For PostgreSQL backends, it is recommended to specify postgresql://.

New features

  • A transaction now covers the whole request/response cycle (#510, Kinto/kinto#194). If an error occurs during the request processing, every operation performed is rolled back. Note: This is only enabled with PostgreSQL backends. In other words, the rollback has no effect on backends like Redis or Memory.
  • Add the protocol_version to tell which protocol version is implemented by the service in the hello page. (#324)
  • New settings for backends when using PostgreSQL: *_pool_maxoverflow, *_pool_recycle, *_pool_timeout to control connections pool behaviour.
  • Add custom pool supporting a max_backlog parameter that limits the number of threads waiting for a connection (#509)
  • Add impacted_records attribute on ResourceChanged event (#501) This also allows listeners to react on particular field change, since old and new version of records is provided.

Bug fixes

  • Fix Service CORS not being set when plugins are included
  • Fix crash with Redis backend if record parent/id is unicode (fixes #556)
  • Fix principals of permission backend not being plugged by default (#573)
  • Fix Redis error traces not being logged (#560)
  • Fix principals of permission backend not being plugged by default. (#573)
  • Maintain pagination offset to prevent pagination loop in some cases. (#366)

Internal changes

  • Switch to SQLAlchemy for smarter connections pools.
  • Added a simple end-to-end test on a Cliquet sample application, using Loads. (fixes #512)
  • Switched to SQLAlchemy sessions instead of raw connections and cursors. (#510)
  • Refactor Redis clients instantiation to avoid repeated defaults. (#567, #568)
  • Initialize Service class attributes before including plugins. (#578)
  • Add a statsd_count helper function to ease the usage of statsd. (#574)
  • Mention SQLAlchemy on missing PostgreSQL dependencies. (#545)

2.10.2 (2015-11-10)

Bug fixes

  • Fix sharing records with ProtectedResource (fixes #549)
  • Fix notifications on protected resources (#548)
  • Log any heartbeat exception (fixes #559)
  • Fix crash with Redis backend if record parent/id is unicode (fixes #556)
  • Fix Redis client instantiation (fixes #564)

2.10.1 (2015-11-03)

Bug fixes

  • Make sure read enpoints (GET, OPTIONS, HEAD) are activated in readonly mode. (#539)

2.10.0 (2015-10-30)

Protocol

  • Moved userid attribute to a dedicated user mapping in the hello view.
  • Fixed 503 error message to mention backend errors in addition to unavailability.
  • Set cache headers only when anonymous (fixes #449)
  • Follow redirections in batch subrequests (fixes #511)
  • When recreating a record that was previously deleted, status code is now 201 (ref #530).

New features

  • Follow redirections in batch subrequests (fixes #511)
  • Add a readonly setting to run the service in read-only mode. (#525)
  • If no client cache is set, add Cache-Control: no-cache by default, so that clients are forced to revalidate their cache against the server (#522, ref Kinto/kinto#231)

Bug fixes

  • Fix PostgreSQL error when deleting an empty collection in a protected resource (fixes #528)
  • Fix PUT not using create() method in storage backend when tombstone exists (fixes #530)
  • Delete tombstone when record is re-created (fixes #518)
  • Fix crash with empty body for PATCH (fixes #477, fixes #516)
  • Fix english typo in 404 error message (fixes #527)

Internal changes

  • Better __pycache__ cleaning

2.9.0 (2015-10-27)

New features

  • Added Pyramid events, triggered when the content of a resource has changed. (#488)
  • Added cliquet.includes setting allowing loading of plugins once Cliquet is initialized (unlike pyramid.includes). (#504)

Protocol

  • Remove the broken git revision commit field in the hello page. (#495).

