RubyMotion: Ruby for iOS development

RubyMotion is a new (commercial) way to develop apps for iOS, this time with Ruby

RubyMotion is a new (commercial) way to develop apps for iOS, this time with Ruby. So why do I think this is better than the traditional way using ObjectveC or other alternatives?

Advantages to other alternatives

Other alternatives often use a wrapper or a different runtime. The problem is that you have to wait for the library/wrapper vendor to include new APIs when iOS gets a new update. RubyMotion instead has a static compiler which compiles to the same code as ObjectiveC. So you can use the myriads of ObjectiveC libraries or even the interface builder. You can even mix your RubyMotion code with existing ObjectiveC programs. Also the static compilation gives you the performance advantages of real native code so that you don’t suffer from the penalties of using another layer. So you could write your programs like you would in ObjectiveC with the same performance and using the same libraries, then why choose RubyMotion?

Advantages to the traditional way

First: Ruby. The Ruby language has a very nice foundation: everything is an expression. And everything can be evaluated with logic operators (only nil and false is false).
In ObjectiveC you would write:

  cell = tableView.dequeueReusableCellWithIdentifier(reuseId);
  if (!cell) {
    cell = [[TableViewCell alloc] initWithStyle: cellStyle, reuseIdentifier: reuseId]];
  }

whereas in Ruby you can write

cell = tableView.dequeueReusableCellWithIdentifier(@reuse_id)
  || TableViewCell.alloc.initWithStyle(@cell_style, reuseIdentifier:@reuse_id)

As you can see you can use the Cocoa APIs right away. But what excites me even more is the community which builds around RubyMotion. RubyMotion is only some months old but many libraries and even award winning apps have been written. Some libraries wrap so called boiler plate code and make it more pleasant you to use. Other introduce new metaphors which change the way apps are written entirely.
I see a bright future for RubyMotion. It won’t replace ObjectiveC for everyone but it is a great alternative.

Grails and the query cache

The principle of least astonishment can be violated in the unusual places like using the query cache on a Grails domain class.

Look at the following code:

class Node {
  Node parent
  String name
  Tree tree
}

Tree tree = new Tree()
Node root = new Node(name: 'Root', tree: tree)
root.save()
new Node(name: 'Child', parent: root, tree: tree).save()

What happens when I query all nodes by tree?

List allNodesOfTree = Node.findAllByTree(tree, [cache: true])

Of course you get 2 nodes, but what is the result of:

allNodesOfTree.contains(Node.get(rootId))

It should be true but it isn’t all the time. If you didn’t implement equals and hashCode you get an instance equals that is the same as ==.
Hibernate guarantees that you get the same instance out of a session for the same domain object. (Node.get(rootId) == Node.get(rootId))

But the query cache plays a crucial role here, it saves the ids of the result and calls Node.load(id). There is an important difference between Node.get and Node.load. Node.get always returns an instance of Node which is a real node not a proxy. For this it queries the session context and hits the database when necessary. Node.load on the other hand never hits the database. It returns a proxy and only when the session contains the domain object it returns a real domain object.

So allNodesOfTree returns

  • two proxies when no element is in the session
  • a proxy and a real object when you call Node.get(childId) beforehand
  • two real objects when you call get on both elements first

Deactivating the query cache globally or for this query only, returns two real objects.

Testing C programs using GLib

Writing programs in good old C can be quite refreshing if you use some modern utility library like GLib. It offers a comprehensive set of tools you expect from a modern programming environment like collections, logging, plugin support, thread abstractions, string and date utilities, different parsers, i18n and a lot more. One essential part, especially for agile teams, is onboard too: the unit test framework gtest.

Because of the statically compiled nature of C testing involves a bit more work than in Java or modern scripting environments. Usually you have to perform these steps:

  1. Write a main program for running the tests. Here you initialize the framework, register the test functions and execute the tests. You may want to build different test programs for larger projects.
  2. Add the test executable to your build system, so that you can compile, link and run it automatically.
  3. Execute the gtester test runner to generate the test results and eventually a XML-file to you in your continuous integration (CI) infrastructure. You may need to convert the XML ouput if you are using Jenkins for example.

