![]() |
Pylint User Manual |
Pylint User Manual This document is meant to be the reference user manual for Pylint. This is a
work in progress so some sections or parts may be missing (sometimes marked by a
XXX). If you think it's lacking some important information, please talk about
it on the python-projects mailing list (see the Mailing lists section for
more information about the list).
Pylint is a tool that checks for errors in python code, tries to enforce a
coding standard and looks for smelling code. This is similar but nevertheless
different from what pychecker provides, especially since pychecker explicitely
does not bother with coding style. The default coding style used by pylint is
close to PEP 008 (aka Guido's style guide). For more information about
code smells, refer to Martin Fowler's refactoring book
One important thing to note is that Pylint isn't smarter than you are: it may
warn you about things that you have conscientiously done. That's for example
because it tries to detect things that may be dangerous in a context, but maybe
not in others, or because it checks for some things that you don't care
about. Generally, you shouldn't expect pylint to be totally quiet about your
code, so don't necessarily be alarmed if it gives you a hell lot of messages for
your proudly(XXX) project ;)
Pylint will display a number of messages as it analyzes the code, as well as
some statistics about the number of warnings and errors found in different
files. The messages are classified under various categories such as errors and
warnings (more below). If you run pylint twice, it will display the statistics
from the previous run together with the ones from the current run, so that you
can see if the code has improved or not.
Last but not least, the code is given an overall mark, based on the number an
severity of the warnings and errors. This has proven to be very motivating for
programmers.
Pylint requires the latest logilab-astng and logilab-common
packages. It should be compatible with any python version greater than
2.2.0 (python 2.2 users will have to install the optik package).
The source tarball is available at ftp://ftp.logilab.fr/pub/pylint.
You may apt-get a debian package by adding
deb ftp://ftp.logilab.org/pub/debian unstable/
to your /etc/apt/sources.list file. Pylint is also available in the
standard Debian distribution (but add our public debian repository
anyway if you want to get the latest releases and upgrades earlier)
Contributed RPM packages for pylint and logilab-common are available at
ftp://ftp.nest.pld-linux.org/test.
Pylint is also available in Gentoo, Fedora 4, Ubuntu, FreeBSD, Darwin
(and maybe other, if si drop us a note please!).
From the source distribution, extract the tarball, go to the extracted
directory and simply run
python setup.py install
You'll have to install dependancies in a similar way.
Windows users may get valuable information about pylint installation on
this page.
On Windows, once you have installed pylint, the command line usage is
pylint.bat [options] module_or_package
But this will only work if pylint.bat is either in the current
directory, or on your system path. (setup.py install install python.bat
to the Scripts subdirectory of your Python installation -- e.g.
C:Python24Scripts.) You can do any of the following to solve this:
To effect (2), simply append the appropriate directory name to the PATH=
statement in autoexec.bat. Be sure to use the Windows directory
separator of ';' between entries. Then, once you have rebooted (this is
necessary so that the new path statement will take effect when
autoexec.bat is run), you will be able to invoke PyLint with
pylint.bat on the command line.
(3) is the best solution. Once done, you can call pylint at the command
line without the .bat, just as do non-Windows users by typing:
pylint [options] module_or_package
To effect option (3), simply create a plain text file pylint.bat with
the single line:
C:\PythonDirectory\Scripts\pylint.bat
(where PythonDirectory is replaced by the actual Python installation
directory on your system -- e.g. C:Python24Scriptspylint.bat).
Pylint is meant to be called from the command line. The usage is
pylint [options] module_or_package
You should give pylint the name of a Python package or module. Pylint
will import this package or module, so you should pay attention to
your PYTHONPATH, since it is a common error to analyze an
installed version of a module instead of the development version.
It is also possible to analyze python files, with a few
restriction. The thing to keep in mind is that pylint will try to
convert the file name to a module name, and only be able to process
the file if it succeeds.
pylint mymodule.py
should always work since the current working
directory is automatically added on top of the python path
pylint directory/mymodule.py
will work if "directory" is a python package (i.e. has an __init__.py
file) or if "directory" is in the python path.
