[schema] load schema from modules names instead of directories
Introspect cubicweb, cubes and apphome using pkgutil to generate the full list
of modules names for loading the schema.
Keep historical behavior and check if source .py file exists if a module is
found using python bytecode file (.pyc and .pyo)
Loading schema from apphome require apphome to be present in sys.path and that
"schema" module resolve to a file located in apphome.
Update migraction tests to explicitely update sys.path when loading schema from
different apps, use a contextmanager for this so it's more readable.
Require updated logilab-common and yams
.. _profiling:
Profiling
=========
Profiling of requests by the pyramid debug toolbar can be a little restrictive
when a specific url needs thin profiling that includes the whole pyramid
dispatch.
Pyramid CubicWeb provides facilities to profile requests as a
:func:`wsgi middleware <cubicweb.pyramid.profile.wsgi_profile>`, and a few
views that facilitate profiling of basic features.
The views and the wsgi middleware are activated when the 'profile' option is
given. This can be done on the command line
(:option:`cubicweb-ctl pyramid --profile`) or in the :ref:`pyramid_settings`.
Views
-----
The following routes and corresponding views are provided when profiling is on:
- ``/_profile/ping``: Reply 'ping' without doing anything else. See also
:func:`cubicweb.pyramid.profile.ping`.
- ``/_profile/cnx``: Reply 'ping' after getting a cnx. See also
:func:`cubicweb.pyramid.profile.cnx`.
Typical Usage
-------------
Let's say we want to measure the cost of having a ``cnx``.
- Start the application with profile enabled:
.. code-block:: console
$ cubicweb-ctl pyramid --no-daemon --profile --profile-dump-every 100
- Use 'ab' or any other http benchmark tool to throw a lot of requests:
.. code-block:: console
$ ab -c 1 -n 100 http://localhost:8080/_profile/cnx
- Analyse the results. I personnaly fancy SnakeViz_:
.. code-block:: console
$ snakeviz program.prof
.. _SnakeViz: http://jiffyclub.github.io/snakeviz/