ipython-qtconsole now uses the newapplication and config system.
Draft of new main BaseIPythonApplication.
Updates to config/application.
Fixing minor bug in config.application.
Fixing bugs in BaseIPythonApplication.
* app_name -> name.
* config_file_paths is properly typed.
allow utils.text.indent to [optionally] flatten existing indentation.
include default value in help output
also include choices in help output for Enums
exit cleanly on help/version
add `--help-all` flag, and don't print all configurables by default
update configurable test_help to match new format
raise test log_level, to prevent output during tests
add subcommand support
make config_file_paths List instead of Unicode
print usage on invalid command-line arguments
make subcommands optional
Merge branch 'subcommand' into newapp
add initialize() to BaseIPythonApplication
initialize loads config, and ensures command-line opts override those specified in config files.
reorganize default config files to match profiles as directories
Old files are left in place, until applications make the transition to newapp.
Application profiles are now directories
ProfileDir is adapted from old ClusterDir
Why do we need these to be classmethods?
These are how ProfileDirs are created. See BaseIPApp.init_profile_dir().
Never mind, too tired: I was thinking of staticmethods and got my brain wires crossed. Looks OK, sorry for the noise.
Min pointed out that this PATH variable was a similar one to the one for cluster dirs, but I think it's worth reconsidering whether we truly need multiple PATH-style search variables for IPython. @ellisonbg, did you have some specific use cases in mind that warrant adding full-blown search paths to IPython, that couldn't be addressed with manual specification of the ipython/cluster/profile dir alone at the command line?
I just think that people go crazy enough with PYTHONPATH, LD_LIBRARY_PATH and other PATH variables in various systems, often leading to much confusion. So I think it's worth that we think a bit more about this one, and see if we can get away with no additional environment variables for IPython, especially not of the search path variety.
fix _profile_changes/d typo in newapplication
prevent flags from clobbering entire config sections
previously failing test included
allow extra_args in applications
strip whitespace from configurable help
allow empty strings in kv_pattern
doing ipython foo="" would fail as invalid.
zmq kernels now started via newapp
QtConsole now uses newapp
qt font family/size configurable
The inline backend now has preliminary config support for rc and figure format.
cleanup channel names to match function not socket
wrong pull target...
The docstring has 'Inconsolata' as the Linux default, but the code has 'Anonymous Pro'. Typo?
The default in Ubuntu is just called 'Monospace'. I think we should go with that. I've never heard of Anonymous Pro, so I imagine that most distros aren't shipping it by default, right?
The fallback is still Monospace. This is the 'preferred' default. I can just make it Monospace, too, so there's no choice to be made.
Ah, OK. Your call then.
I changed it back to just Monospace. That way it's only Windows that has two defaults, which matches the platform - Consolas is the logical default, but users on XP without VS might not have it. The defaults on the other two should always exist.