Breaking changes

  • Renamed internal backend classes for better consistency. Settings remain unchanged, but if you imported the backend classes in your Cliquet application, it will break (#491).
  • cliquet.schema is now deprecated, and was moved to a cliquet.resource module. (#505)
  • Resource collection attribute is now deprecated. Use model attribute instead. (#506)

Internal changes

  • Rework PostgreSQL backends to use composition instead of inheritance for the client code. (#491)
  • Replace DROP INDEX by a conditional creation in PostgreSQL schemas (#487, #496 thanks @rodo)
  • Documentation and minor refactors in viewset code (#490, #498, #502)
  • Add the build-requirements, distclean and maintainer-clean Makefile rules.
  • Documentation JSON patch format. (#484)
  • Fix for permission among record fields in 412 errors. (#499)

2.8.2 (2015-10-22)

Bug fixes

  • Fix crash on settings with list values (#481)
  • Fix crash in Redis permission backend (ref Kinto/kinto#215)

Internal changes

  • Use tox installed in virtualenv (#486)
  • Skip python versions unavailable in tox (#486)

2.8.1 (2015-10-14)

  • Expose public settings without prefix, except if we explicitely configure public_settings to expose them (with cliquet. or project_name.) (ref #476)

2.8.0 (2015-10-06)

Breaking changes

  • Deprecated settings cliquet.cache_pool_maxconn, cliquet.storage_pool_maxconn and cliquet.basic_auth_enabled were removed (ref #448)
  • Prefixed settings will not work if project_name is not defined. (either with cliquet.initialize() or with the cliquet.project_name configuration variable).
  • Settings should now be read without their prefix in the code: request.registry.settings['max_duration'] rather than request.registry.settings['cliquet.max_duration']

New features

  • Add cache CORS headers. (ref #466)
  • Use the project name as setting prefix (ref #472)

Internal changes

  • Expose statsd client so that projects using cliquet can send statsd metrics. (ref #465)
  • Refactor BaseWebTest. (ref #468)
  • Remove hard coded CORS origins in order to be able to override it with config. (ref #467)
  • Allow overridding 405 response error to give context (ref #471)
  • Allow overridding 503 response error to give context (ref #473)

2.7.0 (2015-09-23)

Breaking changes

  • Backends are not instantiated by default anymore (used to be with Redis) (#461)

New features

  • Redirect to remove trailing slash in URLs (fixes Kinto/kinto#112)
  • Add resource cache control headers via settings (fixes #401)
  • Add request bound_data attribute, shared with subrequests. Useful to share context or cache values between BATCH requests for example (#459)

Bug fixes

  • Fix Werkzeug profiling setup docs and code (#451)
  • Fix logger encoding error with UTF-8 output (#455)
  • Do not instantiate backends if not configured (fixes #386)

Internal changes

  • Huge refactoring the interaction between Resource and Permission backend (#454)
  • Fetch record only once from storage with PUT requests on resources (#452)
  • Index permissions columns, bringing huge performance gain for shared collections (#458, ref #354)
  • Add instructions to mention contributors list in documentation (#408)
  • Explicitly call to collection create_record on PUT (#460)

2.6.2 (2015-09-09)

Bug fixes

  • Expose CORS headers on subrequest error response and for non service errors (#435).
  • Make sure a tuple is passed for Postgresql list comparisons even for ids (#443).

Internal changes

  • Use the get_bound_permissions callback to select shared records in collection list (#444).

2.6.1 (2015-09-08)

Bug fixes

  • Make sure a tuple is passed for Postgresql in conditions (#441).

2.6.0 (2015-09-08)

Protocol

  • Fix consistency in API to modify permissions with PATCH (#437, ref Kinto/kinto#155). The list of principals for each specified permission is now replaced by the one provided.

New features

  • Partial collection of records for ProtectedResource when user has no read permission (fixes #354). Alice can now obtain a list of Bob records on which she has read/write permission.

