fancycompleter
is a module to improve your experience in Python by
adding TAB completion to the interactive prompt. It is an extension of
the stdlib's
rlcompleter module.
Its best feature is that the completions are displayed in different colors, depending on their type:
In the image above, strings are shown in green, functions in blue,
integers and boolean in yellows, None
in gray, types and classes in
fuchsia. Everything else is plain white.
fancycompleter
is compatible with Python 3. However, by default colors
don't work on Python 3, see the section How do I get
colors? for details.
- To save space on screen,
fancycompleter
only shows the characters "after the dot". By contrast, in the example aboverlcompleter
shows everything prepended by"sys."
. - If we press
<TAB>
at the beginning of the line, a real tab character is inserted, instead of trying to complete. This is useful when typing function bodies or multi-line statements at the prompt. - Unlike
rlcompleter
,fancycompleter
does complete expressions containing dictionary or list indexing. For example,mydict['foo'].<TAB>
works (assuming thatmydict
is a dictionary and that it contains the key'foo'
, of course :-)). - Starting from Python 2.6, if the completed name is a callable,
rlcompleter
automatically adds an open parenthesis(
. This is annoying in case we do not want to really call it, sofancycompleter
disable this behaviour.
First, install the module with pip
or easy_install
:
$ pip install fancycompleter
Then, at the Python interactive prompt:
>>> import fancycompleter
>>> fancycompleter.interact(persist_history=True)
>>>
If you want to enable fancycompleter
automatically at startup, you can
add those two lines at the end of your
PYTHONSTARTUP
script.
If you do not have a PYTHONSTARTUP
script, the
following command will create one for you in ~/python_startup.py
:
$ python -m fancycompleter install
On Windows, install
automatically sets the PYTHONSTARTUP
environment
variable. On other systems, you need to add the proper command in
~/.bashrc
or equivalent.
Note: depending on your particular system, interact
might need to
play dirty tricks in order to display colors, although everything should
"just work". In particular, the call to interact
should be the last
line in the startup file, else the next lines might not be executed. See
section What is really going on? for
details.
If you are using PyPy, you can stop reading now, as fancycompleter
will work out of the box.
If you are using CPython on Linux/OSX and you installed
fancycompleter
with pip
or easy_install
, they automatically
installed pyrepl
as a requirement, and you should also get colors out
of the box. If for some reason you don't want to use pyrepl
, you
should keep on reading.
By default, in CPython line input and TAB completion are handled by GNU
readline (at least
on Linux). However, readline
explicitly strips escape sequences from
the completions, so completions with colors are not displayed correctly.
There are two ways to solve it:
- (suggested) don't use
readline
at all and rely on pyrepl- use a patched version of
readline
to allow colors
By default, fancycompleter
tries to use pyrepl
if it finds it. To
get colors you need a recent version, >= 0.8.2.
Starting from version 0.6.1, fancycompleter
works also on Windows,
relying on pyreadline. At the
moment of writing, the latest version of pyreadline
is 2.1, which does
not support colored completions; here is the pull
request which adds
support for them. To enable colors, you can install pyreadline
from
this fork using the following
command:
pip install --upgrade https://github.com/antocuni/pyreadline/tarball/master
If you are using Python 3, pyrepl
does not work, and thus is not
installed. Your only option to get colors is to use a patched
readline
, as explained below.
This method is not really recommended, but if you really want, you can
use use a patched readline: you can find the patches in the misc/
directory:
- for readline-5.2
- for readline-6.0
You can also try one of the following precompiled versions, which has
been tested on Ubuntu 10.10: remember to put them in a place where the
linker can find them, e.g. by setting LD_LIBRARY_PATH
:
Once it is installed, you should double-check that you can find it, e.g.
by running ldd
on Python's readline.so
module:
$ ldd /usr/lib/python2.6/lib-dynload/readline.so | grep readline
libreadline.so.6 => /home/antocuni/local/32/lib/libreadline.so.6 (0x00ee7000)
Finally, you need to force fancycompleter
to use colors, since by
default, it uses colors only with pyrepl
: you can do it by placing a
custom config file in ~/.fancycompleterrc.py
. An example config file
is
here
(remind that you need to put a dot in front of the filename!).
To customize the configuration of fancycompleter, you need to put a file
named .fancycompleterrc.py
in your home directory. The file must
contain a class named Config
inheriting from DefaultConfig
and
overridding the desired values.
The default and preferred way to get colors is to use pyrepl
. However,
there is no way to tell CPython to use pyrepl
instead of the built-in
readline at the interactive prompt: this means that even if we install
our completer inside pyrepl's readline library, the interactive prompt
won't see it.
The issue is simply solved by avoiding to use the built-in prompt: instead, we use a pure Python replacement based on code.InteractiveConsole. This brings us also some niceties, such as the ability to do multi-line editing of the history.
The console is automatically run by fancycompleter.interact()
,
followed by sys.exit()
: this way, if we execute it from the script in
PYTHONSTARTUP
, the interpreter exits as soon as we finish the use the
prompt (e.g. by pressing CTRL-D, or by calling quit()
). This way, we
avoid to enter the built-in prompt and we get a behaviour which closely
resembles the default one. This is why in this configuration lines after
fancycompleter.interact()
might not be run.
Note that if we are using readline
instead of pyrepl
, the trick is
not needed and thus interact()
will simply returns, letting the
built-in prompt to show up. The same is true if we are running PyPy, as
its built-in prompt is based on pyrepl anyway.