Lambda Content Filter

Concept

With this example you’ll learn how to implement a generic ContentFilter to create ContentFilteredTopics or QueryConditions using lambda expressions. The code is completely generic and you can easily include it in your applications.

Connext DDS provides built-in content filters, most notably the SQL filter. This filter parses ("compiles") a SQL string expression to create a set of steps that, when executed on each data sample, determine if it passes the filter or not. The SQL filter main advantage is its ease of use and flexibility: you just need to type a SQL expression. However it is prone to runtime errors—Connext DDS detects syntactic or semantic errors in the expression when it parses it—and relatively inefficient. (For more information about the SQL filter see the Content Filtered Topic example.)

The lambda filter in this example provides similar ease of use and even greater flexibility—the lambda function can run arbitrary code. Even better, the lambda filter is defined by actual C++ code, so the compiler will detect errors and it can be more efficient. It has one disadvantage, though. DataWriters can discover the SQL filter expression used by a DataReader and automatically filter samples before delivering them. With the lambda filter we need to explicitly register the same filter on the publication side if we want writer-side filtering.

The lambda filter is a custom content filter.  For more information, see the Custom Content Filters example)

Example Description

  • LambdaFilter.hpp implements the generic LambdaFilter class (an implementation of ContentFilter) and two helper functions: create_lambda_filter() and create_lambda_cft().

  • LambdaFilterExample_subscriber.cxx implements a subscribing application that creates a DataReader with a ContentFilteredTopic for the Stock type.
    This is how you create a DataReader with a lambda-based ContentFilteredTopic:

    dds::domain::DomainParticipant participant(domain_id);
    
    dds::topic::Topic<Stock> topic(participant, "Example Stock");
    
    auto lambda_cft = create_lambda_cft<Stock>(
        "stock_cft",
        topic,
        [](const Stock& stock)
        {
            return stock.symbol() == "GOOG" || stock.symbol() == "IBM";
        }
    );
    
    dds::sub::DataReader<Stock> reader(
        dds::sub::Subscriber(participant), lambda_cft);
    
  • LambdaFilterExample_publisher.cxx implements an application that publishes Stock samples each with one of four random symbols (two of which pass the filter). It also registers the lambda filter to perform writer-side filtering.

You can simply include LambdaFilter.hpp in your own Connext DDS application to start using your own lambda filters.

Download Example

Browse Example

Or browse it in GitHub.

Languages: