Category: Programming Language

How to teach C++

In the closing Keynote of this year’s Meeting C++, Nicolai Josuttis remarked how hard it can be to teach C++ with its ever expanding complexity. His example was teaching rookies about initialization in C++, i.e. whether to use assignment =, parens () or curly braces {}. He also asked for more application-level programmers to participate.

Well, I am an application programmer, and I also have experience with teaching C++. Last year I held a C++ introductory course for experienced C programmers who mostly had never used C++ before. From my experience, I can completely agree with what Nico had to say about teaching C++. The language and its subtlety can be truly overwhelming.

Most of the complexity in C++ boils down to tuning your code for optimal performance. We cannot just leave that out, can we? After all, the sole reason to use C++ is performance, right?

My approach

Performance is one key reason for using C++, no doubt about it. There is a few more, but let us not get distracted. What is even more important is the potential to optimize for performance. But you can do that later! It is actually quite crazy what performance crimes you can get away with in C++ and still have something pretty fast overall, especially with move-semantics and improved RVO.

Given that, I picked a simple subset to start with:

  • Pass by-value only, do not use pointers nor references.
  • Structure your programs around simple data-only structs and functions transforming them.
  • Make good use of the data structures and algorithms from std.

Believe me, you too can write pretty useful programs this way. This is not too far from a data-oriented style anyways.
So, yes, we can actually leave out the performance specific parts for quite some time, while still making sure our programs can be optimized eventually.

And from there?

You can gradually start introducing references and move semantics, when measurement shows copying affecting the performance. This way you can introduce tooling like a profiler and see the effects off passing things around by reference in a language where everything, by default, is passed by value. These things will start to make sense in a context.

The arguably better approach to move semantics is of course types that prevent you from copying, but allow moving. Most resources with side-effects are like this: file handles, locks, threads. You will need to introduce RAII for this to make sense, e.g. writing ctors and dtors.

But you still do not need to write templates, use virtual-function polymorphism, or even pointers. But when you get to those, you will have a good context to use them.

Have you tought C++ and some experience to share? I would like to hear about it!

The sorry state of Grails (Plugins)

We have been developing and maintaining a complex web application on Grails since summer of 2008. By then Grails had passed the 1.0 release milestone and was really hot. A good 10 years later the application is still in use and we are trying to upgrade from Grails 2.4 to 3.3.

Upgrading Grails – a rough ride

Similar to past upgrade experiences the ride is not very smooth. Besides the major changes like the much welcomed switch to the gradle build system, interceptors instead of filters and streamlined configuration there are again a host of more subtle changes. The biggest problem for us though is the plugin situation.

It’s the plugins

In the past we had tough breaks like the abandoned selenium plugin in favor of the much better geb for functional testing. That had cost us a lot of work and many lost and not yet rewritten functional tests.

This time it seems especially hard because you two of our central plugins are not readily available anymore:

  1. Apache Shiro Plugin
  2. Compass-based Searchable Plugin

1. Shiro authentication

There still is no official release of the shiro plugin for Grails 3.x. After some searching and researching the initial port on github we decided to fork and maintain the most current forked version ourselves and try to work with it. Fortunately it was relatively easy to integrate and to update some dependencies. Our authentication and authorization works at least as good as before and we do not face additional problems. Working with interceptors feels quite good, too.

2. Search

The situation is harder with search. Compass and the searchable plugin are dead – plain and simple. The replacement for grails is the elasticsearch plugin which mostly adopted the API of the searchable plugin. Getting it to work is not that easy though. You have different versions depending on the grails 3 version you are targetting. Each plugin version targets a specific elasticsearch server version and so on. Often times (like in the default configuration) you will need a matching mapper-attachment plugin that is not available on maven in newer versions. This is mentioned somewhere in the midst of the plugin documentation.

Furthermore the plugin itself has some problems with hibernate proxies and concurrency so here we have to mess around with the plugin code once more. Once we have everything working for us like before we will try to get our patches upstream.

Marching forward

The upgrade from 2.x to 3.x is the biggest (and best) step of Grails into the right direction. On the downside it places a lot of burden on the application and plugin developers. That again increases the cost of maintaining proven applications further.