A basic test looks quite simple, see the code below:

#include <glib.h>
#include "computations.h"

void computationTest(void)
{
    g_assert_cmpint(1234, ==, compute(1, 1));
}

int main(int argc, char** argv)
{
    g_test_init(&argc, &argv, NULL);
    g_test_add_func("/package_name/unit", computationTest);
    return g_test_run();
}

To run the test and produce the xml-output you simply execute the test runner gtester like so:

gtester build_dir/computation_tests --keep-going -o=testresults.xml

GTester unfortunately produces a result file which is incompatible with Jenkins’ test result reporting. Fortunately R. Tyler Croy has put together an XSL script that you can use to convert the results using

xsltproc -o junit-testresults.xml tools/gtester.xsl testresults.xml

That way you get relatively easy to use unit tests working on your code and nice some CI integration for your modern C language projects.

Update:

Recent gtester run the test binary multiple times if there are failing tests. To get a report of all (passing and failing) tests you may want to use my modified gtester.xsl script.

Building Windows C++ Projects with CMake and Jenkins

An short and easy way to build Windows C++ Software with CMake in Jenkins (with the restriction to support Visual Studio 8).

The C++ programming environment where I feel most comfortable is GCC/Linux (lately with some clang here and there). In terms of build systems I use cmake whenever possible. This environment also makes it easy to use Jenkins as CI server and RPM for deployment and distribution tasks.

So when presented with the task to set up a C++ windows project in Jenkins I tried to do it the same way as much as possible.

The Goal:

A Jenkins job should be set up that builds a windows c++ project on a Windows 7 build slave. For reasons that I will not get into here, compatibility with Visual Studio 8 is required.

The first step was to download and install the correct Windows SDK. This provides all that is needed to build C++ stuff under windows.

Then, after installation of cmake, the first naive try looked like this (in an “execute Windows Batch file” build step)

cmake . -DCMAKE_BUILD_TYPE=Release

This cannot work of course, because cmake will not find compilers and stuff.

Problem: Build Environment

When I do cmake builds manually, i.e. not in Jenkins, I open the Visual Studio 2005 Command Prompt which is a normal windows command shell with all environment variables set. So I tried to do that in Jenkins, too:

call “c:\Program Files\Microsoft SDKs\Windows\v6.0\Bin\SetEnv.Cmd” /Release /x86

cmake . -DCMAKE_BUILD_TYPE=Release

This also did not work and even worse, produced strange (to me, at least) error messages like:

‘Cmd’ is not recognized as an internal or external command, operable program or batch file.

The system cannot find the batch label specified – Set_x86

After some digging, I found the solution: a feature of windows batch programming called delayed expansion, which has to be enabled for SetEnv.Cmd to work correctly.

Solution: SetEnv.cmd and delayed expansion

setlocal enabledelayedexpansion

call “c:\Program Files\Microsoft SDKs\Windows\v6.0\Bin\SetEnv.Cmd” /Release /x86

cmake . -DCMAKE_BUILD_TYPE=Release

nmake

Yes! With this little trick it worked perfectly. And feels almost as with GCC/CMake under Linux:  nice, short and easy.

Why do (different) programming languages matter?

One common saying in software development is: use the best tool for the job. But what is the best tool? I think the best tool is determined by two things: how it fits the problem domain and how it fits your mental model.

One common saying in software development is: use the best tool for the job. But what is the best tool? I think the best tool is determined by two things: how it fits the problem domain and how it fits your mental model. Why your mental model? Just use the best language available! you might think. But as humans we think in languages and even inside these languages everybody has a typical way of expressing himself. Even own words and if they become common we even have a name for it: a dialect. But it is all that you should consider when choosing a programming language? Certainly there are the tools of the trade: the IDE, debugger, profiler, etc. Here is comes down to personal preferences and most of the shortcomings in this field are short term: better tool support is on the way.
There’s another more important aspect though: the community and therefore the mindset which is brought along. The communities form how the languages are used, where the most libraries and frameworks are developed, which problem domains are tackled and what the values are. Values can be testing, elegance, simplicity, robustness, …
Since communities are consisted of individuals, individuals form what the values are. But I think the language designer lays a foundation here: take Ruby for example, Ruby was designed with the intention to make programming fun. This is one of the things that appeals to many developers and the whole community which uses Ruby. Ruby is fun.
These environments spawn amazing things like Rails or more recently RubyMotion Because of the mindset of the community and the foundation inside the language there are these fruits. Last but not least another reason to choose a language is your familiarity with it. You might choose an inferior tool or language because you know it inside out.

Basic Image Processing Tasks with OpenCV

2D detectors and scientific CCD cameras produce many megabytes of image data. Open source library OpenCV is highly recommended as your work horse for all kinds of image processing tasks.

