Module base

Given path to a file or directory, checks that the directory (or that containing the file) exists and creates it if not. This is useful to put in before you try creating or outputing to a file.

Parameters:

• is_dir (bool) - in general, we assume the input name is a file and we want to check its directory. Set is_dir=True if the input is in fact a directory

Returns:

True if the directory was created, False if it already existed.

Returns a tuple containing information about the currently raised exception. This is (type,value,traceback).

Parameters:

• str_tb (bool) - format the traceback as a string instead of returning it as an object. You'll need to do this if you want to pickle the result.

Given the full python path to a package, imports it if possible. If the import fails, raises an OptionalImportError. This is designed for loading packages that come from optional dependencies, so that a sensible error message can uniformly be displayed.

Only works with absolute paths, because relative paths would be relative to this function's module. Relative paths are explicitly disabled.

Parameters:

• dependency_name (string) - a readable name for the package being loaded.
• task (string) - a string identifying the task you're loading it for. This is for readable error output.

Works in the same way as load_optional_package, but corresponds to doing a from X import Y. Note that X and Y do not correspond directly to package_name and object_name, though. Part of X may be found in object_name: this is so that the import of the optional package can be tested before we try loading anything from it, which may generate unrelated import errors, even if the optional package is installed correctly.

E.g. instead of doing:

from nltk.model.ngram import NgramModel

we would do:

load_from_optional_package('nltk', 'model.ngram.NgramModel')

so that we can distinguish between errors loading 'nltk' due to its not being installed and errors importing 'nltk.model.ngram.NgramModel' due to bugs in the module, incorrect class name, etc.

Parameters:

• package_name - Python path to the optional package root
• object_name - remainder of the Python path to the object to be returned.

Decorator to make a method abstract. Just raises a NotImplementedError when the method is called.