Internal changes

  • Fix Wheel packaging for Pypy (fixes Kinto/kinto#177)
  • Add additional test to make sure 400 errors returns CORS Allowed Headers

2.5.0 (2015-09-04)

Protocol

  • Collection records can now be filtered using multiple values (?in_status=1,2,3) (fixes #39)
  • Collection records can now be filtered excluding multiple values (?exclude_status=1,2,3) (fixes mozilla-services/readinglist#68)

Internal changes

  • We can obtains accessible objects_id in a collection from user principals (fixes #423)

2.4.3 (2015-08-26)

Bug fixes

  • Fix the packaging for cliquet (#430)

2.4.2 (2015-08-26)

Internal changes

  • Remove the symlink to cliquet_docs and put the documentation inside cliquet_docs directly (#426)

2.4.1 (2015-08-25)

Internal changes

  • Make documentation available from outside by using cliquet_docs (#413)

2.4.0 (2015-08-14)

Protocol

  • Userid is now provided when requesting the hello endpoint with an Authorization header (#319)
  • UUID validation now accepts any kind of UUID, not just v4 (fixes #387)
  • Querystring parameter _to was renamed to _before (the former is now deprecated) (#391)

New features

  • Cliquet Service class now has the default error handler attached (#388)
  • Allow to configure info link in error responses with cliquet.error_info_link setting (#395)
  • Storage backend now has a purge_deleted() to get rid of tombstones (#400)

Bug fixes

  • Fix missing Backoff header for 304 responses (fixes #416)
  • Fix Python3 encoding errors (#328)
  • data is not mandatory in request body if the resource does not define any schema or if no field is mandatory (fixes mozilla-services/kinto#63)
  • Fix no validation error on PATCH with unknown attribute (fixes #374)
  • Fix permissions not validated on PATCH (fixes #375)
  • Fix CORS header missing in 404 responses for unknown URLs (fixes #414)

Internal changes

  • Renamed main documentation sections to HTTP Protocol and Internals (#394)
  • Remove mentions of storage in documentation to avoid confusions with the Kinto project.
  • Add details in timestamp documentation.
  • Mention talk at Python Meetup Barcelona in README
  • Fix documentation about postgres-contrib dependancy (#409)
  • Add cliquet.utils to Internals documentation (#407)
  • Default id generator now accepts dashes and underscores (#411)

2.3.1 (2015-07-15)

Bug fixes

  • Fix crash on hello view when application is not deployed from Git repository (fixes #382)
  • Expose Content-Length header to Kinto.js (#390)

2.3 (2015-07-13)

New features

  • Provide details about existing record in 412 error responses (fixes mozilla-services/kinto#122)
  • Add ETag on record PUT/PATCH responses (fixes #352)
  • Add StatsD counters for the permission backend

Bug fixes

  • Fix crashes in permission backends when permission set is empty (fixes #368, #371)
  • Fix value of ETag on record: provide collection timestamp on collection endpoints only (fixes #356)
  • Default resources do accept permissions attribute in payload anymore
  • Default resources do not require a root factory (fixes #348)
  • Default resources do not hit the permission backend anymore
  • Default viewset was split and does not handle permissions anymore (fixes #322)
  • Permissions on views is now set only on resources
  • Fix missing last_modified field in PATCH response when no field was changed (fixes #371)
  • Fix lost querystring during version redirection (fixes #364)

Internal changes

  • Document the list of public settings in hello view (mozilla-services/kinto#133)

2.2.1 (2015-07-06)

Bug fixes

  • Fix permissions handling on PATCH /resource (#358)

2.2.0 (2015-07-02)

New features

  • Add public settings in hello view (#318)

Bug fixes

  • Fix version redirection behaviour for unsupported versions (#341)
  • PostgreSQL dependencies are now fully optional in code (#340)
  • Prevent overriding final settings from default_settings parameter in cliquet.initialize() (#343)

Internal changes

  • Fix installation documentation regarding PostgreSQL 9.4 (#338, thanks @elemoine!)
  • Add detail about UTC and UTF-8 for PostgreSQL (#347, thanks @elemoine!)
  • Remove UserWarning exception when running tests (#339, thanks @elemoine!)
  • Move build_request and build_response to cliquet.utils (#344)
  • Pypy is now tested on Travis CI (#337)

2.1.0 (2015-06-26)

New features

  • Cliquet does not require authentication policies to prefix user ids anymore (fixes #299).
  • Pypy support (thanks Balthazar Rouberol #325)
  • Allow to override parent id of resources (#333)