For one of our customers in the scientific domain we do a lot of integration of pieces of hardware into the existing measurement- and control network. A good part of these are 2D detectors and scientific CCD cameras, which have all sorts of interfaces like ethernet, firewire and frame grabber cards. Our task is then to write some glue software that makes the camera available and controllable for the scientists.

One standard requirement for us is to do some basic image processing and analytics. Typically, this entails flipping the image horizontally and/or vertically, rotating the image around some multiple of 90 degrees, and calculcating some statistics like standard deviation.

The starting point there is always some image data in memory that has been acquired from the camera. Most of the time the image data is either gray values (8, or 16 bit), or RGB(A).

As we are generally not falling victim to the NIH syndrom we use open source image processing librarys. The first one we tried was CImg, which is a header-only (!) C++ library for image processing. The header-only part is very cool and handy, since you just have to #include <CImg.h> and you are done. No further dependencies. The immediate downside, of course, is long compile times. We are talking about > 40000 lines of C++ template code!

The bigger issue we had with CImg was that for multi-channel images the memory layout is like this: R1R2R3R4…..G1G2G3G4….B1B2B3B4. And since the images from the camera usually come interlaced like R1G1B1R2G2B2… we always had to do tricks to use CImg on these images correctly. These tricks killed us eventually in terms of performance, since some of these 2D detectors produce lots of megabytes of image data that have to be processed in real time.

So OpenCV. Their headline was already very promising:

OpenCV (Open Source Computer Vision) is a library of programming functions for real time computer vision.

Especially the words “real time” look good in there. But let’s see.

Image data in OpenCV is represented by instances of class cv::Mat, which is, of course, short for Matrix. From the documentation:

The class Mat represents an n-dimensional dense numerical single-channel or multi-channel array. It can be used to store real or complex-valued vectors and matrices, grayscale or color images, voxel volumes, vector fields, point clouds, tensors, histograms.

Our standard requirements stated above can then be implemented like this (gray scale, 8 bit image):

void processGrayScale8bitImage(uint16_t width, uint16_t height,
                               const double& rotationAngle,
                               uint8_t* pixelData)
{
  // create cv::Mat instance
  // pixel data is not copied!
  cv::Mat img(height, width, CV_8UC1, pixelData);

  // flip vertically
  // third parameter of cv::flip is the so-called flip-code
  // flip-code == 0 means vertical flipping
  cv::Mat verticallyFlippedImg(height, width, CV_8UC1);
  cv::flip(img, verticallyFlippedImg, 0);

  // flip horizontally
  // flip-code > 0 means horizontal flipping
  cv::Mat horizontallyFlippedImg(height, width, CV_8UC1);
  cv::flip(img, horizontallyFlippedImg, 1);

  // rotation (a bit trickier)
  // 1. calculate center point
  cv::Point2f center(img.cols/2.0F, img.rows/2.0F);
  // 2. create rotation matrix
  cv::Mat rotationMatrix =
    cv::getRotationMatrix2D(center, rotationAngle, 1.0);
  // 3. create cv::Mat that will hold the rotated image.
  // For some rotationAngles width and height are switched
  cv::Mat rotatedImg;
  if ( (rotationAngle / 90.0) % 2 != 0) {
    // switch width and height for rotations like 90, 270 degrees
    rotatedImg =
      cv::Mat(cv::Size(img.size().height, img.size().width),
              img.type());
  } else {
    rotatedImg =
      cv::Mat(cv::Size(img.size().width, img.size().height),
              img.type());
  }
  // 4. actual rotation
  cv::warpAffine(img, rotatedImg,
                 rotationMatrix, rotatedImg.size());

  // save into TIFF file
  cv::imwrite("myimage.tiff", gray);
}

The cool thing is that almost the same code can be used for our other image types, too. The only difference is the image type for the cv::Mat constructor:


8-bit gray scale: CV_U8C1
16bit gray scale: CV_U16C1
RGB : CV_U8C3
RGBA: CV_U8C4

Additionally, the whole thing is blazingly fast! All performance problems gone. Yay!

Getting basic statistical values is also a breeze:

void calculateStatistics(const cv::Mat& img)
{
  // minimum, maximum, sum
  double min = 0.0;
  double max = 0.0;
  cv::minMaxLoc(img, &min, &max);
  double sum = cv::sum(img)[0];

  // mean and standard deviation
  cv::Scalar cvMean;
  cv::Scalar cvStddev;
  cv::meanStdDev(img, cvMean, cvStddev);
}

All in all, the OpenCV experience was very positive, so far. They even support CMake. Highly recommended!

