digitalmars.D.learn - Template type deduction
- Phil Lavoie (13/13) Jan 20 2013 Imagine you have something like this:
- Philippe Sigaud (4/15) Jan 20 2013 This should work:
- Phil Lavoie (2/26) Jan 20 2013 And it does buddy, thanks!
- Phil Lavoie (15/39) Jan 20 2013 Is there also a way to extract per parameter information (ex:
- Philippe Sigaud (29/40) Jan 20 2013 Sure, the U... part is accessible inside the static if. Just define a
- Phil Lavoie (23/71) Jan 20 2013 YES, it works! I just didn't realized you could only use it
- Andrej Mitrovic (7/13) Jan 20 2013 Use isInstanceOf in std.traits:
- Phil Lavoie (2/21) Jan 20 2013 Nice, I am glad to learn it's in the stdlib.
- Philippe Sigaud (3/6) Jan 20 2013 I just rebuild DMD 2.062 from git head, not there yet... Oh well.
- Andrej Mitrovic (4/10) Jan 20 2013 My bad, it wasn't pulled yet:
- Philippe Sigaud (6/18) Jan 20 2013 What? No need to apologize! It's just a small script I launch once or
- Andrej Mitrovic (5/9) Jan 20 2013 Is this something taken from the DWiki? If not, it might be a good
- Philippe Sigaud (8/17) Jan 20 2013 I just use the dlang-workspace script that was posted some weeks ago,
Imagine you have something like this:
struct MyParamType( T1, T2 ) { ... }
and then a compound:
struct MyCoumpound( T ) if( isMyParamType!T ) { ... }
template isMyParamType( T ) {
enum isMyParamType = ... //Complete this
}
Is it possible to deduce that T is a parameterized struct using
isExpressions:
enum isMyParamType = is( T U: MyParamType!U ); //Works for 1
parameter, but what do you do when you have plenty?
Thanks!
Phil
Jan 20 2013
On Sun, Jan 20, 2013 at 3:59 PM, Phil Lavoie <maidenphil hotmail.com> wrote:
Imagine you have something like this:
struct MyParamType( T1, T2 ) { ... }
and then a compound:
struct MyCoumpound( T ) if( isMyParamType!T ) { ... }
template isMyParamType( T ) {
enum isMyParamType = ... //Complete this
}
Is it possible to deduce that T is a parameterized struct using
isExpressions:
enum isMyParamType = is( T U: MyParamType!U ); //Works for 1 parameter, but
what do you do when you have plenty?
This should work:
is(T _ : MyParamType!U, U...)
( '_' just to tell someone reading the code this part is not used)
Jan 20 2013
"Phil Lavoie" <maidenphil hotmail.com> On Sunday, 20 January 2013 at 15:39:53 UTC, Philippe Sigaud wrote:On Sun, Jan 20, 2013 at 3:59 PM, Phil Lavoie <maidenphil hotmail.com> wrote:And it does buddy, thanks!Imagine you have something like this: struct MyParamType( T1, T2 ) { ... } and then a compound: struct MyCoumpound( T ) if( isMyParamType!T ) { ... } template isMyParamType( T ) { enum isMyParamType = ... //Complete this } Is it possible to deduce that T is a parameterized struct using isExpressions: enum isMyParamType = is( T U: MyParamType!U ); //Works for 1 parameter, but what do you do when you have plenty?This should work: is(T _ : MyParamType!U, U...) ( '_' just to tell someone reading the code this part is not used)
Jan 20 2013
On Sunday, 20 January 2013 at 15:39:53 UTC, Philippe Sigaud wrote:On Sun, Jan 20, 2013 at 3:59 PM, Phil Lavoie <maidenphil hotmail.com> wrote:Is there also a way to extract per parameter information (ex: type) using type deduction. Let me illustrate: struct Toto( T1, T2 ) { alias Key = T1; alias Value = T2; } struct OuterToto( T ) if( isToto!T ) { T.Key key; T.Value value; } Imagine I would want that but without using inner aliases, is this possible? Thanks! PhilImagine you have something like this: struct MyParamType( T1, T2 ) { ... } and then a compound: struct MyCoumpound( T ) if( isMyParamType!T ) { ... } template isMyParamType( T ) { enum isMyParamType = ... //Complete this } Is it possible to deduce that T is a parameterized struct using isExpressions: enum isMyParamType = is( T U: MyParamType!U ); //Works for 1 parameter, but what do you do when you have plenty?This should work: is(T _ : MyParamType!U, U...) ( '_' just to tell someone reading the code this part is not used)
Jan 20 2013
On Sun, Jan 20, 2013 at 5:01 PM, Phil Lavoie <maidenphil hotmail.com> wrote:
Is there also a way to extract per parameter information (ex: type) using
type deduction. Let me illustrate:
struct Toto( T1, T2 ) {
alias Key = T1;
alias Value = T2;
}
struct OuterToto( T ) if( isToto!T ) {
T.Key key;
T.Value value;
}
Imagine I would want that but without using inner aliases, is this possible?