Right now we are close to a Grails 3.3 version of our application but have invested considerable effort into this upgrade.

Our current recommendation and practice is to not start new web applications based on the grails framework because there have been too many breaking changes and the maintainance cost is high. But we are keeping a close look at grails because the increased modularization and and new options like the grails-react-profile may keep grails interesting in the future.

Bending the Java syntax until it breaks

Did you ever bend a programming language until it breaks? This blog entry explores some rather unknown and silly regions of Java and connects them with upcoming language features.

Java is a peculiar programming language. It is used in lots of professional business cases and yet regarded as an easy language suitable for beginner studies. Java’s syntax in particular is criticized as bloated and overly strict and on the next blog praised as lenient and powerful. Well, lets have a look at a correct, runnable Java program that I like to show my students:

class $ {
	{
		System.out.println("hello world");
	}

	static {
		System.out.println("hello, too");
	}

	$() {
		http://www.softwareschneiderei.de
		while ($()) {
			break http;
		}
	}

	static boolean $() {
		return true;
	}

	public static void main(String[] _$) {
		System.out.println($.$());
	}
}

This Java code compiles, perhaps with some warnings about unlucky naming and can be run just fine. You might want to guess what its output to the console is. Notice that the code is quiet simple. No shenanigans with Java’s generics or the dark corners of lambda expressions and method handles. In fact, this code compiles and runs correctly since more than 20 years.

Lets dissect the code into its pieces. To really know what’s going on, you need to look into Java’s Language Specification, a magnificent compendium about everything that can be known about Java on the syntax level. If you want to dive even deeper, the Java Virtual Machine Specification might be your cup of tea. But be warned! Nobody on this planet understands everything in it completely. Even the much easier Java Language Specification contains chapters of pure magic, according to Joshua Bloch (you might want to watch the whole presentation, the statement in question is around the 6 minute mark). And in the code example above, we’ve used some of the magic, even if they are beginner level tricks.

What about the money?

The first glaring anomaly in the code is the strange names that are dollar signs or underscores. These are valid names, according to chapter 3.8 about Identifiers. And we’ve done great by choosing them. Let me quote the relevant sentence from this chapter:

“The “Java letters” include uppercase and lowercase ASCII Latin letters A-Z (\u0041-\u005a), and a-z (\u0061-\u007a), and, for historical reasons, the ASCII dollar sign ($, or \u0024) and underscore (_, or \u005f). The dollar sign should be used […]”

Oh, and by the way: Java identifiers are of unlimited length. You could go and write valid Java code that never terminates. We’ve gone the other way and made our names as short as possible – one character. Since identifiers are used as class names, method names, variable names and (implicitly) constructor names, we can name them all alike.

The variable name of the arguments in the main method used to be just an underscore, but somebody at Oracle changed this section of the Language Specification and added the following sentence:

“The underscore may be used in identifiers formed of two or more characters, but it cannot be used as a one-character identifier due to being a keyword.”

This change happened in Java 9. You can rename the variable “_$” to just “_” in Java 8 and it will work (with a warning).

URLs as first-class citizens?

The next thing that probably caught your eye is the URL in the first line of the constructor. It just stands there. And as I told you, the code compiles. This line is actually a combination of two things: A labeled statement and a comment. You already know about end-of-line comments that are started with a double slash. The rather unknown thing is the labeled statement before it, ending with a colon. This is one of the darker regions of the Language Specification, because it essentially introduces a poor man’s goto statement. And they knew it, because they explicitly talk about it:

“Unlike C and C++, the Java programming language has no goto statement; identifier statement labels are used with break…”

And this explains the weird line in the while loop: “break http” doesn’t command Java to do harm to the computer’s Internet connection, but to leave and complete the labeled statement, in our case the while loop. This spares us from the looming infinite loop, but raises another question: What names are allowed as labels? You’ve guessed it, it’s a Java identifier. We could have named our label “$:” instead of “http:” and chuckled about the line “break $”.

So, Java has a goto statement, but it isn’t called as such and it’s crippled to the point of being useless in practice. In my 20+ years of Java programming, I’ve seen it used just once in the wild.

What about it all?

