On demand recomputing: the Memory class

Usecase

The Memory class defines a context for lazy evaluation of function, by storing the results to the disk, and not rerunning the function twice for the same arguments.

It works by explicitly saving the output to a file and it is designed to work with non-hashable and potentially large input and output data types such as numpy arrays.

A simple example:

First we create a temporary directory, for the cache:

>>> from tempfile import mkdtemp
>>> cachedir = mkdtemp()

We can instantiate a memory context, using this cache directory:

>>> from joblib import Memory
>>> memory = Memory(cachedir=cachedir, verbose=0)

Then we can decorate a function to be cached in this context:

>>> @memory.cache
... def f(x):
...     print('Running f(%s)' % x)
...     return x

When we call this function twice with the same argument, it does not get executed the second time, and the output gets loaded from the pickle file:

>>> print(f(1))
Running f(1)
1
>>> print(f(1))
1

However, when we call it a third time, with a different argument, the output gets recomputed:

>>> print(f(2))
Running f(2)
2

Comparison with memoize

The memoize decorator (http://code.activestate.com/recipes/52201/) caches in memory all the inputs and outputs of a function call. It can thus avoid running twice the same function, but with a very small overhead. However, it compares input objects with those in cache on each call. As a result, for big objects there is a huge overhead. More over this approach does not work with numpy arrays, or other objects subject to non-significant fluctuations. Finally, using memoize with large object will consume all the memory, where with Memory, objects are persisted to the disk, using a persister optimized for speed and memory usage (joblib.dump()).

In short, memoize is best suited for functions with “small” input and output objects, whereas Memory is best suited for functions with complex input and output objects, and aggressive persistence to the disk.

Using with numpy

The original motivation behind the Memory context was to be able to a memoize-like pattern on numpy arrays. Memory uses fast cryptographic hashing of the input arguments to check if they have been computed;

An example

We define two functions, the first with a number as an argument, outputting an array, used by the second one. We decorate both functions with Memory.cache:

>>> import numpy as np

>>> @memory.cache
... def g(x):
...     print('A long-running calculation, with parameter %s' % x)
...     return np.hamming(x)

>>> @memory.cache
... def h(x):
...     print('A second long-running calculation, using g(x)')
...     return np.vander(x)

If we call the function h with the array created by the same call to g, h is not re-run:

>>> a = g(3)
A long-running calculation, with parameter 3
>>> a
array([ 0.08,  1.  ,  0.08])
>>> g(3)
array([ 0.08,  1.  ,  0.08])
>>> b = h(a)
A second long-running calculation, using g(x)
>>> b2 = h(a)
>>> b2
array([[ 0.0064,  0.08  ,  1.    ],
       [ 1.    ,  1.    ,  1.    ],
       [ 0.0064,  0.08  ,  1.    ]])
>>> np.allclose(b, b2)
True

Using memmapping

To speed up cache looking of large numpy arrays, you can load them using memmapping (memory mapping):

>>> cachedir2 = mkdtemp()
>>> memory2 = Memory(cachedir=cachedir2, mmap_mode='r')
>>> square = memory2.cache(np.square)
>>> a = np.vander(np.arange(3)).astype(np.float)
>>> square(a)
________________________________________________________________________________
[Memory] Calling square...
square(array([[ 0.,  0.,  1.],
       [ 1.,  1.,  1.],
       [ 4.,  2.,  1.]]))
___________________________________________________________square - 0.0s, 0.0min
memmap([[  0.,   0.,   1.],
       [  1.,   1.,   1.],
       [ 16.,   4.,   1.]])

Note

Notice the debug mode used in the above example. It is useful for tracing of what is being reexecuted, and where the time is spent.

If the square function is called with the same input argument, its return value is loaded from the disk using memmapping:

>>> res = square(a)
>>> print(repr(res))
memmap([[  0.,   0.,   1.],
       [  1.,   1.,   1.],
       [ 16.,   4.,   1.]])

We need to close the memmap file to avoid file locking on Windows; closing numpy.memmap objects is done with del, which flushes changes to the disk

>>> del res

Note

If the memory mapping mode used was ‘r’, as in the above example, the array will be read only, and will be impossible to modified in place.

On the other hand, using ‘r+’ or ‘w+’ will enable modification of the array, but will propagate these modification to the disk, which will corrupt the cache. If you want modification of the array in memory, we suggest you use the ‘c’ mode: copy on write.

Warning

Because in the first run the array is a plain ndarray, and in the second run the array is a memmap, you can have side effects of using the Memory, especially when using mmap_mode=’r’ as the array is writable in the first run, and not the second.

Gotchas

  • Function cache is identified by the function’s name. Thus if you have the same name to different functions, their cache will override each-others (you have ‘name collisions’), and you will get unwanted re-run:

    >>> @memory.cache
    ... def func(x):
    ...     print('Running func(%s)' % x)
    
    >>> func2 = func
    
    >>> @memory.cache
    ... def func(x):
    ...     print('Running a different func(%s)' % x)
    
    >>> func(1)
    Running a different func(1)
    >>> func2(1)
    memory.rst:0: JobLibCollisionWarning: Possible name collisions between functions 'func' (<doctest memory.rst>:30) and 'func' (<doctest memory.rst>:28)
    Running func(1)
    >>> func(1)
    memory.rst:0: JobLibCollisionWarning: Possible name collisions between functions 'func' (<doctest memory.rst>:28) and 'func' (<doctest memory.rst>:30)
    Running a different func(1)
    >>> func2(1)
    Running func(1)
    

    Beware that with Python 2.6 lambda functions cannot be separated out:

    >>> def my_print(x):
    ...     print(x)
    
    >>> f = memory.cache(lambda : my_print(1))
    >>> g = memory.cache(lambda : my_print(2))
    
    >>> f()
    1
    >>> f()
    >>> g() 
    memory.rst:0: JobLibCollisionWarning: Cannot detect name collisions for function '<lambda>'
    2
    >>> g() 
    >>> f() 
    1
    
  • memory cannot be used on some complex objects, e.g. a callable object with a __call__ method.

