pickler

IPython Notebook magic to make and load pickle dumps of all assignments in a cell or a line

Description

pickler provides dumpit magic function that allows to save cell computation results, and later load them instead of recomputing. The function finds every assignment in a cell, and saves/loads all affected variables. The dump filename is chosen to be a hash of cell code, so if the cell code changes, it will be recomputed.

Note that dumpit only finds assignments in the cell, so if some function called from the cell changes global variables, these changes won't be preserved or loaded.

Usage

%%dumpit
transformed_train = computationally_hard_transform(train)
transformed_test = computationally_hard_transform(test)
call_some_function()
a, b = function_that_returns_tuple()

%dumpit data = pandas.load_csv('a_large_dataset.csv')

By default, dumps are stored in pickler_dumps/ folder, to change it set pickler.FOLDER variable.