On Tue, 15 Mar 2005 10:29:32 +0000 (UTC), AEon <AEon_member@pathlink.com> wrote:
> Derek Parnell wrote:
>
>> For example, you can also write ...
>>  writefln("Line: ", __LINE__);
>> or
>>  writefln("Line: %d", __LINE__);
>> to get the same results.
>
> Indeed. Both work, you can even do this (as was pointed out, writefln is *very*
> flexible), though the formatting would be less than optimal:
>
> // Line: test.d 96Mar 15 200511:17:12Tue Mar 15 11:17:12 2005
> writefln("Line: %d ", __FILE__,__LINE__,__DATE__,__TIME__,__TIMESTAMP__);
>
> // Line: test.d97Mar 15 200511:17:12Tue Mar 15 11:17:12 2005
> writefln("Line: ", __FILE__,__LINE__,__DATE__,__TIME__,__TIMESTAMP__);

writefln accepts the same formatting as printf, eg.

printf("Line: %s:%d %s %s %s\n",__FILE__,__LINE__,__DATE__,__TIME__,__TIMESTAMP__);
writefln("Line: %s:%d %s %s %s",__FILE__,__LINE__,__DATE__,__TIME__,__TIMESTAMP__);

should be identical. Or you could write it like:

writefln("Line: ",__FILE__,":",__LINE__," ",__DATE__," ",__TIME__," ",__TIMESTAMP__);

or even better (safer, no chance of stray %s causing problems)

writefln("Line: %s",__FILE__,":%d",__LINE__," %s",__DATE__," %s",__TIME__," %s",__TIMESTAMP__);

notice, writef<etc> can accept formatting strings anywhere.

> I did a full search in the dmd\ development tree for "__TIMESTAMP__", other than
> in the dmd, and dmd.exe files, the only source related mention of the latter is
> in
>
> E:\d\dmd\src\dmd\idgen.c
>
> struct Msgtable
> {
> char *ident;	// name to use in DMD source
> char *name;	// name in D executable
> };
>
> Msgtable msgtable[] =
> {
> { "LINE", "__LINE__" },
> { "FILE", "__FILE__" },
> { "DATE", "__DATE__" },
> { "TIME", "__TIME__" },
> { "TIMESTAMP", "__TIMESTAMP__" },
> ..
> }
>
> Would there have been any way to look up what type "__LINE__" is for printf?

I notice Anders has given an example, which should work, but appears to highlight a bug in the compiler.

> I
> had assumed that simply all those __ vars are strings (as we found out D Strings
> to be exact)?

I can see why you'd assume that. The only reason I knew __LINE__ was an 'int' is because it's an int in C/C++.

> PS: Any chance to fix this forum to actually respect indentations, to make the reading of source samples easier to read?

I suspect not, because it has been voiced before and nothing has happened.
The best you can do is prefix every line with a character i.e.

# void main() {
#     int i;

..etc..

Regan

On Wed, 16 Mar 2005 09:02:11 +1300, Regan Heath wrote:

[snip]

>> I
>> had assumed that simply all those __ vars are strings (as we found out D
>> Strings
>> to be exact)?
> 
> I can see why you'd assume that. The only reason I knew __LINE__ was an 'int' is because it's an int in C/C++.

Also it is documented  http://www.digitalmars.com/d/lex.html#specialtokens

-- 
Derek Parnell
Melbourne, Australia
16/03/2005 7:12:42 AM

>> import std.stdio;
>> void main()
>> {
>>   writefln("__FILE__ ", typeid(typeof(__FILE__)));
>>   writefln("__LINE__ ", typeid(typeof(__LINE__)));
>>   writefln("__DATE__ ", typeid(typeof(__DATE__)));
>>   writefln("__TIME__ ", typeid(typeof(__TIME__)));
>> }
> 
> *Should* print the types, but:
> 
> __FILE__ TypeInfo
> __LINE__ long
> __DATE__ TypeInfo
> __TIME__ TypeInfo
> 
> It seems broken for char[]... ?

Duh, forgot that string literals are of an unnamed type...
They're converted into char[]/wchar[]/dchar[] from context.

In this case, there was no context, so they renamed unnamed.

> import std.stdio;
> void main()
> {
>   writefln("%s", typeid(typeof("")));
>   writefln("%s", typeid(typeof(cast( char[]) "")));
>   writefln("%s", typeid(typeof(cast(wchar[]) "")));
>   writefln("%s", typeid(typeof(cast(dchar[]) "")));
> }

TypeInfo
char[]
wchar[]
dchar[]

Not a bug.

*However*, string literals in D should all be zero-terminated...
These new ones are not, and that's a genuine implementation bug.

--anders

(continuing talking to myself, thanks for listening :-) )

> *However*, string literals in D should all be zero-terminated...
> These new ones are not, and that's a genuine implementation bug.

Never mind, they are all zero-terminated. (sorry, Walter!)

Just "forgot" that when passed as variadic arguments they
are dynamic arrays and not pointers as when passed to
the first argument of printf ("char*") for instance...

Thus:
  printf(__FILE__); // works
and:
  printf("%s\n", __FILE__); // segfaults

but:
  printf("%.*s\n", __FILE__); // works
  printf("%s\n", cast(char*) __FILE__); // works

But it's easier to use writef, which does the right thing.
Regular string literal and printf/writef handling, in D...

--anders

Regan Heath wrote...

>writefln("Line: %s",__FILE__,":%d",__LINE__," %s",__DATE__," %s",__TIME__," %s",__TIMESTAMP__);
>
>notice, writef<etc> can accept formatting strings anywhere.

Very interesting tip, I am starting to learn what amazing things you can *finally* do, but I would never have tried the above.


>Also it is documented  http://www.digitalmars.com/d/lex.html#specialtokens

Sigh, I had read that but somehow I seem to have skipped the  "__LINE__ integer literal" part every time.

Thanx :)

AEon