    However, it works on numpy ufuncs:

    >>> sin = memory.cache(np.sin)
    >>> print(sin(0))
    0.0
    
  • caching methods: you cannot decorate a method at class definition, because when the class is instantiated, the first argument (self) is bound, and no longer accessible to the Memory object. The following code won’t work:

    class Foo(object):
    
        @mem.cache  # WRONG
        def method(self, args):
            pass
    

    The right way to do this is to decorate at instantiation time:

    class Foo(object):
    
        def __init__(self, args):
            self.method = mem.cache(self.method)
    
        def method(self, ...):
            pass
    

Ignoring some arguments

It may be useful not to recalculate a function when certain arguments change, for instance a debug flag. Memory provides the ignore list:

>>> @memory.cache(ignore=['debug'])
... def my_func(x, debug=True):
...     print('Called with x = %s' % x)
>>> my_func(0)
Called with x = 0
>>> my_func(0, debug=False)
>>> my_func(0, debug=True)
>>> # my_func was not reevaluated

Reference documentation of the Memory class

class joblib.memory.Memory(cachedir, mmap_mode=None, compress=False, verbose=1)

A context object for caching a function’s return value each time it is called with the same input arguments.

All values are cached on the filesystem, in a deep directory structure.

see Reference documentation of the Memory class

__init__(cachedir, mmap_mode=None, compress=False, verbose=1)
Parameters :

cachedir: string or None :

The path of the base directory to use as a data store or None. If None is given, no caching is done and the Memory object is completely transparent.

mmap_mode: {None, ‘r+’, ‘r’, ‘w+’, ‘c’}, optional :

The memmapping mode used when loading from cache numpy arrays. See numpy.load for the meaning of the arguments.

compress: boolean, or integer :

Whether to zip the stored data on disk. If an integer is given, it should be between 1 and 9, and sets the amount of compression. Note that compressed arrays cannot be read by memmapping.

verbose: int, optional :

Verbosity flag, controls the debug messages that are issued as functions are evaluated.

cache(func=None, ignore=None, verbose=None, mmap_mode=False)

Decorates the given function func to only compute its return value for input arguments not cached on disk.

Parameters :

func: callable, optional :

The function to be decorated

ignore: list of strings :

A list of arguments name to ignore in the hashing

verbose: integer, optional :

The verbosity mode of the function. By default that of the memory object is used.

mmap_mode: {None, ‘r+’, ‘r’, ‘w+’, ‘c’}, optional :

The memmapping mode used when loading from cache numpy arrays. See numpy.load for the meaning of the arguments. By default that of the memory object is used.

Returns :

decorated_func: MemorizedFunc object :

The returned object is a MemorizedFunc object, that is callable (behaves like a function), but offers extra methods for cache lookup and management. See the documentation for joblib.memory.MemorizedFunc.

eval(func, *args, **kwargs)

Eval function func with arguments *args and **kwargs, in the context of the memory.

This method works similarly to the builtin apply, except that the function is called only if the cache is not up to date.

clear(warn=True)

Erase the complete cache directory.

Useful methods of decorated functions

Function decorated by Memory.cache() are MemorizedFunc objects that, in addition of behaving like normal functions, expose methods useful for cache exploration and management.

class joblib.memory.MemorizedFunc(func, cachedir, ignore=None, mmap_mode=None, compress=False, verbose=1, timestamp=None)

Callable object decorating a function for caching its return value each time it is called.

All values are cached on the filesystem, in a deep directory structure. Methods are provided to inspect the cache or clean it.

Attributes :

func : callable

The original, undecorated, function.

cachedir : string

Path to the base cache directory of the memory context.

ignore : list or None

List of variable names to ignore when choosing whether to recompute.

mmap_mode : {None, ‘r+’, ‘r’, ‘w+’, ‘c’}

The memmapping mode used when loading from cache numpy arrays. See numpy.load for the meaning of the arguments.

compress : boolean, or integer

Whether to zip the stored data on disk. If an integer is given, it should be between 1 and 9, and sets the amount of compression. Note that compressed arrays cannot be read by memmapping.

verbose : int, optional

The verbosity flag, controls messages that are issued as the function is evaluated.

__init__(func, cachedir, ignore=None, mmap_mode=None, compress=False, verbose=1, timestamp=None)
Parameters :

func: callable :

The function to decorate

cachedir: string :

The path of the base directory to use as a data store

ignore: list or None :

List of variable names to ignore.

mmap_mode: {None, ‘r+’, ‘r’, ‘w+’, ‘c’}, optional :

The memmapping mode used when loading from cache numpy arrays. See numpy.load for the meaning of the arguments.

compress : boolean, or integer

Whether to zip the stored data on disk. If an integer is given, it should be between 1 and 9, and sets the amount of compression. Note that compressed arrays cannot be read by memmapping.

verbose: int, optional :

Verbosity flag, controls the debug messages that are issued as functions are evaluated. The higher, the more verbose

timestamp: float, optional :

The reference time from which times in tracing messages are reported.

call(*args, **kwargs)

Force the execution of the function with the given arguments and persist the output values.

clear(warn=True)

Empty the function’s cache.

format_call(*args, **kwds)

Returns a nicely formatted statement displaying the function call with the given arguments.

get_output_dir(*args, **kwargs)

Returns the directory in which are persisted the results of the function corresponding to the given arguments.

The results can be loaded using the .load_output method.

load_output(output_dir)

Read the results of a previous calculation from the directory it was cached in.