For more details on this see the FAQ_.
You can also start a thin gui around pylint (require TkInter) by
typing
pylint-gui
This should open a window where you can enter the name of the package
or module to check, at pylint messages will be displayed in the user
interface.
The default format for the output is raw text. But passing pylint the
--html option will produce an HTML document.
There are several sections in pylint's output.
For each python module,
pylint will first display a few '*' characters followed by the name
of the module. Then, a number of messages with the following
format:
MESSAGE_TYPE: LINE_NUM:[OBJECT:] MESSAGE
You can get another output format, useful since it's recognized by
most editors or other development tools using the --parseable=y
option.
The message type can be:
Sometimes the line of code which caused the error is displayed with
a caret pointing to the error. This may be generalized in future
versions of pylint.
Example (extracted from a run of pylint on itself...):
************* Module pylint.checkers.format
W: 50: Too long line (86/80)
W:108: Operator not followed by a space
print >>sys.stderr, 'Unable to match %r', line
^
W:141: Too long line (81/80)
W: 74:searchall: Unreachable code
W:171:FormatChecker.process_tokens: Redefining built-in (type)
W:150:FormatChecker.process_tokens: Too many local variables (20/15)
W:150:FormatChecker.process_tokens: Too many branchs (13/12)
Following the analysis message, pylint will display a set of reports,
each one focusing on a particular aspect of the project, such as number
of messages by categories, modules dependancies...
For instance, the metrics report displays summaries gathered from the
current run.
Also, a global evaluation for the code is computed, and an
optional witty comment is displayed (if --comment=y was
specified on the command line).
First of all, we have two basic (but useful) options.
Pylint is architectured around several checkers. By default all
checkers are enabled. You can disable a specific checker by specifying
--enable-<checker>=n, or disable all checkers using
--disable-all and afterwards enable specific checkers with
--enable-<checker>=y. See the list of available features for a
description of provided checkers with their functionalities.
Each checker has some specific options, which can take either a yes/no
value, an integer, a python regular expression, or a comma separated
list of values (which are generally used to override a regular
expression in special cases). For a full list of options, use --help
Specifying all the options suitable for your setup and coding
standards can be tedious, so it is possible to use a rc file to
specify the default values. Pylint looks for /etc/pylintrc and
~/.pylintrc. The --generate-rcfile option will generate a
commented configuration file according to the current configuration on
standard output and exit. You can put other options before this one to
use them in the configuration, or start with the default values and
hand tune the configuration.
Other useful global options include:
What pylint says is not to be taken as gospel. While getting as
few false positives for errors as possible is a goal for us -- and
python makes it hard enough, it is not the case for warnings.
Quoting Alexandre
My usage pattern for pylint is to generally run pylint -e quite often to
get stupid errors flagged before launching an application (or before
comitting). I generally run pylint with all the bells and whistles
activated some time before a release, when I want to cleanup the code.
And when I do that I simply ignore tons of the false warnings (and I
can do that without being driven mad by this dumb program which is not
smart enough to understand the dynamicity of Python because I only run
it once or twice a week in this mode)
Quoting Marteen Ter Huurne
In our project we just accepted that we have to make some modifications in our
code to please PyLint:
The effort is worth it, since PyLint helps us a lot in keeping the code clean
and finding errors early. Although most errors found by PyLint would also be
found by the regression tests, by fixing them before committing, we save time.
And our regression tests do not cover all code either, just the most complex
parts.
You think you have found a bug in Pylint? Well, this may be the case
since Pylint is under development. Please take the time to send a bug
report to python-projects@logilab.org if you've not found it already reported on
the tracker page. This mailing list is also a nice place to
discuss Pylint issues, see below for more information about pylint's related
lists.
