www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Heretical template syntax proposal

reply "Bill Baxter" <wbaxter gmail.com> writes:
In another thread the idea was discussed of using [x] for template arguments.

The main objection to this was that it would be ambiguous with array indexing.
I know [x] has meant array indexing for a long time, but I think
there's not really any longstanding connection between that and any
math notation that existed previously.
In fact (x) is arguably a more mathematical way to index an array.  An
array is just a function that given some input returns some output.
So array(i) actually makes a lot of sense.

This is actually how Matlab behaves, in fact.   In Matlab you don't
index with [i] you index with (i).

There's no reason as far as I can tell that [i] needs to mean array indexing.

So if you change array indexing to (i) (both std arrays and AAs) then
you can use [x] for template syntax.

I think the ambiguity between arrays and functions that would result
is a non-issue.  An array is a kind of function so there's no real
issue.  In fact in Matlab you have exactly that ambiguity and it works
just fine.  In fact it's even more natural since it allows you to
replace an array with a function any time your app's needs grow beyond
the simple mapping function provided by the array.


Ok, it's probably totally unpalatable to anyone who has ever
programmed in a C family language, but I think it could work if we had
the stomach to do it.

--bb
Oct 22 2008
next sibling parent reply Janderson <ask me.com> writes:
Bill Baxter wrote:

 In another thread the idea was discussed of using [x] for template arguments.
 
 The main objection to this was that it would be ambiguous with array indexing.
 I know [x] has meant array indexing for a long time, but I think
 there's not really any longstanding connection between that and any
 math notation that existed previously.
 In fact (x) is arguably a more mathematical way to index an array.  An
 array is just a function that given some input returns some output.
 So array(i) actually makes a lot of sense.
 
 This is actually how Matlab behaves, in fact.   In Matlab you don't
 index with [i] you index with (i).
 
 There's no reason as far as I can tell that [i] needs to mean array indexing.
 
 So if you change array indexing to (i) (both std arrays and AAs) then
 you can use [x] for template syntax.
 
 I think the ambiguity between arrays and functions that would result
 is a non-issue.  An array is a kind of function so there's no real
 issue.  In fact in Matlab you have exactly that ambiguity and it works
 just fine.  In fact it's even more natural since it allows you to
 replace an array with a function any time your app's needs grow beyond
 the simple mapping function provided by the array.
 
 
 Ok, it's probably totally unpalatable to anyone who has ever
 programmed in a C family language, but I think it could work if we had
 the stomach to do it.
 
 --bb


Its also the way ADA worked.  It was a great feature because it made 
programming more generic.  Imagine being able to use an array in the 
same template without having to write a wrapper function (and visa 
versa).  Sadly I don't think D will ever adopt this as its a pretty 
radical change from C and D.  Also it means all the libs would have to 
be re-written to work with this when they get ported to D 2 or 3.

-Joel
Oct 22 2008
parent "Bill Baxter" <wbaxter gmail.com> writes:
On Thu, Oct 23, 2008 at 2:36 PM, Janderson <ask me.com> wrote:

 Bill Baxter wrote:


 Its also the way ADA worked.  It was a great feature because it made
 programming more generic.  Imagine being able to use an array in the same
 template without having to write a wrapper function (and visa versa).  Sadly
 I don't think D will ever adopt this as its a pretty radical change from C
 and D.  Also it means all the libs would have to be re-written to work with
 this when they get ported to D 2 or 3.


For the most part an automatic tool that changed !() to [] and [] to
() would work, but any class or struct that has both opIndex and
opCall would need to be redesigned.

So yeh, it would be a difficult transition.   I guess I'll just jot it
down as a reminder in case I ever decided to implement my own language
some day. :-)


--bb



 Oct 22 2008



prev sibling  parent reply Bruno Medeiros <brunodomedeiros+spam com.gmail>  writes:


Bill Baxter wrote:

 In another thread the idea was discussed of using [x] for template arguments.
 
 The main objection to this was that it would be ambiguous with array indexing.
 I know [x] has meant array indexing for a long time, but I think
 there's not really any longstanding connection between that and any
 math notation that existed previously.
 In fact (x) is arguably a more mathematical way to index an array.  An
 array is just a function that given some input returns some output.
 So array(i) actually makes a lot of sense.
 