This example of Java code serves me as a reminder that a programming language is what we make out of it. Our Java programs could easily all look like this if we wanted to. It’s not Java’s merit that our code is readable. And it’s not Java’s fault that we write bloated code. Both are results of our choices as programmers, not an inevitableness from the language we program in.

I sometimes venture to the darker regions of programming languages to see what the language could look and feel like if somebody makes the wrong decisions. This code example is from one of those little journeys several years ago. And it proved its worth once again when I tried to compile it with Java 9. Remember the addition in the Language Specification that made the single underscore a keyword? It wasn’t random. Java’s authors want to improve the lambda expressions in Project Amber, specifically in the JEP 302 (Lambda Leftovers). This JDK Enhancement Proposal (JEP) was planned for Java 10, is still not included in Java 11 and has no clear release date yet. My code gave me the motivation to dig into the topic and made me watch presentations like the one from Brian Goetz at Devoxx 2017 that’s really interesting and a bit unsettling.

Bending things until they break is one way to learn about their limits. What are your ways to learn about programming languages? Do you always stay in the middle lane? Leave a comment on your journeys.

Consistency over magic, please

The Groovy programming language is a JVM based scripting language. It is used by the Grails web framework and the Gradle build automation system.

Groovy has a language feature called Named argument constructors. This means that given a class with properties, for example

class Example {
  String text
}

you can initialize the properties directly when calling the constructor:

def example = new Example(text: ' This is an example. ')
assert example.text == ' This is an example. '

This is basically a shortcut for initializing the properties via explicit assignment:

def example = new Example()
example.text = ' This is an example. '
assert example.text == ' This is an example. '

So far so good.

Enter Grails

We use the aforementioned Grails framework for some of our web application projects. It is advertised on its website as featuring “convention-over-configuration” and “sensible defaults”. Grails uses the Groovy programming language, and a simple domain class looks just like a plain old Groovy class, except that it lives under the grails-app/domain directory (this is one of the convention-over-configuration aspects):

class Example {
  String text
}

As expected, you can initialize the property via regular assignment:

def example = new Example()
example.text = ' This is an example. '
assert example.text == ' This is an example. '

So one might expect that you can initialize it via a named argument constructor call as well:

def example = new Example(text: ' This is an example. ')
assert example.text == ' This is an example. '

And indeed, you can. But what’s this? Our assertion fails:

assert example.text == ' This is an example. '
               |    |
               |    false
               This is an example.

It is not directly obvious from the assertion failure output, but the property value is indeed no longer equal to the expected text: the leading and trailing spaces got trimmed!

I was surprised, but after some research in Grails documentation it turned out that it’s not a bug, but a feature. In the section on Data Binding, you can find the following sentence:

The mass property binding mechanism will by default automatically trim all Strings at binding time. To disable this behavior set the grails.databinding.trimStrings property to false in grails-app/conf/application.groovy.

Groovy’s named argument constructor feature is used as a data binding mechanism by Grails to bind web request parameters to a domain object. For this the default behavior was modified, so that strings are automatically trimmed. I can only guess that this is considered to be an instance of the “sensible defaults” mentioned on the Grails homepage.

To me personally this kind of surprising behavior is not a sensible default, and I think it goes against the Principle of least astonishement. I prefer consistency over “magic”.

luabind deboostified tips and tricks

luabind deboostified is a fork of the luabind project that helps exposing APIs to Lua. As the name implies, it replaces the boost dependency with modern C++, which makes it a lot more pleasant to work with.

Here are a few tips and tricks I learned while working with it. Some tricks might be applicable to the original luabind – I do not know.

