Trying to add automatic radix detection in std.conv.parse
hello, I'd like to test if it's possible to add an automatic
radix detection in std.conv, parse.
I've added...
if (s.length >= 2)
{
if (s[0..2] == "0x" || s[0..2] == "0X")
{
uint radix = 16;
Source nbr = s[2..$];
return .parse!Target(nbr, radix);
}
else if (s[0..2] == "0b" || s[0..2] == "0B")
{
uint radix = 2;
Source nbr = s[2..$];
return .parse!Target(nbr, radix);
}
}
...to the first parse version. The (maybe `naive`) idea is to
redirect to the right parse version if the prefixes are found.
But it seems to break the purity of format() because I get , when
compiling phobos, the following errors:
std\utf.d(71): Error: pure function 'std.utf.UTFException.this'
cannot call impure function 'std.str
ing.format!(char, uint).format'
std\utf.d(71): Error: safe function 'std.utf.UTFException.this'
cannot call system function 'std.str
ing.format!(char, uint).format'
std\uuid.d(1536): Error: pure function
'std.uuid.UUIDParsingException.this' cannot call impure funct
ion 'std.string.format!(char, string, string, uint).format'
What's wrong ?
Re: Converting a POD struct to a class at compile-time ?
On Wednesday, 16 July 2014 at 18:09:10 UTC, Adam D. Ruppe wrote: On Wednesday, 16 July 2014 at 17:43:03 UTC, Klb wrote: auto names = __traits(allMembers, S); Error: static variable _names_field_0 cannot be read at compile time. The problem there is names is a regular runtime variable and mixins need to use compile time stuff. If you make that enum instead of auto, it'll probably work. Unfortunately I can't encapsulate and alias. Behind the question stands another idea: I do something with an interface implementer, as the "something" is not compat. with structs the idea was to generate a class as a string, with the interface and its default method...
Re: Converting a POD struct to a class at compile-time ?
On Wednesday, 16 July 2014 at 18:09:10 UTC, Adam D. Ruppe wrote: On Wednesday, 16 July 2014 at 17:43:03 UTC, Klb wrote: auto names = __traits(allMembers, S); Error: static variable _names_field_0 cannot be read at compile time. The problem there is names is a regular runtime variable and mixins need to use compile time stuff. If you make that enum instead of auto, it'll probably work. Nice ! It works. :)
Converting a POD struct to a class at compile-time ?
Hello, I'd like to know if it's possible, using CTFE, mixin etc
to convert a POD struct to a class at compile time.
I've reached the step where the string to mix-in is generated but
I cant mix it:
-
import std.stdio;
import std.traits, std.typetuple;
static private template genClassFromStruct(S) if (is(S == struct)
&(__traits(isPOD, S)))
{
auto values = S.init.tupleof;
auto types = typeid(typeof(values));
auto names = __traits(allMembers, S);
string genClassFromStruct()
{
string members;
foreach(int i,t; RepresentationTypeTuple!S)
{
members ~= t.stringof ~ " " ~ names[i] ~ ";\r";
}
return
"class c" ~ S.stringof ~
"{" ~ ";\r"~
members ~
"}";
}
}
struct foo{ int a; float b;}
//mixin( genClassFromStruct!foo );
void main(string args[])
{
foo Foo;
//auto c = new cfoo;
writeln( genClassFromStruct!foo );
}
-
The problem appends when un-commenting the mixin:
Error: static variable _names_field_0 cannot be read at compile
time.
what is exactly stack stomp "-gx" new switch ?
...and any example where this switch will be usefull ?
Re: syntax for calling to with a getter as source argument
On Tuesday, 15 July 2014 at 00:19:15 UTC, Ali Çehreli wrote:
On 07/14/2014 05:10 PM, Ali Çehreli wrote:
> On 07/14/2014 04:04 PM, H. S. Teoh via Digitalmars-d-learn
wrote:
>
> > On Mon, Jul 14, 2014 at 09:12:30PM +0000, Klb via
Digitalmars-d-learn
> wrote:
> >> hello what is the right syntax for this:
> >>
> >>
> >> import std.stdio, std.conv;
> >>
> >> void main(string args[])
> >> {
> >> ubyte[3] src = [0, 1, 2];
> >> string trg = "";
> >>
> >> @property ubyte[3] srcAsProp(){return src;}
> >>
> >> // Error: template std.conv.to cannot deduce
function from
> argument types
> >> // !(string)(ubyte[3]), candidates are:
> >> trg = to!(string)(srcAsProp());
> >> }
> >>
> >>
> >> In a real-world application I'd use an intermediate value
but I'd like
> >> to know If it's possible...The strange fact is that it
doesn't trig an
> >> error if src is a dyn. array. (if you remove the 3 from
[3] then it
> >> passes).
> >
> > You need to slice the static array:
> >
> > trg = to!string(srcAsProp()[]);
>
> There seems to be an attempt in Phobos to support it without
needing an
> explicit slice.
>
> From std/phobos/conv.d:
I meant phobos/std/conv.d there.
> /*
>Converting static arrays forwards to their dynamic
counterparts.
> */
> T toImpl(T, S)(ref S s)
> if (isRawStaticArray!S)
Ok, I think I see now (after two minutes after posting it :) )
why it doesn't work. The overload above takes by-ref, which
does not bind to rvalues in D.
And the problem is, what srcAsProp() returns is an rvalue
because static arrays are value-types and they get copied.
> {
> return toImpl!(T, typeof(s[0])[])(s);
> }
>
> @safe pure nothrow unittest
> {
> char[4] test = ['a', 'b', 'c', 'd'];
> static assert(!isInputRange!(Unqual!(char[4])));
> assert(to!string(test) == test);
> }
>
> I don't know why that overload does not take care of OP's
case. (I am be
> completely off here. :) )
I hope I have it now. :p
Ali
Thanks for this accurate explanation. The problem is quite clear
now.
syntax for calling to with a getter as source argument
hello what is the right syntax for this:
import std.stdio, std.conv;
void main(string args[])
{
ubyte[3] src = [0, 1, 2];
string trg = "";
@property ubyte[3] srcAsProp(){return src;}
// Error: template std.conv.to cannot deduce function from
argument types
// !(string)(ubyte[3]), candidates are:
trg = to!(string)(srcAsProp());
}
In a real-world application I'd use an intermediate value but I'd
like to know If it's possible...The strange fact is that it
doesn't trig an error if src is a dyn. array. (if you remove the
3 from [3] then it passes).