 This is actually how Matlab behaves, in fact.   In Matlab you don't
 index with [i] you index with (i).
 
 There's no reason as far as I can tell that [i] needs to mean array indexing.
 
 So if you change array indexing to (i) (both std arrays and AAs) then
 you can use [x] for template syntax.
 
 I think the ambiguity between arrays and functions that would result
 is a non-issue.  An array is a kind of function so there's no real
 issue.  In fact in Matlab you have exactly that ambiguity and it works
 just fine.  In fact it's even more natural since it allows you to
 replace an array with a function any time your app's needs grow beyond
 the simple mapping function provided by the array.
 
 
 Ok, it's probably totally unpalatable to anyone who has ever
 programmed in a C family language, but I think it could work if we had
 the stomach to do it.
 
 --bb


Hum, that's a hard break from the C language family. Too much of a break 
I think.

Hum, on a side note, that makes me think how it would be like for a 
language like D, with the same semantics as D, but having a completely 
redesigned syntax, from the grounds up. I wonder if there could be 
significant advantages in such a language (in clarity, consistency, 
expressiveness, etc).
Because it is the case that D is continuously getting new additions in 
syntax and semantics. But while semantics are generally easier to change 
and refactor, because there is an infinite "domain space" for semantics, 
the "domain space" for (reasonable) syntax is limited, as we saw 
recently with the desire in the new template instantiation discussions 
about a new kind of brackets. And refactoring/restructuring the syntax 
of a language to accommodate such changes is such a breaking change that 
is nearly impractical without creating a new language.

-- 
Bruno Medeiros - Software Developer, MSc. in CS/E graduate
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D



 Oct 24 2008



  parent reply "Jarrett Billingsley" <jarrett.billingsley gmail.com>  writes:


On Fri, Oct 24, 2008 at 9:19 AM, Bruno Medeiros
<brunodomedeiros+spam com.gmail> wrote:

 Hum, on a side note, that makes me think how it would be like for a language
 like D, with the same semantics as D, but having a completely redesigned
 syntax, from the grounds up. I wonder if there could be significant
 advantages in such a language (in clarity, consistency, expressiveness,
 etc).


Delight is partly there.  http://delight.sourceforge.net/



 Oct 24 2008



  parent reply Bruno Medeiros <brunodomedeiros+spam com.gmail>  writes:


Jarrett Billingsley wrote:

 On Fri, Oct 24, 2008 at 9:19 AM, Bruno Medeiros
 <brunodomedeiros+spam com.gmail> wrote:

 Hum, on a side note, that makes me think how it would be like for a language
 like D, with the same semantics as D, but having a completely redesigned
 syntax, from the grounds up. I wonder if there could be significant
 advantages in such a language (in clarity, consistency, expressiveness,
 etc).


 
 Delight is partly there.  http://delight.sourceforge.net/


Interesting (even though it has Python indentation :P ). But does it 
have all D features? Can a D program be automatically translated to a 
Delight program?

-- 
Bruno Medeiros - Software Developer, MSc. in CS/E graduate
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D



 Oct 25 2008



 next sibling parent reply bearophile <bearophileHUGS lycos.com>  writes:


Bruno Medeiros:

 Interesting (even though it has Python indentation :P ).


I think giving it a Python-style syntax is the main (and in the beginning
almost only) purpose of Delight. So if you don't like that thing, then probably
Delight isn't for you.



 But does it have all D features?


Once debugged I think it will have all features of D2, minus few things left
out on purpose, like global variables.

Note that Delight adds some other things, fixes some of the glaring holes of D2
(like in its module system), adds a built-in log system, nullable types, and
I'm discussing with the author regarding other small things to add, like a "in"
lookup for arrays, OpCmp and opEquals among AAs, lazy/nonlazy list comps syntax
copied exactly from Python, a set data type, etc.
The hope is to produce a language nicer to program, more concise, and a little
less bug prone than D. I don't know if it will succeed.



 Can a D program be automatically translated to a Delight program?


Why do you want to do this?
I think that in most cases a translation can be quite mechanical, the
translator can automatically add a class to store the global variables used in
the D program...