1. Splitting module registration

You can split the registration code for different classes. I usually add a register function per class, like this:

struct A {
  void doSomething();
  static luabind::scope registerWithLua();
};

struct B {
  void goodStuff();
  static luabind::scope registerWithLua();
};

You can then combine their registration code into a single module on the Lua side: 

void registerAll(lua_State* L) {
  luabind::module(L)[
    A::registerWithLua(),
    B::registerWithLua()];
}

The implementation of a registration function looks like this:

luabind::scope A::registerWithLua()
{
  return luabind::class_<A>("A")
    .def("doSomething", &A::doSomething);
}

2. Multiple policies and multiple return values

Unlike C++, Lua has real multiple return values. You can use that by utilizing the return value policies that luabind offers. Lets say, you want to write this in Lua:

local x, y = a.getPosition()

The C++ side could look like this:

void getPosition(A const& a, float& x, float& y);

The deboostified fork needs its policies supplied in a type list. Let’s use a small helper meta-function to build that:

template <typename... T>
using joined = 
  typename luabind::meta::join<T...>::type;

Once you have that, you can expose it like this: 

luabind::def("getPosition", &getPosition,
              joined<
                luabind::pure_out_value<2>,
                luabind::pure_out_value<3>
              >());

3. Specialized data structures using luabind::object

Using the converters in luabind is not the only way to make Lua values from C++. Almost everything you can do in Lua itself, you can do with luabind::object. Here is a somewhat contrived example:

luabind::object repeat(luabind::object what,
                       int count) {
  // Create a new table object
  auto result = luabind::newtable(
    what.interpreter());
  // Fill it as an array [1..N]
  for (int i = 1; i <= count; ++i)
    result[i] = what;
  return result;
}

This function can then be exported via luabind::def and used just like any other function. This is just the tip of the iceberg, though. For example, you can also write functions that, at runtime, behave differently when a number is passed in as when a table is passed in. You can find out the Lua type with luabind::type(myObject).

Of course, as soon as you want to create new objects to return to Lua, you need the lua_State pointer in that function. Using the interpreter from a passed-in luabind::object is one way, but I have yet to find another pleasant way to do this. It is probably possible to use the policies to do this, and have them pass that in as a special parameter, but for now I am using some complicated machinery to bind lambda functions that capture the Lua interpreter.

That’s it for now..

Keep in mind that these are not thoroughly researched best-practices, but patterns I have used to solve actual problems. There might be better solutions out there – if you know any, please let me know. Hope this helped!

Client-side web development: Drink the Kool-Aid or be cautious?

Client side web development is a fast-changing world. JavaScript libraries and frameworks come and go monthly. A couple of years ago jQuery was a huge thing, then AngularJS, and nowadays people use React or Vue.js with a state container like Redux. And so do we for new projects. Unfortunately, these modern client-side frameworks are based on the npm ecosystem, which is notoriously known for its dependency bloat. Even if you only have a couple of direct dependencies the package manager lock file will list hundreds of indirect dependencies. Experience has shown that lots of dependencies will result in a maintenance burden as time passes, especially when you have to do major version updates. Also, as mentioned above, frameworks come and then go out of fashion, and the maintainers of a framework move on to their next big thing pet project, leaving you and your project sitting on a barely or no longer maintained base, and frameworks can’t be easily replaced, because they tend to permeate every aspect of your application.

With this frustrating experience in mind we recently did an experiment for a new medium sized web project. We avoided frameworks and the npm ecosystem and only used JavaScript libraries with no or very few indirect dependencies, which really were necessary. Browsers have become better at being compatible to web standards, at least regarding the basics. Libraries like jQuery and poly-fills that paper over the incompatibilities can mostly be avoided — an interesting resource is the website You Might Not Need jQuery.

We still organised our views as components, and they are communicating via a very simple event dispatcher. Some things had to be done by foot, but not too much. It works, although the result is not as pure as it would have been with declarative views as facilitated by React and a functional state container like Redux. We’re still fans of the React+Redux approach and we’re using it happily (at least for now) for other projects, but we’re also skeptical regarding the long term costs, especially from relying on the npm ecosystem. Which approach will result in less maintenance burden? We don’t know yet. Time will tell.

Decoding non-utf8 server responses using the Fetch API

The new Javascript Fetch API is really nice addition to the language and my preferable, and in fact the only bearable, way to do server requests.
The Promise based API is a lot nicer than older, purely callback-based, approaches.

The usual approach to get a text response from a server using the Fetch API looks like this:

let request = fetch(url)
  .then(response => response.text())
  .then(handleText);

But this has one subtle problem:

I was building a client application that reads weather data from a small embedded device. We did not have direct access to changing the functionality of that device, but we could upload static pages to it, and use its existing HTML API to query the amount of registered rainfall and lightning strikes.

