Files
@ 4791487dbec1
Branch filter:
Location: kallithea/docs/administrator_guide/vcs_setup.rst - annotation
4791487dbec1
1.7 KiB
text/prs.fallenstein.rst
api: stop using 'Optional', 'OAttr'/'OptionalAttr' classes
There does not seem to be a good reason to use the 'Optional' and
'OptionalAttr' classes.
It makes the code harder to understand. And worse, the 'default value'
specified is not always used, which can thus give false information to
users.
The way Optional was used in the API calls is twofold:
1.either by effectively extracting a value, via Optional.extract(param).
If 'param' was indeed specified by the user, then this would yield that
user-specified value. Otherwise, it would yield the value declared in the
parameter declaration, e.g. param=Optional(defaultvalue).
2.or by checking if a variable is an instance of the Optional class. In case
a user effectively passed a value, this value will not be of the
Optional class. So if a parameter is an object of class Optional, we know
the user did not pass a value, and we can apply some default.
In the declaration of the parameter, the specified default value will only
be used if the 'extract' method is used, i.e. method 1 above.
A simpler way to address this problem of default values is just with Python
default values, using 'None' as magic value if the default will be
calculated inside the method.
The docstrings still specify something like:
type: Optional(bool)
which is humanly readable and does not necessarily refer to a class called
'Optional', so such strings are kept.
There does not seem to be a good reason to use the 'Optional' and
'OptionalAttr' classes.
It makes the code harder to understand. And worse, the 'default value'
specified is not always used, which can thus give false information to
users.
The way Optional was used in the API calls is twofold:
1.either by effectively extracting a value, via Optional.extract(param).
If 'param' was indeed specified by the user, then this would yield that
user-specified value. Otherwise, it would yield the value declared in the
parameter declaration, e.g. param=Optional(defaultvalue).
2.or by checking if a variable is an instance of the Optional class. In case
a user effectively passed a value, this value will not be of the
Optional class. So if a parameter is an object of class Optional, we know
the user did not pass a value, and we can apply some default.
In the declaration of the parameter, the specified default value will only
be used if the 'extract' method is used, i.e. method 1 above.
A simpler way to address this problem of default values is just with Python
default values, using 'None' as magic value if the default will be
calculated inside the method.
The docstrings still specify something like:
type: Optional(bool)
which is humanly readable and does not necessarily refer to a class called
'Optional', so such strings are kept.
2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 52f823b92614 52f823b92614 52f823b92614 52f823b92614 52f823b92614 52f823b92614 52f823b92614 52f823b92614 52f823b92614 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe 2bb5e9ee49fe | .. _vcs_setup:
=============================
Version control systems setup
=============================
Kallithea supports Git and Mercurial repositories out-of-the-box.
For Git, you do need the ``git`` command line client installed on the server.
You can always disable Git or Mercurial support by editing the
file ``kallithea/__init__.py`` and commenting out the backend. For example, to
disable Git but keep Mercurial enabled:
.. code-block:: python
BACKENDS = {
'hg': 'Mercurial repository',
#'git': 'Git repository',
}
Git-specific setup
------------------
Web server with chunked encoding
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Large Git pushes require an HTTP server with support for
chunked encoding for POST. The Python web servers waitress_ and
gunicorn_ (Linux only) can be used. By default, Kallithea uses
waitress_ for `gearbox serve` instead of the built-in `paste` WSGI
server.
The web server used by gearbox is controlled in the .ini file::
use = egg:waitress#main
or::
use = egg:gunicorn#main
Also make sure to comment out the following options::
threadpool_workers =
threadpool_max_requests =
use_threadpool =
Increasing Git HTTP POST buffer size
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
If Git pushes fail with HTTP error code 411 (Length Required), you may need to
increase the Git HTTP POST buffer. Run the following command as the user that
runs Kallithea to set a global Git variable to this effect::
git config --global http.postBuffer 524288000
.. _waitress: http://pypi.python.org/pypi/waitress
.. _gunicorn: http://pypi.python.org/pypi/gunicorn
.. _subrepositories: http://mercurial.aragost.com/kick-start/en/subrepositories/
|