Cliquet

------------------

**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/readinlist170)
- 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/readinglist70)
- Move initialization functions to dedicated module (ref 137)
- Got rid of request custom attributes for storage and cache (245)

------------------

**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 <http://raven.readthedocs.io/en/latest/integrations/pyramid.htmllogger-setup>`_.

**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)

------------------

**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``)

------------------

**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)

------------------

**Bug fixes**

- Rely on Pyramid API to build pagination Next-Url (147)

------------------

**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)