Bug fixes

  • Fix crash in authorization on OPTIONS requests (#331)
  • Fix crash when If-Match is provided without If-None-Match (#335)

Internal changes

  • Fix docstrings and documentation (#329)

2.0.0 (2015-06-16)

New features

  • Authentication and authorization policies, as well as group finder function can now be specified via configuration (fixes #40, #265)
  • Resources can now be protected by fine-grained permissions (#288 via #291, #302)

Minor

  • Preserve provided id field of records using POST on collection (#293 via #294)
  • Logging value for authentication type is now available for any kind of authentication policy.
  • Any resource endpoint can now be disabled from settings (#46 via #268)

Bug fixes

  • Do not limit cache values to string (#279)
  • When PUT creates the record, the HTTP status code is now 201 (#298, #300)
  • Add safety check in utils.current_service() (#316)

Breaking changes

  • cliquet.storage.postgresql now requires PostgreSQL version 9.4, since it now relies on JSONB. Data will be migrated automatically using the migrate command.
  • the @crud decorator was replaced by @register() (fixes #12, #268)
  • Firefox Accounts code was removed and published as external package cliquet-fxa
  • The Cloud storage storage backend was removed out of Cliquet and should be revamped in Kinto repository (mozilla-services/kinto#45)

API

  • Resource endpoints now expect payloads to have a data attribute (#254, #287)
  • Resource endpoints switched from If-Modified-Since and If-Unmodified-Since to Etags (fixes #251 via #275), thanks @michielbdejong!

Minor

  • existing attribute of conflict errors responses was moved inside a generic details attribute that is also used to list validation errors.
  • Setting cliquet.basic_auth_enabled is now deprecated. Use pyramid_multiauth configuration instead to specify authentication policies.
  • Logging value for authentication type is now authn_type (with FxAOAuth or BasicAuth as default values).

Internal changes

  • Cliquet resource code was split into Collection and Resource (fixes #243, #282)
  • Cleaner separation of concern between Resource and the new notion of ViewSet (#268)
  • Quickstart documentation improvement (#271, #312) thanks @N1k0 and @brouberol!
  • API versioning documentation improvements (#313)
  • Contribution documentation improvement (#306)

1.8.0 (2015-05-13)

Breaking changes

  • Switch PostgreSQL storage to JSONB: requires 9.4+ (#104)
  • Resource name is not a Python property anymore (ref #243)
  • Return existing record instead of raising 409 on POST (fixes #75)
  • cliquet.storage.postgresql now requires version PostgreSQL 9.4, since it now relies on JSONB. Data will be migrated automatically using the migrate command.
  • Conflict errors responses existing attribute was moved inside a generic details attribute that is also used to list validation errors.
  • In heartbeat end-point response, database attribute was renamed to storage

New features

  • Storage records ids are now managed in python (fixes #71, #208)
  • Add setting to disable version redirection (#107, thanks @hiromipaw)
  • Add response behaviour headers for PATCH on record (#234)
  • Provide details in error responses (#233)
  • Expose new function cliquet.load_default_settings() to ease reading of settings from defaults and environment (#264)
  • Heartbeat callback functions can now be registered during startup (#261)

Bug fixes

  • Fix migration behaviour when metadata table is flushed (#221)
  • Fix backoff header presence if disabled in settings (#238)

Internal changes

  • Require 100% of coverage for tests to pass
  • Add original error message to storage backend error
  • A lots of improvements in documentation (#212, #225, #228, #229, #237, #246, #247, #248, #256, #266, thanks Michiel De Jong)
  • Migrate Kinto storage schema on startup (#218)
  • Fields id and last_modified are not part of resource schema anymore (#217, mozilla-services/readinlist#170)
  • Got rid of redundant indices in storage schema (#208, ref #138)
  • Disable Cornice schema request binding (#172)
  • Do not hide FxA errors (fixes mozilla-services/readinglist#70)
  • Move initialization functions to dedicated module (ref #137)
  • Got rid of request custom attributes for storage and cache (#245)

1.7.0 (2015-04-10)

Breaking changes

  • A command must be ran during deployment for database schema migration:

    $ cliquet –ini production.ini migrate

  • Sentry custom code was removed. Sentry logging is now managed through the logging configuration, as explained in docs.