Game of Life: TDD style in Java

I always got problems finding the right track with test driven development (TDD), going down the wrong track can get you stuck.
So here I document my experience with tdd-ing Conway’s Game of Life in Java.

I always got problems finding the right track with test driven development (TDD), going down the wrong track can get you stuck.
So here I document my experience with tdd-ing Conway’s Game of Life in Java.

The most important part of a game of life implementation since the rules are simple is the datastructure to store the living cells.
So using TDD we should start with it.
One feature of our cells should be that they are equal according to their coordinates:

@Test
public void positionsShouldBeEqualByValue() {
  assertEquals(at(0, 1), at(0, 1));
}

The JDK features a class holding two coordinates: java.awt.Point, so we can use it here:

public class Board {
  public static Point at(int x, int y) {
    return new Point(x, y);
  }
}

You could create your own Position or Cell class and implementing equals/hashCode accordingly but I want to keep things simple so we stick with Point.
A board should holding the living cells and we need to compare two boards according to their living cells:

@Test
public void boardShouldBeEqualByCells() {
  assertEquals(new Board(at(0, 1)), new Board(at(0, 1)));
}

Since we are only interested in living cells (all other cells are considered dead) we store only the living cells inside the board:

public class Board {
  private final Set<Point> alives;

  public Board(Point... points) {
    alives = new HashSet<Point>(Arrays.asList(points));
  }

  @Override
  public boolean equals(Object o) {
    if (this == o) return true;
    if (o == null || getClass() != o.getClass()) return false;

    Board board = (Board) o;

    if (alives != null ? !alives.equals(board.alives) : board.alives != null) return false;

    return true;
  }

  @Override
  public int hashCode() {
    return alives != null ? alives.hashCode() : 0;
  }
}

If you take a look at the rules you see that you need to have a way to count the neighbours of a cell:

@Test
public void neighbourCountShouldBeZeroWithoutNeighbours() {
  assertEquals(0, new Board(at(0, 1)).neighbours(at(0, 1)));
}

Easy:

public int neighbours(Point p) {
  return 0;
}

Neighbours are either vertically adjacent:

@Test
public void neighbourCountShouldCountVerticalOnes() {
  assertEquals(1, new Board(at(0, 0), at(0, 1)).neighbours(at(0, 1)));
}
public int neighbours(Point p) {
  int count = 0;
  for (int yDelta = -1; yDelta <= 1; yDelta++) {
    if (alives.contains(at(p.x, p.y + yDelta))) {
      count++;
    }
  }
  return count;
}

Hmm now both neighbour tests break, oh we forgot to not count the cell itself:
First the test…

@Test
public void neighbourCountShouldNotCountItself() {
  assertEquals(0, new Board(at(0, 0)).neighbours(at(0, 0)));
}

Then the fix:

public int neighbours(Point p) {
  int count = 0;
  for (int yDelta = -1; yDelta <= 1; yDelta++) {
    if (!(yDelta == 0) && alives.contains(at(p.x, p.y + yDelta))) {
      count++;
    }
  }
  return count;
}

And the horizontal adjacent ones:

@Test
public void neighbourCountShouldCountHorizontalOnes() {
  assertEquals(1, new Board(at(0, 1), at(1, 1)).neighbours(at(0, 1)));
}
public int neighbours(Point p) {
  int count = 0;
  for (int yDelta = -1; yDelta <= 1; yDelta++) {
    for (int xDelta = -1; xDelta <= 1; xDelta++) {
      if (!(xDelta == 0 && yDelta == 0) && alives.contains(at(p.x + xDelta, p.y + yDelta))) {
        count++;
      }
    }
  }
  return count;
}

And the diagonal ones are also included in our implementation:

@Test
public void neighbourCountShouldCountDiagonalOnes() {
  assertEquals(2, new Board(at(-1, 1), at(1, 0), at(0, 1)).neighbours(at(0, 1)));
}

So we set the stage for the rules. Rule 1: Cells with one neighbour should die:

@Test
public void cellWithOnlyOneNeighbourShouldDie() {
  assertEquals(new Board(), new Board(at(0, 0), at(0, 1)).next());
}

A simple implementation looks like this:

public Board next() {
  return new Board();
}

OK, on to Rule 2: A living cell with 2 neighbours should stay alive:

