About

A pure-Python (2 and 3) package for manipulating:

• Binary decision diagrams (BDDs).
• Multi-valued decision diagrams (MDDs).

as well as Cython bindings to the C libraries:

• CUDD
• BuDDy

These bindings expose almost identical interfaces as the Python implementation. The intended workflow is:

• develop your algorithm in pure Python (easy to debug and introspect),
• use the bindings to benchmark and deploy

Your code remains the same.

An ordered BDD is represented using dictionaries for the successors, unique table, and reference counts. Nodes are positive integers, and edges signed integers. A complemented edge is represented as a negative integer. Garbage collection uses reference counting.

Contains:

• All the standard functions defined, e.g., by Bryant.
• Rudell's sifting algorithm for variable reordering.
• Reordering to obtain a given order.
• Quantified Boolean expression parser that creates BDD nodes.
• Pre/Image computation (relational product).
• Renaming variables to their neighbors.
• Conversion from BDDs to MDDs.
• Conversion functions to networkx and pydot graphs.
• BDDs have methods to dump and load them as nested dicts using pickle.
• BDDs dumped by CUDD can be loaded using a PLY-based parser for the header, and a fast simple by-line parser for the main body of nodes.
• Cython bindings to CUDD
• Cython bindings to BuDDy

Documentation

In the Markdown file doc.md.

Examples

Two interfaces are available:

• convenience: the module dd.autoref wraps dd.bdd and takes care of reference counting, using __del__.

• "low level": the module dd.bdd requires that the user in/decrement the reference counters associated with nodes that are used outside of a BDD.

Automated reference counting

The module dd.autoref wraps the pure-Python BDD implementation in dd.bdd. A Function object wraps a node and decrements its reference count when disposed by Python's garbage collector:

from dd.autoref import BDD, Function

bdd = BDD()
[bdd.add_var(var) for var in ['x', 'y']
u = bdd.add_expr('x -> y')

# alternative
x = bdd.var('x')
not_x = ~ x
y = bdd.var('y')
v = not_x | y
assert u == v

CUDD

The interface to CUDD in dd.cudd looks similar to dd.autoref, including automated reference counting:

from dd import cudd

bdd = cudd.BDD()
[bdd.add_var(var) for var in ['x', y'']]
u = bdd.add_expr('\E x, y: x & y')
assert u == bdd.true, u

# longer alternative
xy = bdd.add_expr('x & y')
u = bdd.exist(['x', 'y'], xy)
assert u == bdd.true, u

Reference counting by the user

The pure-Python module dd.bdd can be used directly, which allows access more extensive than dd.autoref. The n variables in a dd.bdd.BDD are ordered from 0 (top level) to n-1 (bottom level). The terminal node 1 is at level n.

from dd.bdd import BDD

ordering = dict(x=0, y=1)
bdd = BDD(ordering)
bdd.add_var('z')

Boolean expressions can be added with the method BDD.add_expr:

u = bdd.add_expr('x | y')
v = bdd.add_expr('!x | z')
w = bdd.apply('and', u, v)
w = bdd.apply('&', u, v)
r = bdd.apply('->', u, w)

Garbage collection is triggered either explicitly by the user, or when invoking the reordering algorithm. If we invoked garbage collection next, then the nodes u, v, w would be deleted. To prevent this from happening, their reference counts should be incremented. For example, if we want to prevent w from being collected as garbage, then

bdd.incref(w)

To decrement the reference count:

bdd.decref(w)

The more useful functions in dd.bdd are: rename, image, preimage, reorder, to_nx, to_pydot.

Use the method BDD.dump to write a BDD to a pickle file, and BDD.load to load it back. A CUDD dddmp file can be loaded using the function dd.dddmp.load.

Examples of how dd can be used to implement symbolic algorithms can be found in the omega package.

Installation

pure-Python

Recommended to use pip, because the latest version will install dependencies before dd:

pip install dd

Otherwise:

python setup.py install

If you use the latter, remember to install ply before dd. If ply is absent, then the parser tables will not be cached, affecting performance. For graph layout with pydot, graphviz needs to be installed.

Cython bindings

dd fetching CUDD

By default, the package installs only the Python modules. You can select to install either or both Cython extensions by the setup.py options --cudd and --buddy.

Pass --fetch to setup.py to tell it to download, unpack, and make CUDD. If building from the repository, then first install cython. For example:

pip install cython  # not needed if building from PyPI distro
python setup.py install --fetch --cudd

These options can be passed to pip too, using the --install-option in a requirements file, for example:

dd >= 0.1.1 --install-option="--fetch" --install-option="--cudd"

The command line behavior of pip is currently different, so

pip install --install-option="--fetch" dd

will propagate option --fetch to dependencies, and so raise an error.

User fetching build dependencies

If you build and install CUDD or BuDDy yourself, then ensure that:

• the header files and libraries of either CUDD or BuDDy are present, and
• suitable compiler, include, linking, and library flags are passed, either by setting environment variables prior to calling pip, or by editing the file download.py.

Tests

Require nose and the extras. Run with:

cd tests/
nosetests

If the extension module dd.cudd has not been compiled and installed, then the CUDD tests will fail.

License

BSD-3, see file LICENSE.