Using the fetch API, I quickly got the data and extracted it from the HTML but some of the identifiers had some screwed up characters that looked like decoding problems. So I checked whether the HTTP Content-Type was set correctly in the response. To my surprise it was correctly set as Content-Type: text/html; charset=iso-8859-1.

So why did my Javascript Application not get that? After some digging, it turned out that Response’s text() function always decodes the payload as utf-8. The mismatch between that and the charset explained the problem!

Obviously, I had to do the decoding myself. The solution I picked was to use the TextDecoder class. It can decode an ArrayBuffer with a given encoding. Luckily, that is easy to get from the response:

let request = fetch(url)
  .then(response => response.arrayBuffer())
  .then(buffer => {
    let decoder = new TextDecoder("iso-8859-1");
    let text = decoder.decode(buffer);
    handleText(text);
  });

Since I only had to support that single encoding, that worked well for me. Keep in mind that the TextDecoder is still experimental Technology. However, we had a specific browser as a target and it works there. Lucky us!

For what the javascript!

The setting

We are developing and maintaining an important web application for one of our clients. Our application scrapes a web page and embeds our own content into that page frame.

One day our client told us of an additional block of elements at the bottom of each page. The block had a heading “Image Credits” and a broken image link strangely labeled “inArray”. We did not change anything on our side and the new blocks were not part of the HTML code of the pages.

Ok, so some new Javascript code must be the source of these strange elements on our pages.

The investigation

I started the investigation using the development tools of the browser (using F12). A search for the string “Image Credits” instantly brought me to the right place: A Javascript function called on document.ready(). The code was basically getting all images with a copyright attribute and put the findings in an array with the text as the key and the image url as the value. Then it would iterate over the array and add the copyright information at the bottom of each page.

But wait! Our array was empty and we had no images with copyright attributes. Still the block would be put out. I verified all this using the debugger in the browser and was a bit puzzled at first, especially by the strange name “inArray” that sounded more like code than some copyright information.

The cause

Then I looked at the iteration and it struck me like lightning: The code used for (name in copyrightArray) to iterate over the elements. Sounds correct, but it is not! Now we have to elaborate a bit, especially for all you folks without a special degree in Javascript coding:

In Javascript there is no distinct notion of associative arrays but you can access all enumerable properties of an object using square brackets (taken from Mozillas Javascript docs):

var string1 = "";
var object1 = {a: 1, b: 2, c: 3};

for (var property1 in object1) {
  string1 = string1 + object1[property1];
}

console.log(string1);
// expected output: "123"

In the case of an array the indices are “just enumerable properties with integer names and are otherwise identical to general object properties“.

So in our case we had an array object with a length of 0 and a property called inArray. Where did that come from? Digging further revealed that one of our third-party libraries added a function to the array prototype like so:

Array.prototype.inArray = function (value) {
  var i;
  for (i = 0; i &lt; this.length; i++) {
    if (this[i] === value) {
      return true;
    }
  }
  return false;
};

The solution

Usually you would iterate over an array using the integer index (old school) or better using the more modern and readable for…of (which also works on other iterable types). In this case that does not work because we do not use integer indices but string properties. So you have to use Object.keys().forEach() or check with hasOwnProperty() if in your for…in loop if the property is inherited or not to avoid getting unwanted properties of prototypes.

The takeaways

Iteration in Javascript is hard! See this lengthy discussion…The different constructs are named quite similar an all have subtle differences in behaviour. In addition, libraries can mess with the objects you think you know. So finally some advice from me:

  • Arrays are only true arrays with positive integer indices/property names!
  • Do not mess with the prototypes of well known objects, like our third-party library did…
  • Use for…of to iterate over true arrays or other iterables
  • Do not use associative arrays if other options are available. If you do, make sure to check if the properties are own properties and enumerable.

 

Game Optimization Resolved

In my last blog post, I explained a performance problem in my game abstractanks but not how I solved it.

So I had not done any optimization work in a while, so the first thing I did turned out to be an error. And not only in hindsight – I actually knew how to tackle a problem like that – I just temporarily forgot at that point.