@Test
public void livingCellWithTwoNeighboursShouldStayAlive() {
  assertEquals(new Board(at(0, 0)), new Board(at(-1, -1), at(0, 0), at(1, 1)).next());
}

Now we need to iterate over each living cell and count its neighbours:

public class Board {
  public Board(Point... points) {
    this(new HashSet<Point>(Arrays.asList(points)));
  }

  private Board(Set<Point> points) {
    alives = points;
  }

  public Board next() {
    Set<Point> aliveInNext = new HashSet<Point>();
    for (Point cell : alives) {
      if (neighbours(cell) == 2 {
        aliveInNext.add(cell);
      }
    }
    return new Board(aliveInNext);
  }
}

In this step we added a convenience constructor to pass a set instead of some cells.
The last Rule: a cell with 3 neighbours should be born or stay alive (the pattern is called blinker, so we name the test after it):

@Test
public void blinker() {
  assertEquals(new Board(at(-1, 1), at(0, 1), at(1, 1)), new Board(at(0, 0), at(0, 1), at(0, 2)).next());
}

For this we need to look at all the neighbours of the living cells:

public Board next() {
  Set<Point> aliveInNext = new HashSet<Point>();
  for (Point cell : alives) {
    for (int yDelta = -1; yDelta <= 1; yDelta++) {
      for (int xDelta = -1; xDelta <= 1; xDelta++) {
        Point testingCell = at(cell.x + xDelta, cell.y + yDelta);
        if (neighbours(testingCell) == 2 || neighbours(testingCell) == 3) {
          aliveInNext.add(testingCell);
        }
      }
    }
  }
  return new Board(aliveInNext);
}

Now our previous test breaks, why? Well the second rule says: a *living* cell with 2 neighbours should stay alive:

public Board next() {
  Set<Point> aliveInNext = new HashSet<Point>();
  for (Point cell : alives) {
    for (int yDelta = -1; yDelta <= 1; yDelta++) {
      for (int xDelta = -1; xDelta <= 1; xDelta++) {
        Point testingCell = at(cell.x + xDelta, cell.y + yDelta);
        if ((alives.contains(testingCell) && neighbours(testingCell) == 2) || neighbours(testingCell) == 3) {
          aliveInNext.add(testingCell);
        }
      }
    }
  }
  return new Board(aliveInNext);
}

Done!
Now we can refactor and make the code cleaner like removing the logic duplication for iterating over the neighbours, adding methods like toString for output or better failing test messages, etc.

Python Pitfall: Alleged Decrement Operator

The best way to make oneself more familiar with the possibilities and pitfalls of a newly learned programming language is to start pet projects using that language. That’s just what I did to dive deeper into Python. While working on my Python pet project I made a tiny mistake which took me quite a while to figure out. The code was something like (highly simplified):

for i in range(someRange):
  # lots of code here
  doSomething(--someNumber)
  # even more code here

For me, with a strong background in Java and C, this looked perfectly right. Yet, it was not. Since it compiled properly, I immediately excluded syntax errors from my mental list of possible reasons and began to search for a semantic or logical error.

After a while, I remembered that there is no such thing as post-increment or post-decrement operator, so why should there be a pre-decrement? Well, there isn’t. But, if there is no pre-decrement operator, why does –someNumber compile? Basically, the answer is pretty simple: To Python –someNumber is the same as -(-(someNumber)).

A working version of the above example could be:

for i in range(someRange):
  # lots of code here
  someNumber -= 1
  doSomething(someNumber)
  # even more code here

Use Boost’s Multi Index Container!

Boost’s multi index container is a very cool and useful piece of code. Make it a part of your toolbox. You can start slowly by replacing uses of std::set and std::multiset with simple boost::multi_index_containers.

Sometimes, after you have used a special library or other special programming tool for a job, you forget about it because you don’t have that specific use case anymore. Boost’s multi_index container could fall in this category because you don’t have to hold data in memory with the need to access it by different keys all the time.

Therefore, this post is intended to be a reminder for c++ programmers that there exists this pretty cool thing called boost::multi_index_container and that you can use it in more situations than you would think at first.

(If you’re already using it on a regular basis you may stop here, jump directly to the comments and tell us about your typical use cases.)

I remember when I discovered boost::multi_index_container I found it quite intimidating at first sight. All those templates that are used in sometimes weird ways can trigger that feeling if you are not a template metaprogramming specialist (i.e. haven’t yet read Andrei Alexandrescu’s book “Modern C++ Design” ).

