Snuggs are s-expressions for Numpy

>>> snuggs.eval("(+ (asarray 1 1) (asarray 2 2))")
array([3, 3])

Snuggs wraps Numpy in expressions with the following syntax:

expression = "(" (operator | function) *arg ")"
arg = expression | name | number | string

import snuggs
snuggs.eval('(+ 1 2)')
# 3

Arrays can be created using asarray.

snuggs.eval("(* 3.5 (asarray 1 1))")
# array([ 3.5,  3.5])

Expressions can also refer by name to arrays in a local context.

snuggs.eval("(+ (asarray 1 1) b)", b=np.array([2, 2]))
# array([3, 3])

This local context may be provided using keyword arguments (e.g., b=np.array([2, 2])), or by passing a dictionary that stores the keys and associated array values. Passing a dictionary, specifically an OrderedDict, is important when using a function or operator that references the order in which values have been provided. For example, the read function will lookup the i-th value passed:

ctx = OrderedDict((
    ('a', np.array([5, 5])),
    ('b', np.array([2, 2]))
))
snuggs.eval("(- (read 1) (read 2))", ctx)
# array([3, 3])

Arithmetic (* + / -) and logical (< <= == != >= > & |) operators are available. Members of the numpy module such as asarray(), mean(), and where() are also available.

snuggs.eval("(mean (asarray 1 2 4))")
# 2.3333333333333335

snuggs.eval("(where (& tt tf) 1 0)",
    tt=numpy.array([True, True]),
    tf=numpy.array([True, False]))
# array([1, 0])

New in snuggs 1.1 are higher-order functions map and partial.

snuggs.eval("((partial * 2) 2)")
# 4

snuggs.eval('(asarray (map (partial * 2) (asarray 1 2 3)))')
# array([2, 4, 6])

Snuggs makes simple calculator programs possible. None of the optimizations of, e.g., numexpr (multithreading, elimination of temporary data, etc) are currently available.

If you're looking to combine Numpy with a more complete Lisp, see Hy:

=> (import numpy)
=> (* 2 (.asarray numpy [1 2 3]))
array([2, 4, 6])