• bearophile <bearophileHUGS lycos.com> Jul 28 2009
• Bill Baxter <wbaxter gmail.com> Jul 28 2009
• bearophile <bearophileHUGS lycos.com> Jul 28 2009
• bearophile <bearophileHUGS lycos.com> Jul 28 2009
• Bill Baxter <wbaxter gmail.com> Jul 28 2009

bearophile <bearophileHUGS lycos.com> writes:

Bill Baxter:
 Now we may not want to go so hog wild putting  this and  that
 everywhere, but if we did we could get rid of 19 keywords right there,
 and add  property also without adding a new keyword.


Glad to see I'm not the only one to think like this :-)
Later more semantic attributes can be added, for example to express things that
the compiler can use to parallelize code better, to know about pointer
aliasing, to tell the compiler what asm functions can be inlined (currently
done with a pragma in LDC), etc.

What syntax do you want to use when you want more than one attribute?
 something1  somethingelse
...

Bye,
bearophile

Bill Baxter <wbaxter gmail.com> writes:

On Tue, Jul 28, 2009 at 11:12 AM, bearophile<bearophileHUGS lycos.com> wrot=
e:
 Bill Baxter:
 Now we may not want to go so hog wild putting  this and  that
 everywhere, but if we did we could get rid of 19 keywords right there,
 and add  property also without adding a new keyword.


 Glad to see I'm not the only one to think like this :-)
 Later more semantic attributes can be added, for example to express thing=


er aliasing, to tell the compiler what asm functions can be inlined (curren=
tly done with a pragma in LDC), etc.
 What syntax do you want to use when you want more than one attribute?
  something1  somethingelse
 ...


I don't know.  I'm actually not too up on how these things are used in
the languages where they exist (just Java, Python, and C#?).  It just
seems like a good idea to solve a more general problem if possible.
And when someone reminded us that "deprecated" is a keyword, that set
me off.


Something else that occurred to me while reading the keyword list is
that "switch" is a horrible keyword hog.  Three keywords just for that
one construct!   If D didn't have a switch statement today and were
looking to add one, somehow I doubt that there would be sufficient
support for dedicating three whole keywords to the thing.    Switch is
a monstrosity pretty much any way you look at it.  Sh had the right
idea there.  Is "case" really necessary there?  Some syntax should
suffice I would thing.  And "default" is probably the world's most
useless keyword.  Why not "else:" or "*:" instead of introducing a
whole new keyword?

--bb

bearophile <bearophileHUGS lycos.com> writes:

Bill Baxter:
 Switch is a monstrosity pretty much any way you look at it.  Sh had the right
 idea there.  Is "case" really necessary there?  Some syntax should
 suffice I would thing.  And "default" is probably the world's most
 useless keyword.  Why not "else:" or "*:" instead of introducing a
 whole new keyword?


Case of Pascal is better (2 keywords where "of" isn't that useful, begin-end
are optional if you want more than one instruction, Pascal isn't
case-sensitive):

Case X of
  1: Begin
        ...
      End;
  2 .. 10: Begin
        ...
      End;
  ...
  Else Begin
         ...
       End;
End; {CASE}

But D is backward compatible with C (unlike languages like Scala, that
eventually D will need to take a look at), so this is an academic topic.

Bye,
bearophile

bearophile <bearophileHUGS lycos.com> writes:

Bill Baxter:
 Changing the keywords used in a switch would be a very mechanical
 transformation, like changing typedef to alias everywhere it appears
 in C code you're translating.


Once you have changed the keyword used in the case/switch, then you (and me)
probably want to fix other problems of switch (fall through).
Recently here I have suggested the creation of a small tool (written in Python,
not in D) like 2to3 (used to convert Python2 to Python3) to convert C to D :-)
I don't know how much long such program may become (2to3 uses a parser of the
std lib).

Bye,
bearophile

Bill Baxter <wbaxter gmail.com> writes:

On Tue, Jul 28, 2009 at 12:01 PM, bearophile<bearophileHUGS lycos.com> wrot=
e:
 Bill Baxter:
 Switch is a monstrosity pretty much any way you look at it. =A0Sh had th=




 idea there. =A0Is "case" really necessary there? =A0Some syntax should
 suffice I would thing. =A0And "default" is probably the world's most
 useless keyword. =A0Why not "else:" or "*:" instead of introducing a
 whole new keyword?


 Case of Pascal is better (2 keywords where "of" isn't that useful, begin-=


ensitive):
 Case X of
 =A01: Begin
 =A0 =A0 =A0 =A0...
 =A0 =A0 =A0End;
 =A02 .. 10: Begin
 =A0 =A0 =A0 =A0...
 =A0 =A0 =A0End;
 =A0...
 =A0Else Begin
 =A0 =A0 =A0 =A0 ...
 =A0 =A0 =A0 End;
 End; {CASE}

 But D is backward compatible with C (unlike languages like Scala, that ev=



Except that it's very much not backwards compatible with C syntax in
constructs that are very frequent in C: typedef/alias, #define.
Changing the keywords used in a switch would be a very mechanical
transformation, like changing typedef to alias everywhere it appears
in C code you're translating.

--bb