Bye,
bearophile



 Oct 25 2008



 next sibling parent  Bruno Medeiros <brunodomedeiros+spam com.gmail>  writes:


bearophile wrote:

 Bruno Medeiros:

 Can a D program be automatically translated to a Delight program?


 
 Why do you want to do this?
 I think that in most cases a translation can be quite mechanical, the
translator can automatically add a class to store the global variables used in
the D program...
 


It's not that I want to do that, it was just a question to see if all D 
feature were available in Delight (if not, a direct translation might be 
difficult or impossible)


-- 
Bruno Medeiros - Software Developer, MSc. in CS/E graduate
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D



 Oct 27 2008



prev sibling  parent reply Bruno Medeiros <brunodomedeiros+spam com.gmail>  writes:


bearophile wrote:

 Bruno Medeiros:

 Interesting (even though it has Python indentation :P ).


 
 I think giving it a Python-style syntax is the main (and in the beginning
almost only) purpose of Delight. So if you don't like that thing, then probably
Delight isn't for you.
 
 


It may not be exactly what I was thinking then. I was thinking of a 
completely new syntax design, so as to be able to be fully "optimized" 
for efficiency/productivity.

-- 
Bruno Medeiros - Software Developer, MSc. in CS/E graduate
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D



 Oct 27 2008



  parent reply bearophile <bearophileHUGS lycos.com>  writes:


Bruno Medeiros:

 It may not be exactly what I was thinking then. I was thinking of a 
 completely new syntax design, so as to be able to be fully "optimized" 
 for efficiency/productivity.


I see. (Note that even if Python syntax isn't D syntax, lot of people are used
to it, so it's a kind of "standard" by itself.) Do you have some specific
ideas? The author of Delight may be interested in what you have to say, because
Delight is very young, so several improvement are possible still.

Bye,
bearophile



 Oct 27 2008



  parent  Bruno Medeiros <brunodomedeiros+spam com.gmail>  writes:


bearophile wrote:

 Bruno Medeiros:

 It may not be exactly what I was thinking then. I was thinking of a 
 completely new syntax design, so as to be able to be fully "optimized" 
 for efficiency/productivity.


 
 I see. (Note that even if Python syntax isn't D syntax, lot of people are used
to it, so it's a kind of "standard" by itself.) Do you have some specific
ideas? The author of Delight may be interested in what you have to say, because
Delight is very young, so several improvement are possible still.
 
 Bye,
 bearophile


I don't have any ideas for a specific change. Working on something like 
this would probably take a fair amount of time, as like I said, any 
potential change in syntax could have several implications in other 
syntax constructs.
But I can give some examples I found interesting: Nemerle for instance. 

example, is that it changed the if-statement to be an expression instead 
of just a statement, thus making the C-famlily ?: ternary operator 
redundant, so that it could be removed.


-- 
Bruno Medeiros - Software Developer, MSc. in CS/E graduate
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D



 Oct 31 2008



prev sibling  parent  Thomas Leonard <talex5+d gmail.com>  writes:


On Sat, 25 Oct 2008 20:05:48 +0100, Bruno Medeiros wrote:


 Jarrett Billingsley wrote:

 On Fri, Oct 24, 2008 at 9:19 AM, Bruno Medeiros
 <brunodomedeiros+spam com.gmail> wrote:

 Hum, on a side note, that makes me think how it would be like for a
 language like D, with the same semantics as D, but having a completely
 redesigned syntax, from the grounds up. I wonder if there could be
 significant advantages in such a language (in clarity, consistency,
 expressiveness, etc).


 
 Delight is partly there.  http://delight.sourceforge.net/


 
 Interesting (even though it has Python indentation :P ). But does it
 have all D features? Can a D program be automatically translated to a
 Delight program?


Almost. Just change the header generation functions to output in Delight 
syntax instead of D syntax. I used this to compile Phobos at one point.

But that meant I could never make Delight smaller than D, since I had to 
support every feature of D, so I changed the compiler to allow parsing 
both D and Delight code. The things I've removed are listed here in the 
"D" section:

  http://delight.sourceforge.net/compare.html

If anything else doesn't work, it's a bug ;-)



 Oct 26 2008