Filter Specification

Filters take data from one stream, performs an operation on the data and optionally writes the output to another stream.

While filters are essential to most Stroom setups, there is very little to the specification. If runs continuously and thus needs to maintain its state, it should do so in the ".state" metadata stream of the output stream, not the input stream. This allows for multiple filter processes to run independently of one source stream while maintaining their state separately.

The current reference implementation stores the state of its filter processes as a very small object with the index of the document that was last processed by the filter.

{"i":923898}

However, since the state stream for a filter process is not intended to be read by any other processes - there are no further requirements.

For samples of different filter implementations, see the filter section of the reference implementation.