You can check for already reported bugs, planned features on pylint's tracker
web page: http://www.logilab.org/project/name/pylint
Notice that if you don't find something you have expected in pylint's
tracker page, it may be on the tracker page of one of its dependancies, namely
astng and common:
Use the python-projects@logilab.org mailing list for anything related
to Pylint. This is in most cases better than sending an email directly
to the author, since others will benefit from the exchange, and you'll
be more likely answered by someone subscribed to the list. This is a
moderated mailing list, so if you're not subscribed email you send will have to
be validated first before actually being sent on the list.
You can subscribe to this mailing list at
http://lists.logilab.org/mailman/listinfo/python-projects
Archives are available at
http://lists.logilab.org/pipermail/python-projects/
If you prefer speaking french instead of english, you can use the
generic forum-fr@logilab.org mailing list:
Notice though that this list has a very low traffic since most pylint related
discussions are done on the python-projects mailing list.
To be written...
To be written...
An example available from the examples directory:
"""pylint option block-disable-msg"""
__revision__ = None
class Foo(object):
"""block-disable-msg test"""
def __init__(self):
pass
def meth1(self, arg):
"""this issues a message"""
print self
def meth2(self, arg):
"""and this one not"""
# pylint: disable-msg=W0613
print self\
+ "foo"
def meth3(self):
"""test one line disabling"""
# no error
print self.bla # pylint: disable-msg=E1101
# error
print self.blop
def meth4(self):
"""test re-enabling"""
# pylint: disable-msg=E1101
# no error
print self.bla
print self.blop
# pylint: enable-msg=E1101
# error
print self.blip
def meth5(self):
"""test IF sub-block re-enabling"""
# pylint: disable-msg=E1101
# no error
print self.bla
if self.blop:
# pylint: enable-msg=E1101
# error
print self.blip
else:
# no error
print self.blip
# no error
print self.blip
def meth6(self):
"""test TRY/EXCEPT sub-block re-enabling"""
# pylint: disable-msg=E1101
# no error
print self.bla
try:
# pylint: enable-msg=E1101
# error
print self.blip
except UndefinedName: # pylint: disable-msg=E0602
# no error
print self.blip
# no error
print self.blip
def meth7(self):
"""test one line block opening disabling"""
if self.blop: # pylint: disable-msg=E1101
# error
print self.blip
else:
# error
print self.blip
# error
print self.blip
def meth8(self):
"""test late disabling"""
# error
print self.blip
# pylint: disable-msg=E1101
# no error
print self.bla
print self.blop
To be written...
To be written...
You can find some simple examples in the examples
directory of the distribution (custom.py and custom_raw.py). I'll try to
quickly explain the essentials here.
First, there are two kinds of checkers :
* raw checkers, which are analysing each module as a raw file stream
* ast checkers, which are working on an ast representation of the module
The ast representation used is an extension of the one provided with the
standard python distribution in the compiler package. The extension
adds additional information and methods on the tree nodes to ease
navigation and code introspection.
An AST checker is a visitor, and should implement
visit_<lowered class name>
leave_<lowered class name>
methods for the nodes it's interested in. To get description of the different
classes used in an ast tree, look at the compiler.ast documentation.
Checkers are ordered by priority. For each module, pylint's engine:
Notice that the source code is probably the best source of
documentation, it should be clear and well documented. Don't hesitate to
ask for any information on the python-projects mailing list.
All our software is developped using the mercurial version control
system. This is a very cool distributed vcs and its usage is very similar to
other ones such as cvs or subversion (though the distributed feature introduced
some different usage patterns). See mercurial home page for installation on
your computer and basic usage. Note that it's very easy to send us patches using
hg email command ;).
You can get the in-development pylint source code from our public mercurial
repository:
The same is true for pylint dependancies (if you use pylint code from the
repository, you should usually use code from the repository as well for astng
and common):
Pylint is integrated in the following editors/IDEs:
To use pylint from within vim, see
http://www.gonzo.kiev.ua/projects/pylint.vim
To use pylint from within komodo, see
http://mateusz.loskot.net/2006/01/15/running-pylint-from-komodo/
To use pylint from within gedit, see
http://live.gnome.org/Gedit/PylintPlugin
The following projects are known to use pylint to help develop better
code:
Also notice that the CheeseCake kwalitee reporting tool uses pylint to
analyze the source code.
