A Python shell that has all configuration loaded into the namespace

[dnouri]

Not sure if trivial or useful. :-)

But here's an example session:

$ pld-shell
Welcome to the palladium shell.  I've loaded your configuration for you,
and here's the variables you have access to:
{'__mode__': 'fit',
 'dataset_loader_test': <palladium.dataset.SQL object at 0x7f745dc19828>,
 'dataset_loader_train': <palladium.dataset.SQL object at 0x7f744c2f9400>,
 'load_data_decorators': [],
 'model': LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=True),
 'model_persister': <palladium.persistence.File object at 0x7f744c2f9390>}

>>> model
LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=True)

[coveralls]

Coverage remained the same at 100.0% when pulling 9d64c32 on samsungaccelerator:feature-shell-command into 7ad551e on ottogroup:develop.

[alattner]

@dnouri: Is there any particular use case you have in mind? It might be helpful for debugging, but I am not sure if a simple pdb wouldn't do the same. Maybe some options would be nice (e.g., --fit fitting the model after loading the config or --devserver loading a model from the model persister). What do you think?

Can you use the shell history in a simple shell (not in emacs or alike)? I get something like "^[[A" in my default terminal.

docopt(list_cmd.doc, argv=argv) should be docopt(shell_cmd.doc, argv=argv)

[dnouri]

@alattner: The use case I had in mind was running a prediction (so I'd still have to call the model_persister) and then plot the predictions against the actual values.

But I'm thinking firing up a Python shell and calling get_config() is easy enough. But maybe add another issue where we add documentation on how to "load" model and environment in a shell to work with, e.g. in an IPython notebook. Pretty easy but not obvious how to do it.