Sure, the U... part is accessible inside the static if. Just define a
generic helper template:
struct Toto(First, Second)
{
}
template TemplateArguments(T)
{
static if (is(T _ : a!(Args), alias a, Args...))
alias TemplateArguments = Args;
else
alias TemplateArguments = void;
}
template TemplateName(T)
{
static if (is(T _ : a!(Args), alias a, Args...))
alias TemplateName = a;
else
alias TemplateName = void;
}
void main(string[] args)
{
alias Type = Toto!(int, double);
writeln(TemplateArguments!(Type).stringof);
writeln(TemplateName!(Type).stringof);
alias T = TemplateName!(Type);
alias Swapped = T!(double, int); // And you can use T as a template.
}
As you can see, you can even use the deduced template name.
Jan 20 2013
On Sunday, 20 January 2013 at 16:11:08 UTC, Philippe Sigaud wrote:On Sun, Jan 20, 2013 at 5:01 PM, Phil Lavoie <maidenphil hotmail.com> wrote:YES, it works! I just didn't realized you could only use it inside the static ifs, thanks for pointing that out! isInstanceOf( alias Template, T ) is nice but lacks support for expressions directly: but that ain't a big deal, example: struct Pair( T1, T2 ) {...} template isPair( alias T ) { static if( is( T ) ) { enum isPair = isInstanceOf!( Pair, T ); } else { enum isPair = isInstanceOf!( Pair, typeof( T ) ); //Must add this to support expressions. } } //Returns the type of the first term of the pair. template FirstType( alias T ) if( isPair!T ) { static if( is( T _: Pair!Args, Args... ) ) { alias FirstType = Args[ 0 ]; } else { alias FirstType = FirstType!( typeof( T ); } } The D programming language ROCKS :).Is there also a way to extract per parameter information (ex: type) using type deduction. Let me illustrate: struct Toto( T1, T2 ) { alias Key = T1; alias Value = T2; } struct OuterToto( T ) if( isToto!T ) { T.Key key; T.Value value; } Imagine I would want that but without using inner aliases, is this possible?Sure, the U... part is accessible inside the static if. Just define a generic helper template: struct Toto(First, Second) { } template TemplateArguments(T) { static if (is(T _ : a!(Args), alias a, Args...)) alias TemplateArguments = Args; else alias TemplateArguments = void; } template TemplateName(T) { static if (is(T _ : a!(Args), alias a, Args...)) alias TemplateName = a; else alias TemplateName = void; } void main(string[] args) { alias Type = Toto!(int, double); writeln(TemplateArguments!(Type).stringof); writeln(TemplateName!(Type).stringof); alias T = TemplateName!(Type); alias Swapped = T!(double, int); // And you can use T as a template. } As you can see, you can even use the deduced template name.