Going down the rabbit hole

Where we left off, my profiler showed FriendlyUnitOccupies as the culprit. That function basically does circle/circle collision detection using a quad-tree as the spatial acceleration structure. Looking at the samples from my profiler, I could see that that it was descending into the tree quite deeply. Like all tree structures, a quad-tree does pointer-chasing which is very bad for modern CPUs. So I figured I should look at how to optimize that. The data structure was implemented in a hurry, so there seemed plenty to do:

  • Instead of recursing into each node, use tail-call optimization and early culling to speed up traversal.
  • Pre-cache the query with the max-search radius and the other requirements to the units, e.g. not dead, same team, etc.. and then use that to build a new tree for the actual queries.

Because the data structure was pretty non-generic, I started to basically rewrite it to use it in this scenario. While I was about half way through with that, it dawned on me that I was barking at the wrong tree.

Taking a step back

The excellent book Video Game Optimization has some great advice on which level to attack an optimization problem.

  1. System-level. Can you change the system to do something differently and still solve your problem?
  2. Algorithm-level. Are you using the most efficient right algorithm for the data you have?
  3. Micro-level. Are you not wasting any processing power on the lower levels?

I was already on the algorithm level. So I went back to the systemic level: What if the AI did not try to change the target position that often, maybe just every few seconds? That effectively meant lowering the AIs APM. It’s not a bad solution, especially since that makes the AI behave more human. But on the other hand, real-time games, as the name implies, have a soft real-time requirement. So you generally like to avoid huge workloads that go over your frame budget. With how slow the algorithm was, that could easily be the case. The solution is then to do the work concurrently, either by splitting it up or doing it in the background. Both solutions seemed difficult, since the AI code does currently not allow for easy concurrency. So that idea was out.

What if the parking-positions where cached? Subsequent calls to get parking positions could probably reuse a lot of the positions that were computed in previous frames, given that the target point only moves by a little bit each frame. I figured that might work, but it requires more housekeeping and data-dependencies – the result of the previous query needs to be used for the next. That seemed complex and therefore brittle.

A Solution?

Temporal coherency was a pretty good idea though, but not the scale was to big this time. What if I exploited it within a single frame? Now the original code did obscure this, but maybe it gets a little more clear if I write it like this:

optional<v2> GameWorld::FindFreePosition(v2 Center, std::vector<v2> const& Occupied)
{
  auto CheckPosition = [&](v2 Candiate)
  {
    if (!IsPassable(Candidate))
      return false;

    if (OverlapsWith(Occupied))
      return false;

    return !FriendlyUnitOccupies(Candidate);  
  };
  auto Samples = SampledPositions(Center, SomeRandomness());
  auto Found = find_if(SampledPositions.begin(), SampledPositions.end(), CheckPosition(Position));
  
  return (Found != SampledPositions.end()) ? *Found : none;
}

Now as I explained in the previous post, this was called in a loop for each unit to be parked. 

std::vector<v2> GameWorld::FindParkingPositions(v2 Center, std::size_t N)
{
  std::vector<v2> Results;
  for (std::size_t i = 0; i < N; ++i)
  {
    auto MaybePosition = FindFreePosition(Center, Results);
    if (!MaybePosition) // No more free space?
      break;
    Results.push_back(*MaybePosition);
  }
  return Results;
}

Easy to see: counting the number of CheckPosition calls, this algorithm is O(n) in number of sampled positions. The number of sampled positions depends linearly on the number of units to be parked, because more units obviously need more parking positions, essentially making this O(n²) for the unit count! But the positions get resampled for each unit – with the only change being the little bit of randomness that is injected everytime. In other words, each call would just test false for sampled positions roughly corresponding to the units that are already placed.

So what I did was a very small change: only inject the randomness once and merge the loops:

auto Samples = SampledPositions(Center, SomeRandomness());
std::vector<v2> Results;

for (auto const& Sample : Samples)
{
  if (CheckPosition(Sample))
    Results.push_back(Sample);

  if (Result.size() >= N)
    break;
}
return Results;

And this did the trick! The algorithm’s run-time when below the 1ms range, and the smaller variation in randomness is not really visible.

