View mode: basic / threaded / horizontal-split · Log in · Help
September 27, 2009
Re: Dispatching on a variant
Christopher Wright wrote:
> Justin Johansson wrote:
>> I've had a good poke around the forums and couldn't find anything on 
>> this so ...
>>
>> What's the recommended method for dispatching code off the runtime 
>> type of a variant variable
>> (Phobos D2 std.variant)?
>>
>> Does one use a bunch of
>>
>> if ( var.peek!(type1)) { ... }
>> else if ( var.peek!(type2)  { ... }
>>
>> for all N possible types, or is there a better & faster way with a 
>> switch or jump table of sorts?
> 
> Variant should have an accessible typeinfo property.

It does, see member function type().

Andrei
September 27, 2009
Re: Dispatching on a variant
Christopher Wright wrote:
> Justin Johansson wrote:
>> I've had a good poke around the forums and couldn't find anything on 
>> this so ...
>>
>> What's the recommended method for dispatching code off the runtime 
>> type of a variant variable
>> (Phobos D2 std.variant)?
>>
>> Does one use a bunch of
>>
>> if ( var.peek!(type1)) { ... }
>> else if ( var.peek!(type2)  { ... }
>>
>> for all N possible types, or is there a better & faster way with a 
>> switch or jump table of sorts?
> 
> Variant should have an accessible typeinfo property. (When I say should, 
> I mean that it would be appropriate for it to have it. I am not saying 
> that I believe it has it.) It should be faster to compare that (or 
> switch on typeinfo.name) than to peek for each type.

I wouldn't make that rely on string switch. While it's a very convenient 
feature to have in D that's usually only seen in scripting, its also a 
*slow* one when compared to integral switch.

string switch actually walks the case strings and compares with the 
source string until it finds a match, a binary search is much faster if 
you don't care about the order of the tests.

Also if someone strips his executable from all names in TypeInfo and 
ClassInfo to prevent people from peeking into his internals, it breaks 
the code. Some companies out there goes to great lengths to make their 
executables as hard to reverse engineer as possible.

A good alternative would be for the algebraic type to define an internal 
enum which is used for switch statements:

alias Algebraic!(int, void*, Object) MyType;

switch(myType.typeId) {
case MyType.INT:
case MyType.VOIDPTR:
case MyType.OBJECT:
}

While not perfect it definitely is fast.
September 27, 2009
Re: Dispatching on a variant
On Sat, Sep 26, 2009 at 11:16 PM, Jeremie Pelletier <jeremiep@gmail.com> wrote:

> string switch actually walks the case strings and compares with the source
> string until it finds a match, a binary search is much faster if you don't
> care about the order of the tests.

FWIW the runtime does perform a binary search of the strings in a
switch. The compiler outputs the string table as a sorted list.
September 27, 2009
Re: Dispatching on a variant
Jarrett Billingsley wrote:
> On Sat, Sep 26, 2009 at 11:16 PM, Jeremie Pelletier <jeremiep@gmail.com> wrote:
> 
>> string switch actually walks the case strings and compares with the source
>> string until it finds a match, a binary search is much faster if you don't
>> care about the order of the tests.
> 
> FWIW the runtime does perform a binary search of the strings in a
> switch. The compiler outputs the string table as a sorted list.

Oh, you're right! I just checked in druntime.
Next ›   Last »
1 2 3
Top | Discussion index | About this forum | D home