Files
@ 877bcf22bf71
Branch filter:
Location: kallithea/docs/installation.rst
877bcf22bf71
7.2 KiB
text/prs.fallenstein.rst
pytest migration: introduce TestControllerPytest
In order to allow tests to benefit from pytest specific functionality, like
fixtures, they can no longer derive from unittest.TestCase. What's more,
while they can derive from any user-defined class, none of the classes
involved (the test class itself nor any of the base classes) can have an
__init__ method.
Converting all tests from unittest-style to pytest-style in one commit is
not realistic. Hence, a more gradual approach is needed.
Most existing test classes derive from TestController, which in turn derives
from BaseTestCase, which derives from unittest.TestCase. Some test classes
derive directly from BaseTestCase.
Supporting both unittest-style and pytest-style from TestController directly
is not possible: pytest-style _cannot_ and unittest-style _must_ derive from
unittest.TestCase. Thus, in any case, an extra level in the class hierarchy
is needed (TestController deriving from Foo and from unittest.TestCase;
pytest-style test classes would then directly derive from Foo).
The requirement that pytest-style test classes cannot have an __init__
method anywhere in the class hierarchy imposes another restriction that
makes it difficult to support both unittest-style and pytest-style test
classes with one class. Any init code needs to be placed in another method
than __init__ and be called explicitly when the test class is initialized.
For unittest-style test classes this would naturally be done with a
setupClass method, but several test classes already use that. Thus, there
would need to be explicit 'super' calls from the test classes. This is
technically possible but not very nice.
A more transparent approach (from the existing test classes point of view),
implemented by this patch, works as follows:
- the implementation of the existing TestController class is now put under
a new class BaseTestController. To accomodate pytest, the __init__ method
is renamed init.
- contrary to the original TestController, BaseTestController does not
derive from BaseTestCase (and neither from unittest.TestCase). Instead,
the 'new' TestController derives both from BaseTestCase, which is
untouched, and from BaseTestController.
- TestController has an __init__ method that calls the base classes'
__init__ methods and the renamed 'init' method of BaseTestController.
- a new class TestControllerPytest is introduced that derives from
BaseTestController but not from BaseTestCase. It uses a pytest fixture to
automatically call the setup functionality previously provided by
BaseTestCase and also calls 'init' on BaseTestController. This means a
little code duplication but is hard to avoid.
The app setup fixture is scoped on the test method, which means that the app
is recreated for every test (unlike for the unittest-style tests where the
app is created per test class). This has the advantage of detecting current
inter-test dependencies and thus improve the health of our test suite. This
in turn is one step closer to allowing parallel test execution.
The unittest-style assert methods (assertEqual, assertIn, ...) do not exist
for pytest-style tests. To avoid having to change all existing test cases
upfront, provide transitional implementations of these methods. The
conversion of the unittest asserts to the pytest/python asserts can happen
gradually over time.
In order to allow tests to benefit from pytest specific functionality, like
fixtures, they can no longer derive from unittest.TestCase. What's more,
while they can derive from any user-defined class, none of the classes
involved (the test class itself nor any of the base classes) can have an
__init__ method.
Converting all tests from unittest-style to pytest-style in one commit is
not realistic. Hence, a more gradual approach is needed.
Most existing test classes derive from TestController, which in turn derives
from BaseTestCase, which derives from unittest.TestCase. Some test classes
derive directly from BaseTestCase.
Supporting both unittest-style and pytest-style from TestController directly
is not possible: pytest-style _cannot_ and unittest-style _must_ derive from
unittest.TestCase. Thus, in any case, an extra level in the class hierarchy
is needed (TestController deriving from Foo and from unittest.TestCase;
pytest-style test classes would then directly derive from Foo).
The requirement that pytest-style test classes cannot have an __init__
method anywhere in the class hierarchy imposes another restriction that
makes it difficult to support both unittest-style and pytest-style test
classes with one class. Any init code needs to be placed in another method
than __init__ and be called explicitly when the test class is initialized.
For unittest-style test classes this would naturally be done with a
setupClass method, but several test classes already use that. Thus, there
would need to be explicit 'super' calls from the test classes. This is
technically possible but not very nice.
A more transparent approach (from the existing test classes point of view),
implemented by this patch, works as follows:
- the implementation of the existing TestController class is now put under
a new class BaseTestController. To accomodate pytest, the __init__ method
is renamed init.
- contrary to the original TestController, BaseTestController does not
derive from BaseTestCase (and neither from unittest.TestCase). Instead,
the 'new' TestController derives both from BaseTestCase, which is
untouched, and from BaseTestController.
- TestController has an __init__ method that calls the base classes'
__init__ methods and the renamed 'init' method of BaseTestController.
- a new class TestControllerPytest is introduced that derives from
BaseTestController but not from BaseTestCase. It uses a pytest fixture to
automatically call the setup functionality previously provided by
BaseTestCase and also calls 'init' on BaseTestController. This means a
little code duplication but is hard to avoid.
The app setup fixture is scoped on the test method, which means that the app
is recreated for every test (unlike for the unittest-style tests where the
app is created per test class). This has the advantage of detecting current
inter-test dependencies and thus improve the health of our test suite. This
in turn is one step closer to allowing parallel test execution.
The unittest-style assert methods (assertEqual, assertIn, ...) do not exist
for pytest-style tests. To avoid having to change all existing test cases
upfront, provide transitional implementations of these methods. The
conversion of the unittest asserts to the pytest/python asserts can happen
gradually over time.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 | .. _installation:
==========================
Installation on Unix/Linux
==========================
The following describes three different ways of installing Kallithea:
- :ref:`installation-source`: The simplest way to keep the installation
up-to-date and track any local customizations is to run directly from
source in a Kallithea repository clone, preferably inside a virtualenv
virtual Python environment.