Conslusions

I was thrown off-track be the false conclusion that CheckPositions was too slow when it was in fact just called too often. Context is key! Always approach these things outside-in.
Using less-than-optimal abstractions obscured the opportunity to hoist out the sample generation from me. Iteration is always a separate concern, even when it is not on containers!

Transforming C-Style arrays in java

Every now and then some customer asks us to fix or improve some important legacy application other people have written. Usually, such projects are fun and it is rewarding to see the improvements both in code and value for the users.

In one of these projects there is a Java GUI application that uses C-style arrays for some of its central data structures:

public class LegoBox  {
  public LegoBrick[] bricks = new LegoBrick[8000];
  public int brickCount = 0;
}

The array-length is a constant upper bound and does not denote the actual elements in the array. Elements are added dynamically to the array and it looks like a typical job for a automatically growing Collection like java.util.ArrayList. Most operations simply iterate over all elements and perform some calculations. But changing such a central part in a performance sensitive application is not only a lot of work but also risky.

We decided to take an incremental approach to improve code readability and maintainability and measured performance with a large, representative dataset between refactorings. There are two easy alternative APIs that improve working with the above data structure.

Imperative API

Smooth migration from the existing imperative “ask”-code (see “Tell, don’t ask”-principle) can be realized by providing an java.util.Iterable to the underlying array.


public int countRedBricks() {
  int redBrickCount = 0;
  for (int i = 0; i < box.brickCount; i++) {
    if (box.bricks[i].isRed()) {
      redBrickCount++;
    }
  }
  return redBrickCount;
}

Code like above is easily transformed to much clearer code like below:

public class LegoBox  {
  public LegoBrick[] bricks = new LegoBrick[8000];
  public int brickCount = 0;

  public Iterable<LegoBrick> allBricks() {
    return Arrays.stream(tr, 0, brickCount).collect(Collectors.toList());
  }
}

public int countRedBricks() {
  int redBrickCount = 0;
  for (LegoBrick brick : box.bricks) {
    if (brick.isRed()) {
      redBrickCount++;
    }
  }
  return redBrickCount;
}

Functional API

A nice alternative to the imperative solution above is a functional interface to the array. In Java 8 and newer we can provide that easily and encapsulate the iteration over our array:

public class LegoBox  {
  public LegoBrick[] bricks = new LegoBrick[8000];
  public int brickCount = 0;

  public <R> R forAllBricks(Function<Brick, R> operation, R identity, BinaryOperator<R> reducer) {
    return Arrays.stream(bricks, 0, brickCount).map(operation).reduce(identity, reducer);
  }

  public void forAllBricks(Consumer<LegoBrick> operation) {
    Arrays.stream(bricks, 0, brickCount).forEach(operation);
  }
}

public int countRedBricks() {
  return box.forAllBricks(brick -> brick.isRed() ? 1 : 0, 0, (sum, current) -> sum + current);
}

The functional methods can be tailored to your specific needs, of course. I just provided two examples for possible functional interfaces and their implementation.

The function + reducer case is a very general interface and used here for an implementation of our “count the red bricks” use case. Alternatively you could implement this use case with a more specific but easier to use filter + count interface:

public class LegoBox  {
  public LegoBrick[] bricks = new LegoBrick[8000];
  public int brickCount = 0;

  public long countBricks(Predicate<Brick> filter) {
    return Arrays.stream(bricks, 0, brickCount).filter(operation).count();
  }
}

public int countRedBricks() {
  return box.countBricks(brick -> brick.isRed());
}

The consumer case is very simple and found a lot in this specific project because mutation of the array elements is a typical operation and all over the place.

The functional API avoids duplicating the iteration all the time and removes the need to access the array or iterable/collection. It is therefore much more in the spirit of “tell”.

Conclusion

The new interfaces allow for much simpler and maintainable client code and remove a lot of duplicated iterations on the client side. They can be introduced on the way when implementing requested features for the customer.

That way we invested only minimal effort in cleaner, better maintainable and more error-proof code. When someday all accesses to the public array are encapsulated we can use the new found freedom to internalize the array and change it to a better fitting data structure like an ArrayList.