Declarative logic overview

ShapeLogic is trying to develop a framework for declarative programming that works well for the computer vision and image processing domain. This has been growing along a letter match example application. It has been a trail and error process that so far has produced these approaches:

3 different approaches to declarative programming in ShapeLogic

Here is a chronological listing of ShapeLogic's 3 different approaches declarative logic:

  1. Declarative goal driven logic engine. From ShapeLogic 0.2 till 0.8.
  2. Logic filter language. From ShapeLogic 0.8 The syntax and development of the logic language is better described in Logic language .
  3. Lazy streams. From ShapeLogic 0.9.

Result of different approaches so far

For the letter matching example, the lazy stream approach have been both simpler and more powerful than the goal driven logic engine. This is also used with the particle analyzer.

The declarative goal driven logic engine is build around an Artificial Intelligence choice tree that is traversed by hierarchical tasks. This approach was only developed until ShapeLogic 0.8, but it might work well when reasoning under uncertainty.

The logic filter language is used with both lazy streams and goal driven approach.

ShapeLogic is a toolkit, all the 3 approaches are available with unit tests.

For now development is focused on the lazy stream approach.

Lazy streams

The idea is that you have definitions of a data stream, say:

  • Natural numbers,
  • Fibonacci numbres
  • Sequence of polygons in an image
  • Sequence of sequence of polygons in an image sequence

The definition does not cause anything to be calculated, but when the data is needed it will be calculated. So it is a generalization of a lazy calculation.

Idea behind lazy streams

Lazy streams have the following features

  • You can define a lazy stream based on other lazy streams
  • They works as a kind of UNIX pipes or Legos
  • They serves as your query construct, you can directly query them

Lazy streams in letter match example and particle analyzer

Stream used in letter match example:

  • Polygon stream, this is generated from either and SVG file or and image
  • Cleaned up polygon stream, this can be done in different ways
  • A lot of feature streams:
    • hole count
    • point count
    • number of end point in lower left corner
    • number of t junction in upper right corner
  • Each letter is an And stream build of a feature stream and a predicate, for letter A:
    • hole count == 1
    • point count > 5
    • number of end point in lower left corner == 1
    • number of t junction on the right side == 1
  • The Letter match stream is an XOr stream of the letter stream.

Technical details about lazy streams

The streams in ShapeLogic are different than the streams in Haskell, Oz, Scheme and Scala. These are all based on data structures similar to a LISP list.

Immutable stream in Java List

In ShapeLogic the streams are assumed to be immutable. The data is put in a normal Java List.

The advantage of Java List
  • Faster access to data
  • Works better with Java
Advantages of LISP list / disadvantage of Java List
  • In a LISP list sub lists are also lists
  • This enables elegant functional style recursive programming

Using streams for concurrent programming

Streams can also support parallel or concurrent programming, which is important with the CPU intensive operations in image processing and computer vision. Especially with the advent of cheap multi processor machines.

Example: Find polygons in a stack of images

You define a lazy data stream for this and set a stream property

randomAccess = true

This indicates that individual elements can be calculated independently. The factory creating the stream could create a parallel version of the stream and assign each operation its own thread.

Note that the result would be a stream of polygons for each image.

Logic filter language

The Logic filter language allows you to do Boolean combinations of filters with parameters

If you need to do more complex boolean combinations of filter on polygons try the following:

"polygon.filter('PointRightOfFilter(0.3) && PointAboveFilter(0.3)')"
"polygon.filter('PointOfTypeFilter(PointType.T_JUNCTION) && PointLeftOfFilter(0.5)').size()"

These will filter:

  • Points that are in in the right 30% and top 30% of the bounding box for the polygon.
  • Points that are in in the left 50% of the bounding box for the polygon and having the annotation T Junction.

Logical operators

Currently ShapeLogic is using the standard Java, C, C++ notation:

  And use &&
  Or use ||
  Not use !

Ideas for future declarative programming constructs

IoC, Inversion of Control

There is an external definition that determine what type of objects should be created, and how they should be instantiated and create and instantiate all dependent object.

Once the system become more complex this will probably be essential.

The system is currently simple so now this would only add complexity, and is not used yet.