Question:
Is it possible to give file as argument to pylint, instead of module ?
Answer:
pylint expects the name of a package or module as argument. As a convenience,
you can give to it a file name if it's possible to guess a module name from
the file's path, using the python path. Some examples :
"pylint mymodule.py" should always works since the current working
directory is automatically added on top of the python path
"pylint directory/mymodule.py" will work if "directory" is a python
package (i.e. has an __init__.py file) or if "directory" is in the
python path.
"pylint /whatever/directory/mymodule.py" will work if either:
Question:
I'm using psyobj from psyco and get a lot of spurious "unused variables
messages". Is it normal ?
Answer:
Yes. That's actually due to a bug in psyco, making the locals()
function for objects inheriting from psyobj returning an empty
dictionary. For the moment, the only way to fix this is to use the
PYLINT_IMPORT environment variable to not use psyco during pylint
checking. Sample code
import os
try:
if os.environ.has_key('PYLINT_IMPORT'):
raise ImportError()
from psyco.classes import psyobj
except ImportError:
class psyobj:
pass
NOTICE: this problem should not occurs with pylint >= 0.5 since from
this version pylint is not looking anymore for information in living
objects (i.e. it doesn't anymore import analysed modules)
Question:
I've a function / method which is a callback where I do not have any
control on received argument, and pylint is complaining about unused
arguments. What can I do to avoid those warnings ?
Answer:
prefix (ui) the callback's name by cb_, as in cb_onclick(...). By
doing so arguments usage won't be checked. Another solution is to
use one of the name defined in the "dummy-variables" configuration
variable for unused argument ("_" and "dummy" by default).
Question:
When is pylint considering a class as an interface ?
Answer:
A class is considered as an interface if there is a class named
"Interface" somewhere in it ancestor's tree.
Question:
When is pylint considering that a class is implementing a given
interface ?
Answer:
Pylint is using the Zope 2 interfaces conventions, and so is
considering that a class is implementing interfaces listed in its
__implements__ attribute.
Question:
When is pylint considering a class as an abstract class ?
Answer:
A class is considered as an abstract class if at least one of its
methods is doing nothing but raising NotImplementedError
Question:
Is there some way to disable some message for a particular module
only ?
Answer:
Yes, you can disable or enable (globally disabled) message at the
module level by adding the corresponding option in a comment at the
top of the file:
# pylint: disable-msg=W0401, E0202
# pylint: enable-msg=C0302
Question:
I have a mixin class relying on attributes of the mixed class, and I
would like to not have the "access to undefined member" message on
this class. Is it possible ?
Answer:
Yes :o) To do so you have to set the ignore-mixin-members option to
"yes" (this is the default value) and to name your mixin class with
a name which ends with "mixin" (whatever case)
Question:
Is it possible to locally disable a particular message for a block
of code or for a single line of code ?
Answer:
Yes, this feature has been added in pylint 0.11. This may be done by
adding "#pylint: disable-msg=W0123,E4567" at the desired block level
or at the end of the desired line of code
Question:
Where is the persistent data stored to make comparison between
two successive runs ?
Answer:
Analysis data are stored as pickle file in a directory which is
localized using the following rules:
Question:
How can I know the option name (for pylintrc) corresponding to a
specific command line option ?
Answer:
You can always generate a sample pylintrc file with --generate-rcfile
Every option present on the command line before this will be included in
the rc file
For example:
pylint --disable-msg=W0702,C0103 --class-rgx='[A-Z][a-z]+' --generate-rcfile
Question:
pychecker has no problem finding the imports and reporting on problems with
them, while pylint seems unable to deal with the same imports. Why ?
Answer:
pychecker and pylint use different approaches. pychecker
imports the modules and rummages around in the result, hence it sees my
mangled sys.path. pylint doesn't import any of the candidate modules and
thus doesn't include any of import's side effects (good and bad). It
traverses an AST representation of the code.
Unknown target name: "faq".
|