New features

  • Add PostgreSQL schema migration system (#139)
  • Add cache and oauth in heartbeat view (#184)
  • Add monitoring features using NewRelic (#189)
  • Add profiling features using Werkzeug (#196)
  • Add ability to override default settings in initialization (#136)
  • Add more statsd counter for views and authentication (#200)
  • Add in-memory cache class (#127)

Bug fixes

  • Fix crash in DELETE on collection with PostgreSQL backend
  • Fix Heka logging format of objects (#199)
  • Fix performance of record insertion using ordered index (#138)
  • Fix 405 errors not JSON formatted (#88)
  • Fix basic auth prompt when disabled (#182)

Internal changes

  • Improve development setup documentation (thanks @hiromipaw)
  • Deprecated cliquet.initialize_cliquet, renamed to cliquet.initialize.
  • Code coverage of tests is now 100%
  • Skip unstable tests on TravisCI, caused by fsync = off in their PostgreSQL.
  • Perform random creation and deletion in heartbeat view (#202)

1.6.0 (2015-03-30)

New features

  • Split schema initialization from application startup, using a command-line tool.
cliquet --ini production.ini init

Bug fixes

  • Fix connection pool no being shared between cache and storage (#176)
  • Default connection pool size to 10 (instead of 50) (#176)
  • Warn if PostgreSQL session has not UTC timezone (#177)

Internal changes

  • Deprecated cliquet.storage_pool_maxconn and cliquet.cache_pool_maxconn settings (renamed to cliquet.storage_pool_size and cliquet.cache_pool_size)

1.5.0 (2015-03-27)

New features

  • Mesure calls on the authentication policy (#167)

Breaking changes

  • Prefix statsd metrics with the value of cliquet.statsd_prefix or cliquet.project_name (#162)
  • http_scheme setting has been replaced by cliquet.http_scheme and cliquet.http_host was introduced ((#151, #166)
  • URL in the hello view now has version prefix (#165)

Bug fixes

  • Fix Next-Page url if service has key in url (#158)
  • Fix some PostgreSQL connection bottlenecks (#170)

Internal changes

  • Update of PyFxA to get it working with gevent monkey patching (#168)
  • Reload kinto on changes (#158)

1.4.1 (2015-03-25)

Bug fixes

  • Rely on Pyramid API to build pagination Next-Url (#147)

1.4.0 (2015-03-24)

Breaking changes

  • Make monitoring dependencies optional (#121)

Bug fixes

  • Force PostgreSQl session timezone to UTC (#122)
  • Fix basic auth ofuscation and prefix (#128)
  • Make sure the paginate_by setting overrides the passed limit argument (#129)
  • Fix limit comparison under Python3 (#143)
  • Do not serialize using JSON if not necessary (#131)
  • Fix crash of classic logger with unicode (#142)
  • Fix crash of CloudStorage backend when remote returns 500 (#142)
  • Fix behaviour of CloudStorage with backslashes in querystring (#142)
  • Fix python3.4 segmentation fault (#142)
  • Add missing port in Next-Page header (#147)

Internal changes

  • Use ujson again, it was removed in the 1.3.2 release (#132)

  • Add index for as_epoch(last_modified) (#130). Please add the following statements to SQL for the migration:

    ALTER FUNCTION as_epoch(TIMESTAMP) IMMUTABLE;
    CREATE INDEX idx_records_last_modified_epoch ON records(as_epoch(last_modified));
    CREATE INDEX idx_deleted_last_modified_epoch ON deleted(as_epoch(last_modified));
    
  • Prevent fetching to many records for one user collection (#130)

  • Use UPSERT for the heartbeat (#141)

  • Add missing OpenSSL in installation docs (#146)

  • Improve tests of basic auth (#128)

1.3.2 (2015-03-20)