But if you look at it after you fought your way through the documentation and after your unit test is green that tests your first example, it doesn’t look that complicated anymore.

My latest use case for boost::multi_index_container was data objects that should be sorted by two different date-times. (For dates and times we use boost::date_time, of course). At first, the requirement was to store the objects sorted by one date time. I used a std::set for that with a custom comparator. Everything was fine.

With changing requirements it became necessary to retrieve objects by another date time, too. I started to use another std::set with a different comparator but then I remembered that there was some cool container somewhere in boost for which you can define multiple indices ….

After I had set it up with the two date time indices, the code also looked much cleaner because in order to update one object with a new time stamp I could just call container->replace(…) instead of fiddling around with the std::set.

Furthermore, I noticed that setting up a boost::multi_index_container with a specific key makes it much clearer what you intend with this data structure than using a std::set with a custom comparator. It is not that much more typing effort, and you can practice template metaprogramming a little bit 🙂

Let’s compare the two implementations:

#include <boost/shared_ptr.hpp>
#include <boost/date_time/posix_time/posix_time.hpp>
using boost::posix_time::ptime;

// objects of this class should be stored
class MyDataClass
{
  public:
    const ptime& getUpdateTime() const;
    const ptime& getDataChangedTime() const;

  private:
    ptime _updateTimestamp;
    ptime _dataChangedTimestamp;
};
typedef boost::shared_ptr<MyDataClass> MyDataClassPtr;

Now the definition of a multi index container:

#include <boost/multi_index_container.hpp>
#include <boost/multi_index/ordered_index.hpp>
#include <boost/multi_index/mem_fun.hpp>
using namespace boost::multi_index;

typedef multi_index_container
<
  MyDataClassPtr,
  indexed_by
  <
    ordered_non_unique
    <
      const_mem_fun<MyDataClass, 
        const ptime&, 
        &MyDataClass::getUpdateTime>
    >
  >
> MyDataClassContainer;

compared to std::set:

#include <set>

// we need a comparator first
struct MyDataClassComparatorByUpdateTime
{
  bool operator() (const MyDataClassPtr& lhs, 
                   const MyDataClassPtr& rhs) const
  {
    return lhs->getUpdateTime() < rhs->getUpdateTime();
  }
};
typedef std::multiset<MyDataClassPtr, 
                      MyDataClassComparatorByUpdateTime> 
   MyDataClassSetByUpdateTime;

What I like is that the typedef for the multi index container reads almost like a sentence. Besides, it is purely declarative (as long as you get away without custom key extractors), whereas with std::multiset you have to implement the comparator.

In addition to being a reminder, I hope this post also serves as motivation to get to know boost::multi_index_container and to make it a part of your toolbox. If you still have fears of contact, start small by replacing usages of std::set/multiset.

Performance Hogs Sometimes Live in Most Unexpected Places

Surprises when measuring performance are common – but sometimes you just can’t believe it.

When we develop software we always apply the best practice of not optimizing prematurely. This plays together with other best practices like writing the most readable code, or YAGNI.

‘Premature’ means different things in different situations. If you don’t have performance problems it means that there is absolutely no point in optimizing code. And if you do have performance problems it means that Thou Shalt Never Guess which code to optimize because software developers are very bad at this. The keyword here is profiling.

Since we don’t like to be “very bad” at something we always try to improve our skills in this field. The skill of guessing which code has to be optimized, or “profiling in your head” is no different in this regard.

So most of the times in profiling sessions, I have a few unspoken guesses at which parts of the code the profiler will point me to. Unfortunately, I have to say that  I am very often very surprised by the outcome.

Surprises in performance fixing sessions are common but they are of different quality. One rather BIG surprise was to find out that std::string::find of the C++ standard library is significantly slower (by factor > 10) than its C library counterpart strstr (discovered with gcc-4.4.6 on CentOS 6, verified with eglibc-2.13 and gcc-4.7).

Yes, you read right and you may not believe it. That was my reaction, too, so I wrote a little test program containing only two strings and calls to std::string::find and std::strstr, respectively. The results were – and I’ve no problem repeating myself here – a BIG surprise.

The reason for that is that std::strstr uses a highly optimized string matching algorithm version whereas std::string::find works with straight-forward memory comparison.

So when doing profiling sessions, always be prepared for shaking-your-world-view kind of surprises. They can even come from your beloved and highly regarded standard library.

UPDATE: See this stackoverflow question for more information.