Microsecond Latency Rules Engine with RoaringBitmap

Implementing a rules engine can shorten development time and remove a lot of tedious if statements from your business logic. Unfortunately they are almost always slow and often bloated. Simple rules engines can be implemented by assigning integer salience to each line in a truth table, with rule resolution treated as an iterative intersection of ordered sets of integers. Implemented in terms of sorted sets, it would be remiss not to consider RoaringBitmap for the engine’s core. The code is at github.

Classification Table and Syntax

This rules engine builds on the simple idea of a truth table usually used to teach predicate logic and computer hardware. Starting with a table and some attributes, interpreting one attribute as a classification, we get a list of rules. It is trivial to load such a table from a database. Since classifications can overlap, we prioritise by putting the rules we care about most – or the most salient rules – at the top of the table. When multiple rules match a fact, we take the last in the set ordered by salience. So we don’t always have to specify all of the attributes to get a classification, we can rank attributes by their importance left to right, where it’s required that all attributes to the left of a specified attribute are also specified when matching a fact against a rule set.

It’s possible to define rules containing wildcards. Wildcard rules will match any query (warning: if these are marked as high salience they will hide more specific rules with lower salience). It’s also possible to specify a prefix with a wildcard, which will match any query that matches at least the prefix.

Below is an example table consisting of rules for classification of regional English accents by phonetic feature.

English Accent Rules

thought cloth lot palm plant bath trap accent
/ɔ/ /ɒ/ /ɑ/ /ɑː/ /ɑː/ /ɑː/

/æ/ Received Pronunciation (UK)
/ɔ/ /ɔ/ /ɑ/ /ɑ/ /æ/ /æ/

/æ/ Georgian (US)
/ɑ/ /ɑ/ /ɑ/ /ɑ/ /æ/ /æ/

/æ/ Canadian
* * /ɑ/ /ɑ/ /æ/ /æ/

/æ/ North American
* * * * * *

/æ/ Non Native
* * * * * *

* French


In the example above, the vowel sounds used in words differentiating speakers of several English accents are configured as a classification table. The accent column is the classification of any speaker exhibiting the properties specified in the six leftmost columns. UK Received Pronunciation is the most specific rule and has high salience, whereas various North American accents differ from RP in their use of short A vowels. A catch all for North American accents would wild card the sounds in thought and caught (contrast Boston pronunciations with Texas). So long as trap has been pronounced with a short A (which all English speakers do), and no other rule would recognise the sounds used in the first six words, the rule engine would conclude the speaker is using English as a second language. If not even the word trap is recognisable, then the speaker is probably unintelligible, or could be French.


A rule with a given salience can be represented by creating a bitmap index on salience by the attribute values of the rules. For instance, to store the rule {foo, bar} -> 42, with salience 10, create a bitmap index on the first attribute of the rule, and set the 10th bit of the “foo” bitmap; likewise for the “bar” bitmap of the second index. Finding rules which match both attributes is a bitwise intersection, and since we rank by salience, the rule that wins is the first in the set. An obvious choice for fast ordered sets is RoaringBitmap.

RoaringBitmap consists of containers, which are fast, cache-friendly sorted sets of integers, and can contain up to 2^16 shorts. In RoaringBitmap, containers are indexed by keys consisting of the most significant 16 bits of the integer. For a rules engine, if you have more than 2^16 rules you have a much bigger problem anyway, so a container could index all the rules you could ever need, so RoaringBitmap itself would be overkill. While RoaringBitmap indexes containers by shorts (it does so for the sake of compression), we can implement wildcard and prefix matching by associating containers with Strings rather than shorts. As the core data structure of the rules engine, a RoaringBitmap container is placed at each node of an Apache commons PatriciaTrie. It’s really that simple – see the source at github.

When the rules engine is queried, a set consisting of all the rules that match is intersected with the container found at the node in the trie matching the value specified for each attribute. When more than one rule matches, the rule with the highest salience is accessed via the Container.first() method, one of the features I have contributed to RoaringBitmap. See example usage at github.



HTTP Content Negotiation

Ever had the situation where you’ve implemented a REST API, and then another team or potential customer comes along and asks if you could just change the serialisation format? Or maybe you are using a textual media type and want to experiment to see if performance would improve with a binary format? I see this done wrong all the time.

There is a mechanism for content negotiation built in to HTTP since the early days. It’s a very simple mechanism so this won’t be a long post. There is a short Java example using Jackson and Jersey to allow parametric content negotiation to be performed transparently to both the server and client.

Never Ever Do This

One option would be to implement a completely new service layer for each content type you need to serve. This not only costs you time, and probably test coverage, but moving between serialisation formats as a client of your API will probably constitute a full-blown migration. Keeping the implementations synchronised will become difficult, and somebody needs to tell the new guy he needs to implement the change in the JSON version as well as the SMILE version, and if the XML version could be updated by the end of the week, that would be great. I have seen commercial products that have actually done this.

Don’t Do This (Unless You’ve Already Done It)

Another option that I see in various APIs is using a query parameter for GET requests. This offends my sensibilities slightly, because it has the potential to conflate application and transport concerns, but it’s not that bad. The SOLR REST API has a query parameter wt which allows the client to choose between JSON and XML, for instance. It’s not exactly cumbersome and the client can easily change the serialisation format by treating it as a concern of the application. But what if you want to query the API from a browser? Browsers speak HTTP ex cathedra, not the local patois your API uses. The browser will have to put up with whatever your defaults are, and hope it can consume that format. It’s unnecessary complexity in the surface area of your API but it’s quite common, and at least it isn’t weird.

1996 Calling

All you have to do is set the Accept header of the request to the media type (for instance – application/cbor) you want to consume in the response.  If the server can produce the media type, it will do so, and then set the Content-Type header on the response. If it can’t produce the media type (and this should come out in integration testing) the server will respond with a 300 response code with textual content listing the supported media types that the client should choose from.

Content Negotiation with Jersey and Jackson JAX-RS providers

Jackson makes this very easy to do in Java. We want to be able to serialise the response below into a MIME type of the client’s choosing – they can choose from JSON, XML, CBOR, SMILE and YAML. The response type produced is not constrained by annotations on the method; none of the formats need ever be formally acknowledged by the programmer and can be added or removed by modifying the application’s classpath.

public class ContentResource {

  public Response getContent() {
    Map<String, Object> content = new HashMap<>();
    content.put("property1", "value1");
    content.put("property2", 10);
    return Response.ok(content).build();

There are Jackson JAX-RS providers included by the maven dependencies below. Their presence is transparent to the application and can be used via the standard JAX-RS Resource SPI. Handler resolution is implemented by a chain-of-responsibility where the first resource accepting the Accept header will serialise the body of the response and set the Content-Type header.






There’s some simple coding to do to configure the Jersey resource and run it embedded with Jetty.

public class ServerRunner {

  public static void main(String[] args) throws Exception {
    ResourceConfig config = new ResourceConfig();
    config.packages(true, "com.fasterxml.jackson.jaxrs");
    ServletHolder servletHolder = new ServletHolder(new ServletContainer(config));
    Server server = new Server(8080);
    ServletContextHandler handler = new ServletContextHandler(server, "/*");
    handler.addServlet(servletHolder, "/*");

If you do this, then you can leave the decision up to your client and just add and remove resource provider jars from your classpath. A silver lining is that you can test the mechanism yourself from IntelliJ:



And you can get non-technical users to check from a browser: