Skip to main content
Software Engineering

Return to Answer

deleted 13 characters in body
Source Link
Philipp
Philipp
  • 23.5k
  • 6
  • 65
  • 69

Using dedicated types has the advantage that it's much harder to screw up the order of arguments. The first version of your example function, for example, begins with a chain of 9 doubles. When someone uses this function, it is very easy to screw up the order and pass the gimbal angles in place of the GPS coordinates and vice versa. This can lead to very strange and hard to trace bugs. When you pass only three arguments with different types instead, this error can be catched by the compiler.

I would in fact consider to go a step further and introduce types which are technically identical but have a different semantic meaning. For example by having a class PlaneAngle3d and a separate class GimbalAngle3d, even though both are a collection of three doubles. This would make it impossible to accidentally use one as the other, unless it's intentional.

I would recommend using typedefs which are just aliases for primitive types (typedef double ZoomLevel) not just for that reason, but also for another: It easily allows you to change the type later. When you get the idea one day that using float could be just as good as double but could be more memory- and CPU-efficient, you just have to change this in one place, and not in dozends.

Using dedicated types has the advantage that it's much harder to screw up the order of arguments. The first version of your example function, for example, begins with a chain of 9 doubles. When someone uses this function, it is very easy to screw up the order and pass the gimbal angles in place of the GPS coordinates and vice versa. This can lead to very strange and hard to trace bugs. When you pass only three arguments with different types instead, this error can be catched by the compiler.

I would in fact consider to go a step further and introduce types which are technically identical but have a different semantic meaning. For example by having a class PlaneAngle3d and a separate class GimbalAngle3d, even though both are a collection of three doubles. This would make it impossible to accidentally use one as the other, unless it's intentional.

I would recommend using typedefs which are just aliases for primitive types (typedef double ZoomLevel) not just for that reason, but also for another: It easily allows you to change the type later. When you get the idea one day that using float could be just as good as double but could be more memory- and CPU-efficient, you just have to change this in one place, and not in dozends.

Using dedicated types has the advantage that it's much harder to screw up the order of arguments. The first version of your example function, for example, begins with a chain of 9 doubles. When someone uses this function, it is very easy to screw up the order and pass the gimbal angles in place of the GPS coordinates and vice versa. This can lead to very strange and hard to trace bugs. When you pass only three arguments with different types instead, this error can be catched by the compiler.

I would in fact consider to go a step further and introduce types which are technically identical but have a different semantic meaning. For example by having a class PlaneAngle3d and a separate class GimbalAngle3d, even though both are a collection of three doubles. This would make it impossible to use one as the other, unless it's intentional.

I would recommend using typedefs which are just aliases for primitive types (typedef double ZoomLevel) not just for that reason, but also for another: It easily allows you to change the type later. When you get the idea one day that using float could be just as good as double but could be more memory- and CPU-efficient, you just have to change this in one place, and not in dozends.

Source Link
Philipp
Philipp
  • 23.5k
  • 6
  • 65
  • 69

Using dedicated types has the advantage that it's much harder to screw up the order of arguments. The first version of your example function, for example, begins with a chain of 9 doubles. When someone uses this function, it is very easy to screw up the order and pass the gimbal angles in place of the GPS coordinates and vice versa. This can lead to very strange and hard to trace bugs. When you pass only three arguments with different types instead, this error can be catched by the compiler.

I would in fact consider to go a step further and introduce types which are technically identical but have a different semantic meaning. For example by having a class PlaneAngle3d and a separate class GimbalAngle3d, even though both are a collection of three doubles. This would make it impossible to accidentally use one as the other, unless it's intentional.

I would recommend using typedefs which are just aliases for primitive types (typedef double ZoomLevel) not just for that reason, but also for another: It easily allows you to change the type later. When you get the idea one day that using float could be just as good as double but could be more memory- and CPU-efficient, you just have to change this in one place, and not in dozends.