3 ms·

Can someone explain why lambda expressions are being picked out?. I.e. why not say 'An expression is not necessarily a closure' as wouldn't this stuff apply to

Can someone explain why lambda expressions are being picked out?. I.e. why not say 'An expression is not necessarily a closure' as wouldn't this stuff apply to any old expressions, functions etc. I am not a lisp person but thinking of this in terms of say C# or Haskell.

dllthomas 11y ago
Because lambdas are sometimes referred to as closures even where they don't close over anything. Other expressions generally aren't.
Turing_Machine 11y ago
They're being picked out because you see people do that all the time (i.e., calling an anonymous function a "closure"), while I don't recall ever seeing someone referring to a random expression as a closure. I'm not overly fond of calling anonymous functions "lambdas", for that matter. Sure, several languages use the term lambda, but many do not. More specifically, "anonymous function" is much more descriptive than "lambda".
- nailer 11y ago
  Swift even does this officially: ALL lambdas are called closures, even though there's nothing being closed over. Similarly, people refer to constructor functions that always return objects of the same class/prototype as factories, despite both the GoF patterns that popularised the term showing dynamic output. IIRC Angular JS misuses the term this way too. AFAICT it's either genuine ignorance of how/why these words describe or some kind of weird demonstration of cleverness by using a particular term, regardless of whether it's applicable or not.
  - Ono-Sendai 11y ago
    It's probably because, in an implementation of a staticly typed language, you have to treat all functions being passed around as closures. E.g. the type of a lambda expression needs to be one thing, and that is a closure. Whether or not any variables are actually closed over is irrelevant.
    - nailer 11y ago
      Couldn't the type of a passed around function simply be 'function'? Why does it have to be inaccurate?
      - Ono-Sendai 11y ago
        Well a function may in general have closed over arguments. Therefore it needs to be a closure. (If you have lexical scoping that is)
sklogic 11y ago
Lambda is easy to reason about, since it is guaranteed to not have any side effects. To juggle the other kinds of expressions like this you'd need a more elaborate purity analysis. So, lambda is a very easy prey for the DCE. Also, there are some specific optimisations, like beta-reduction and closure inlining.

dllthomas 11y ago

Because lambdas are sometimes referred to as closures even where they don't close over anything. Other expressions generally aren't.

Turing_Machine 11y ago

They're being picked out because you see people do that all the time (i.e., calling an anonymous function a "closure"), while I don't recall ever seeing someone referring to a random expression as a closure. I'm not overly fond of calling anonymous functions "lambdas", for that matter. Sure, several languages use the term lambda, but many do not. More specifically, "anonymous function" is much more descriptive than "lambda".

nailer 11y ago
Swift even does this officially: ALL lambdas are called closures, even though there's nothing being closed over. Similarly, people refer to constructor functions that always return objects of the same class/prototype as factories, despite both the GoF patterns that popularised the term showing dynamic output. IIRC Angular JS misuses the term this way too. AFAICT it's either genuine ignorance of how/why these words describe or some kind of weird demonstration of cleverness by using a particular term, regardless of whether it's applicable or not.
- Ono-Sendai 11y ago
  It's probably because, in an implementation of a staticly typed language, you have to treat all functions being passed around as closures. E.g. the type of a lambda expression needs to be one thing, and that is a closure. Whether or not any variables are actually closed over is irrelevant.
  - nailer 11y ago
    Couldn't the type of a passed around function simply be 'function'? Why does it have to be inaccurate?
    - Ono-Sendai 11y ago
      Well a function may in general have closed over arguments. Therefore it needs to be a closure. (If you have lexical scoping that is)

sklogic 11y ago

Lambda is easy to reason about, since it is guaranteed to not have any side effects. To juggle the other kinds of expressions like this you'd need a more elaborate purity analysis. So, lambda is a very easy prey for the DCE. Also, there are some specific optimisations, like beta-reduction and closure inlining.