Filter Junctions

Introduction

The concept of Filter Sequences is somewhat restrictive in the sense that it only allows for a single input. Some filters however, require more than one dataset, e.g. one that is un- and one that has beeen prefiltered by another filtering algorithm.
For such cases we designed a subclass of FilterSequence that still only takes one stream of input data, but can autonomously divide it into a set number of input partitions.
This then in turn allows for a new class of filters: FilterJunctors. Junctors work like filters, except that they

• are stored in junctions and not in sequences,
• can take any number of input data sets.

You can find a list of available junctors here.
FilterJunctions are restricted to symmetrical input, i.e. each input data set must be of identical size and must use the same indexing.
If this is not the case, you must use an Asymmetrical Filter Junction

Configuring Junctions and Junctors

The configuration process of Junctions is identical to that of sequences, except that they allow for more than one specified input identifier in their input XML attribute.
This distinction is made automatically: Specifying more than one input will let the filtering system conclude that you wish to use a junction.

Junctors are configured exactly the way filters are. Note that junctors can only be stored in junctions and not in sequences, while filters can only be part of a sequence, not a junction.

Future Plans

As of now, dynamic junctors are not yet implemented.