  • Revert ujson usage (#132)

1.3.1 (2015-03-20)

Bug fixes

  • Fix packaging (#118)

1.3.0 (2015-03-20)

New features

  • Add PostgreSQL connection pooling, with new settings cliquet.storage_pool_maxconn and cliquet.cache_pool_maxconn (Default: 50) (#112)
  • Add StatsD support, enabled with cliquet.statsd_url = udp://server:port (#114)
  • Add Sentry support, enabled with cliquet.sentry_url = http://user:pass@server/1 (#110)

Bug fixes

  • Fix FxA verification cache not being used (#103)
  • Fix heartbeat database check (#109)
  • Fix PATCH endpoint crash if request has no body (#115)

Internal changes

  • Switch to ujson for JSON de/serialization optimizations (#108)

1.2.1 (2015-03-18)

  • Fix tests about unicode characters in BATCH querystring patch
  • Remove CREATE CAST for the postgresql backend
  • Fix environment variable override

1.2 (2015-03-18)

Breaking changes

  • cliquet.storage.postgresql now uses UUID as record primary key (#70)
  • Settings cliquet.session_backend and cliquet.session_url were renamed cliquet.cache_backend and cliquet.cache_url respectively.
  • FxA user ids are not hashed anymore (#82)
  • Setting cliquet.retry_after was renamed cliquet.retry_after_seconds
  • OAuth2 redirect url now requires to be listed in fxa-oauth.webapp.authorized_domains (e.g. *.mozilla.com)
  • Batch are now limited to 25 requests by default (#90)

New features

  • Every setting can be specified via an environment variable (e.g. cliquet.storage_url with CLIQUET_STORAGE_URL)
  • Logging now relies on structlog (#78)
  • Logging output can be configured to stream JSON (#78)
  • New cache backend for PostgreSQL (#44)
  • Documentation was improved on various aspects (#64, #86)
  • Handle every backend errors and return 503 errors (#21)
  • State verification for OAuth2 dance now expires after 1 hour (#83)

Bug fixes

  • FxA OAuth views errors are now JSON formatted (#67)
  • Prevent error when pagination token has bad format (#72)
  • List of CORS exposed headers were fixed in POST on collection (#54)

Internal changes

  • Added a method in cliquet.resource.Resource to override known fields (required by Kinto)
  • Every setting has a default value
  • Every end-point requires authentication by default
  • Session backend was renamed to cache (#96)

1.1.4 (2015-03-03)

  • Update deleted_field support for postgres (#62)

1.1.3 (2015-03-03)

  • Fix include_deleted code for the redis backend (#60)
  • Improve the update_record API (#61)

1.1.2 (2015-03-03)

  • Fix packaging to include .sql files.

1.1.1 (2015-03-03)

  • Fix packaging to include .sql files.

1.1 (2015-03-03)

New features

  • Support filter on deleted using since (#51)

Internal changes

  • Remove python 2.6 support (#50)
  • Renamed Resource.deleted_mark to Resource.deleted_field (#51)
  • Improve native_value (#56)
  • Fixed Schema options inheritance (#55)
  • Re-build the virtualenv when setup.py changes
  • Renamed storage.url to cliquet.storage_url (#49)
  • Refactored the tests/support.py file (#38)

1.0 (2015-03-02)

  • Initial version, extracted from Mozilla Services Reading List project (#1)

New features

  • Expose CORS headers so that client behind CORS policy can access them (#5)
  • Postgresql Backend (#8)
  • Use RedisSession as a cache backend for PyFxA (#10)
  • Delete multiple records via DELETE on the collection_path (#13)
  • Batch default prefix for endpoints (#14 / #16)
  • Use the app version in the / endpoint (#22)
  • Promote Basic Auth as a proper authentication backend (#37)

Internal changes

  • Backends documentation (#15)
  • Namedtuple for filters and sort (#17)
  • Multiple DELETE in Postgresql (#18)
  • Improve Resource API (#29)
  • Refactoring of error management (#41)
  • Default Options for Schema (#47)

Contributors

Indices and tables