- :ref:`installation-virtualenv`: If you prefer to only use released versions
of Kallithea, the recommended method is to install Kallithea in a virtual
Python environment using `virtualenv`. The advantages of this method over
direct installation is that Kallithea and its dependencies are completely
contained inside the virtualenv (which also means you can have multiple
installations side by side or remove it entirely by just removing the
virtualenv directory) and does not require root privileges.
- :ref:`installation-without-virtualenv`: The alternative method of installing
a Kallithea release is using standard pip. The package will be installed in
the same location as all other Python packages you have ever installed. As a
result, removing it is not as straightforward as with a virtualenv, as you'd
have to remove its dependencies manually and make sure that they are not
needed by other packages.
.. _installation-source:
Installation from repository source
-----------------------------------
To install Kallithea in a virtualenv_ using the stable branch of the development
repository, follow the instructions below::
hg clone https://kallithea-scm.org/repos/kallithea -u stable
cd kallithea
virtualenv ../kallithea-venv
source ../kallithea-venv/bin/activate
pip install --upgrade pip setuptools
python2 setup.py develop
python2 setup.py compile_catalog # for translation of the UI
You can now proceed to :ref:`setup`.
To upgrade, simply update the repository with ``hg pull -u`` and restart the
server.
.. _installation-virtualenv:
Installing a released version in a virtualenv
---------------------------------------------
It is highly recommended to use a separate virtualenv_ for installing Kallithea.
This way, all libraries required by Kallithea will be installed separately from your
main Python installation and other applications and things will be less
problematic when upgrading the system or Kallithea.
An additional benefit of virtualenv_ is that it doesn't require root privileges.
- Assuming you have installed virtualenv_, create a new virtual environment
for example, in `/srv/kallithea/venv`, using the virtualenv command::
virtualenv /srv/kallithea/venv
- Activate the virtualenv_ in your current shell session and make sure the
basic requirements are up-to-date by running::
source /srv/kallithea/venv/bin/activate
pip install --upgrade pip setuptools
.. note:: You can't use UNIX ``sudo`` to source the ``virtualenv`` script; it
will "activate" a shell that terminates immediately. It is also perfectly
acceptable (and desirable) to create a virtualenv as a normal user.
.. note:: Some dependencies are optional. If you need them, install them in
the virtualenv too::
pip install psycopg2
pip install python-ldap
This might require installation of development packages using your
distribution's package manager.
- Make a folder for Kallithea data files, and configuration somewhere on the
filesystem. For example::
mkdir /srv/kallithea
- Go into the created directory and run this command to install Kallithea::
pip install kallithea
Alternatively, download a .tar.gz from http://pypi.python.org/pypi/Kallithea,
extract it and run::
python2 setup.py install
- This will install Kallithea together with pylons_ and all other required
python libraries into the activated virtualenv.
You can now proceed to :ref:`setup`.
.. _installation-without-virtualenv:
Installing a released version without virtualenv
------------------------------------------------
For installation without virtualenv, 'just' use::
pip install kallithea
Note that this method requires root privileges and will install packages
globally without using the system's package manager.
To install as a regular user in ``~/.local``, you can use::
pip install --user kallithea
You can now proceed to :ref:`setup`.
Upgrading Kallithea from Python Package Index (PyPI)
----------------------------------------------------
.. note::
It is strongly recommended that you **always** perform a database and
configuration backup before doing an upgrade.
These directions will use '{version}' to note that this is the version of
Kallithea that these files were used with. If backing up your Kallithea
instance from version 0.1 to 0.2, the ``my.ini`` file could be
backed up to ``my.ini.0-1``.
If using a SQLite database, stop the Kallithea process/daemon/service, and
then make a copy of the database file::
service kallithea stop
cp kallithea.db kallithea.db.{version}
Back up your configuration file::
cp my.ini my.ini.{version}
Ensure that you are using the Python virtual environment that you originally
installed Kallithea in by running::
pip freeze
This will list all packages installed in the current environment. If
Kallithea isn't listed, activate the correct virtual environment::
source /srv/kallithea/venv/bin/activate
Once you have verified the environment you can upgrade Kallithea with::
pip install --upgrade kallithea
Then run the following command from the installation directory::
paster make-config Kallithea my.ini
This will display any changes made by the new version of Kallithea to your
current configuration. It will try to perform an automerge. It is recommended
that you recheck the content after the automerge.
.. note::
Please always make sure your .ini files are up to date. Errors can
often be caused by missing parameters added in new versions.
It is also recommended that you rebuild the whoosh index after upgrading since
the new whoosh version could introduce some incompatible index changes. Please
read the changelog to see if there were any changes to whoosh.
The final step is to upgrade the database. To do this simply run::
paster upgrade-db my.ini
This will upgrade the schema and update some of the defaults in the database,
and will always recheck the settings of the application, if there are no new
options that need to be set.
.. note::
The DB schema upgrade library has some limitations and can sometimes fail if you try to
upgrade from older major releases. In such a case simply run upgrades sequentially, e.g.,
upgrading from 0.1.X to 0.3.X should be done like this: 0.1.X. > 0.2.X > 0.3.X
You can always specify what version of Kallithea you want to install for example in pip
`pip install Kallithea==0.2`
You may find it helpful to clear out your log file so that new errors are
readily apparent::
echo > kallithea.log
Once that is complete, you may now start your upgraded Kallithea Instance::
service kallithea start
Or::
paster serve /srv/kallithea/my.ini
.. note::
If you're using Celery, make sure you restart all instances of it after
upgrade.
.. _virtualenv: http://pypi.python.org/pypi/virtualenv
.. _pylons: http://www.pylonsproject.org/
|