3.5.2. Filters and Predicates

Filter

Operator (predicate)

Output sequence length

\(a' = \filter{p}\ a\)

\(p: A \rightarrow \text{Boolean}\)

\(|a'| \le |a|\)

Any user-defined algorithm or output sink may be configured to operate on data that satisfy a Boolean condition or predicate. The act of restricting the invocation of a function to data that satisfy a predicate is known as filtering. To filter data as presented to a given algorithm, one or more predicates must be specified in a filter clause. Phlex will not schedule a predicate for execution if it is not bound to a filter.

Todo

Define filter clause. Many algorithms can specify the same predicate in their filter clauses without executing the predicate multiple times.

Phlex will only schedule a filter for execution if there is at least one non-filter algorithm or output sink downstream of it. Predicates can be evaluated on (e.g.) run-level data-product sets and applied to algorithms that process data from data-product sets that are subsets of the run (e.g. events).

Operator

Allowed signature

\(p\)

bool function_name(P1, Pn..., Rm...) [qualifiers];

3.5.2.1. Registration interface

class tracks { ... };

bool select_tracks(Tracks const& ts) { ... }

PHLEX_REGISTER_ALGORITHMS(m, config)
{
  auto selected_data_scope = config.get<std::string>("data_scope");

  predicate("good_tracks", select_tracks, concurrency::unlimited)
    .sequence( <input_sequence_spec> );
}