Jan 20 2013
On 1/20/13, Phil Lavoie <maidenphil hotmail.com> wrote:Imagine you have something like this: struct MyParamType( T1, T2 ) { ... } and then a compound: struct MyCoumpound( T ) if( isMyParamType!T ) { ... }Use isInstanceOf in std.traits: struct MyCoumpound( T ) if( isInstanceOf!(MyParamType, T) ) { ... } On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:is(T _ : MyParamType!U, U...) ( '_' just to tell someone reading the code this part is not used)This won't be necessary in the next release thanks to a pull by Kenji, you will be able to use: is(T : MyParamType!U, U...)
Jan 20 2013
On Sunday, 20 January 2013 at 15:59:45 UTC, Andrej Mitrovic wrote:On 1/20/13, Phil Lavoie <maidenphil hotmail.com> wrote:Nice, I am glad to learn it's in the stdlib.Imagine you have something like this: struct MyParamType( T1, T2 ) { ... } and then a compound: struct MyCoumpound( T ) if( isMyParamType!T ) { ... }Use isInstanceOf in std.traits: struct MyCoumpound( T ) if( isInstanceOf!(MyParamType, T) ) { ... } On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:is(T _ : MyParamType!U, U...) ( '_' just to tell someone reading the code this part is not used)This won't be necessary in the next release thanks to a pull by Kenji, you will be able to use: is(T : MyParamType!U, U...)
Jan 20 2013
On Sun, Jan 20, 2013 at 4:59 PM, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:This won't be necessary in the next release thanks to a pull by Kenji, you will be able to use: is(T : MyParamType!U, U...)I just rebuild DMD 2.062 from git head, not there yet... Oh well.
Jan 20 2013
On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:On Sun, Jan 20, 2013 at 4:59 PM, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:My bad, it wasn't pulled yet: https://github.com/D-Programming-Language/dmd/pull/1255 Apologies for taking your time on building DMD.This won't be necessary in the next release thanks to a pull by Kenji, you will be able to use: is(T : MyParamType!U, U...)I just rebuild DMD 2.062 from git head, not there yet... Oh well.
Jan 20 2013
On Sun, Jan 20, 2013 at 5:23 PM, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:What? No need to apologize! It's just a small script I launch once or twice a week. It clones the github repos and builds everything. I just type `bash regenerate_dmd.sh` and wait a few minutes while GCC compiles happily.On Sun, Jan 20, 2013 at 4:59 PM, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:My bad, it wasn't pulled yet: https://github.com/D-Programming-Language/dmd/pull/1255 Apologies for taking your time on building DMD.This won't be necessary in the next release thanks to a pull by Kenji, you will be able to use: is(T : MyParamType!U, U...)I just rebuild DMD 2.062 from git head, not there yet... Oh well.
Jan 20 2013
On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:What? No need to apologize! It's just a small script I launch once or twice a week. It clones the github repos and builds everything. I just type `bash regenerate_dmd.sh` and wait a few minutes while GCC compiles happily.Is this something taken from the DWiki? If not, it might be a good idea to amend the current Posix build docs and maybe add a link to your script: http://wiki.dlang.org/Building_DMD I use something similar on win32 where I can build DMD-git very quickly.
Jan 20 2013
On Sun, Jan 20, 2013 at 7:33 PM, Andrej Mitrovic <andrej.mitrovich gmail.com> wrote:On 1/20/13, Philippe Sigaud <philippe.sigaud gmail.com> wrote:I just use the dlang-workspace script that was posted some weeks ago, wrapped in another script to pull the latest dlang-workspace, should it ever change. It's all hardcoded, because I'm just discovering bash scripts. So, nothing that should be on a wiki :) The dlang-workspace/posix/gen.sh script itself, which does all the hard work, could maybe be posted. But I'm not its author...What? No need to apologize! It's just a small script I launch once or twice a week. It clones the github repos and builds everything. I just type `bash regenerate_dmd.sh` and wait a few minutes while GCC compiles happily.Is this something taken from the DWiki? If not, it might be a good idea to amend the current Posix build docs and maybe add a link to your script: http://wiki.dlang.org/Building_DMD I use something similar on win32 where I can build DMD-git very quickly.
Jan 20 2013