Add binding for anaconda-mode-usages.
Add completion of filenames in repl.
Don't set variables unnecessarily in newer emacsen - these settings are
only needed in emacs 24.3.
Currently, when tab is pressed, spaces are inserted. It makes Python
REPL unable to recognize code indentation. We must use tab to indent.
Setting indent-tabs-mode is buffer local, so it does not affect its
value in other buffers.
When no Python executable is detected, Semantic causes an error that it cannot
retrieve include paths from the executable, causing an error that prevents
Python buffer to be switched and error is repeatedly reported whenever a Python
buffer is opened. This commit ignores the error from
semantic-python-get-system-include-path that is called when Semantic setup
itself for a Python buffer.
- Company should be enabled in Python REPL for displaying completion
candidates. Note that in Python REPL, we must use company-capf to
display the candidates from Python REPL, since in the REPL it uses
completion-at-point to display candidates from the REPL. Also, we
locally set company-minimum-prefix-length to 0, so that completion
starts immediately after the first character. We also need to set the
idle delay a bit higher (0.5) compared with Spacemacs default (0.2),
since setting it too low may cause company unable to retrieve completion
candidates.
- Anaconda should NOT be enabled in Python REPL for retrieving proper
completion candidates. This can be tested in Python REPL: when
anaconda-mode is enabled and without any prefix string before point, run
the command completion-at-point (bound to C-M-i by default), completion
candidates from Anaconda are displayed but NOT the ones from the real
REPL. So, when you import a module i.e. os and try to access the
functions in the module with "os." prefix, nothing inside the module is
displayed because Anaconda does not work in Python REPL. When
anaconda-mode is disabled, candidates in "os." are displayed properly
with company-capf. company-anaconda backend cannot be used in Python
REPL since it is buggy: it cannot retrieve any prefix before point and
narrow properly.
