Re:My thoughts on Google Go
by dkersten on Thursday November 12, 2009 @07:06AM
(#937584)
I assume its because they don't want references everywhere. For example, allowing the programmer to choose if an object is allocated on the stack or heap, passed by reference (using a pointer) or by value.
[ Parent
| on HN ]
Re:My thoughts on Google Go
by euroclydon on Thursday November 12, 2009 @09:17AM
(#937776)
Can you elaborate on when an object is "allocated on the stack" versus being "allocated on the heap" in Go?
The echo program from the tutorial[1], line 21:
var s string = "";
Now, I don't see it written there that it is on the stack, but I imagine so. From the source code of 6g, in cgen.c[2], line 884, you can see this comment:
/*
* n is on stack, either local variable
* or return value from function call.
* return n's offset from SP.
*/
Sounds to me that local variables and return values are always on the stack, which makes sense to me.[1] http://golang.org/doc/go_tutorial.html#tmp_44
[2] http://code.google.com/p/go/source/browse/src/cmd/6g/cgen.c?...
I cant, since I'm not terribly familiar with Go. Like I said, I assume that this is the reasoning behind supporting pointers, rather than references to everything. I did not actually check.
Ah, of course.
Although, gotta wonder how these stack-allocated objects interact with closures.
Other languages that support closures have stack allocated objects too. So its no different in this case.
Maybe this wikipedia article has details: http://en.wikipedia.org/wiki/Thunk
C-style stack allocated objects captured by closures are not memory-safe in the absence of escape analysis. If the closure outlives the stack frame that allocated the object, a dangling pointer may occur, unless the stack frame itself is allocated on the heap, or the captured objects are moved by the compiler into a structure which is itself allocated on the heap.