Difference between revisions of "NewType"
Jump to navigation
Jump to search
(newtype and data comparison.) |
Benmachine (talk | contribs) (redirect to Newtype (if I've remembered syntax correctly)) |
||
| Line 1: | Line 1: | ||
| + | #REDIRECT [[Newtype]] |
||
| − | newtype is not exactly like data. #haskell is frequently being questioned about the difference. |
||
| − | |||
| − | Firstly, as mentioned in most tutorials, newtype can handle only a single data constructor, and a single field in that single data constructor. |
||
| − | |||
| − | Secondly, newtype is strict on its field. |
||
| − | |||
| − | So how is: |
||
| − | |||
| − | <code>newtype NewType = NewType Int</code> |
||
| − | |||
| − | different from: |
||
| − | |||
| − | <code>data NewData = NewData Int</code> |
||
| − | |||
| − | The number of bottoms differ. Since data is lazy, it introduces a new bottom in addition to the bottom that existed in the Int itself. |
||
| − | |||
| − | NewData can be _|_ or it can be (NewData _|_). NewType cannot be _|_ but can only be (NewType _|_). |
||
| − | |||
| − | So, lets try another comparison: |
||
| − | |||
| − | How is: |
||
| − | |||
| − | <code>newtype NewType = NewType Int</code> |
||
| − | |||
| − | different from: |
||
| − | |||
| − | <code>data NewData = NewData !Int</code> |
||
| − | |||
| − | Here, the number of bottoms is the same, but they are not quite the same bottoms. NewType's bottom is: (NewType _|_) (the bottom is <b>inside</b> the pattern), whereas NewData's bottom is (_|_) (the bottom is the <b>whole</b> pattern). |
||
| − | |||
| − | To demonstrate this difference, we can use the following code; |
||
| − | <code> |
||
| − | case (NewType undefined) of |
||
| − | (NewType _) -> "Happy happy joy joy :-)" |
||
| − | case (NewData undefined) of |
||
| − | (DataType _) -> "Won't ever get here :-(" |
||
| − | </code> |
||
| − | |||
| − | Since the newtype bottom is inside the pattern, and we catch that with (_) and never use it, we don't actually hit the bottom. But the NewData bottom is on the whole pattern, so even matching the single data constructor that way is a bottom. |
||
Revision as of 22:00, 29 September 2010
Redirect to: