C++

Determine endianness
The easiest is to determine if the targetted architecture is big- or little-endian at compile time. On linux this is straightforward: just include /usr/include/endian.h. You can then use the defined variables __BYTE_ORDER, __LITTLE_ENDIAN and __BIG_ENDIAN. For example:

Convert a rgba image to grayscale

// read 4 bytes at once
unsigned int * p = reinterpret_cast(some_image);
unsigned int * end = p + width*height;
// the resulting gray scale image
unsigned char out_image[width*height]
unsigned char * p_out = out_image;
for(; p != end; ++p)
{
   unsigned int fourBytes = *p;
#if __BYTE_ORDER == __LITTLE_ENDIAN
   b = fourBytes >> 16;
   g = fourBytes >> 8;
   r = fourBytes;
#elif __BYTE_ORDER == __BIG_ENDIAN
   r = fourBytes >> 24;
   g = fourBytes >> 16;
   b = fourBytes >> 8;
#endif
   *p_out = (r >> 2) + (g >> 1) + (b >> 2);
   p_out++;
}

Singleton pattern
Imagine that your system depends on a set of parameters that, e.g. are read from a configuration file at start, or entered via a GUI, etc. You can put all your configuration parameters in an object. A nice pattern to use in that case is the singleton, which guarantees that there is only single config object at any time during the life of the program.

class Config
{
public:
  static Config & GetConfig()
  {
     static Config c;
     return c;
  }

// some more methods here
// ...

private:
// make the cstor, copy and assignement operator private
  Config()   {}
  Config( const Config & );
  void operator=( const Config & );
};


// then anywhere in your code you can write
Config & MyConfig = Config::GetConfig();
// and MyConfig will always refer to the same object.
MyConfig.SetParameter(...)