www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Just a thought with slices.

reply J Anderson <REMOVEanderson badmama.com.au> writes:
Just a thought with slices.  What about if the namespaces became 
optional inside the square brackets (ie like a with statement).

ie
= test[0..length];

instead of:
= test[0..test.length];

Then you could also do negative versions like:

= test[0..length-10];

Since we use length in slices all the time, this would be handy syntax 
sugar.

This would also apply to UDT's.

-- 
-Anderson: http://badmama.com.au/~anderson/
May 10 2004
next sibling parent reply Norbert Nemec <Norbert.Nemec gmx.de> writes:
I think the '$' idea is much simpler than that. Especially, if we have
multidimensional arrays one day, where there is not only one length but
several ranges for the individual dimensions.

(B.t.w: I chose the term "range" over "length" for multidimensional arrays,
since it would seems strange to talk of several lengths instead of width,
height or whatever. "length" should of course continue to for
one-dimensional arrays as an alternative term)

        test[0..test.range[0]-1,0..test.range[1]]

would then become

        test[0..$-1,0..$-1]



J Anderson wrote:


 Just a thought with slices.  What about if the namespaces became
 optional inside the square brackets (ie like a with statement).
 
 ie
 = test[0..length];
 
 instead of:
 = test[0..test.length];
 
 Then you could also do negative versions like:
 
 = test[0..length-10];
 
 Since we use length in slices all the time, this would be handy syntax
 sugar.
 
 This would also apply to UDT's.
May 10 2004
parent reply J Anderson <REMOVEanderson badmama.com.au> writes:
Norbert Nemec wrote:


I think the '$' idea is much simpler than that. Especially, if we have
multidimensional arrays one day, where there is not only one length but
several ranges for the individual dimensions.

(B.t.w: I chose the term "range" over "length" for multidimensional arrays,
since it would seems strange to talk of several lengths instead of width,
height or whatever. "length" should of course continue to for
one-dimensional arrays as an alternative term)

        test[0..test.range[0]-1,0..test.range[1]]

would then become

        test[0..$-1,0..$-1]
  


New symbols are rarely good ways to make readable code (besides the 
common mathematical ones).  As soon as I see $, it's yet another symbol 
I have to go away and learn.  Besides with UDT's you would have access 
(without the dot) to any public member defined in that class not just 
length.

-- 
-Anderson: http://badmama.com.au/~anderson/
May 10 2004
parent reply Norbert Nemec <Norbert.Nemec gmx.de> writes:
J Anderson wrote:


 Norbert Nemec wrote:
 

I think the '$' idea is much simpler than that. Especially, if we have
multidimensional arrays one day, where there is not only one length but
several ranges for the individual dimensions.

(B.t.w: I chose the term "range" over "length" for multidimensional
arrays, since it would seems strange to talk of several lengths instead of
width, height or whatever. "length" should of course continue to for
one-dimensional arrays as an alternative term)

        test[0..test.range[0]-1,0..test.range[1]]

would then become

        test[0..$-1,0..$-1]
  


 New symbols are rarely good ways to make readable code (besides the
 common mathematical ones).  As soon as I see $, it's yet another symbol
 I have to go away and learn.


That's the same with every new language concept. And the meaning of $ is far
easier to understand than the concept that inside of [] you are in a
different namespace.


 Besides with UDT's you would have access
 (without the dot) to any public member defined in that class not just
 length.


True, but that might just as well be confusing, if the class defines members
that you are not interested in, colliding with whatever you actually want
to access. Besides: what other class members would you access so often in
within brackets, that this would really make a big difference?
May 10 2004
parent reply Kevin Bealer <Kevin_member pathlink.com> writes:
In article <c7o8n6$29p$1 digitaldaemon.com>, Norbert Nemec says...
.

 Besides with UDT's you would have access
 (without the dot) to any public member defined in that class not just
 length.


True, but that might just as well be confusing, if the class defines members
that you are not interested in, colliding with whatever you actually want
to access. Besides: what other class members would you access so often in
within brackets, that this would really make a big difference?


Regard this simple string class; regard this simple find method:

string has_waldo("micro-miniature waldo fabricator");
string wheres_waldo( has_waldo[find("waldo")..length] );

Kevin
May 10 2004
parent Norbert Nemec <Norbert.Nemec gmx.de> writes:
Kevin Bealer wrote:


 In article <c7o8n6$29p$1 digitaldaemon.com>, Norbert Nemec says...
 .

 Besides with UDT's you would have access
 (without the dot) to any public member defined in that class not just
 length.


True, but that might just as well be confusing, if the class defines
members that you are not interested in, colliding with whatever you
actually want to access. Besides: what other class members would you
access so often in within brackets, that this would really make a big
difference?


 
 Regard this simple string class; regard this simple find method:
 
 string has_waldo("micro-miniature waldo fabricator");
 string wheres_waldo( has_waldo[find("waldo")..length] );


OK, I have to yield to that. Personally, I still like the '$' concept, but
that's probably just a matter of taste...
May 10 2004
prev sibling next sibling parent Ilya Minkov <minkov cs.tum.edu> writes:
I like this.

-eye



J Anderson schrieb:


 Just a thought with slices.  What about if the namespaces became 
 optional inside the square brackets (ie like a with statement).
 
 ie
 = test[0..length];
 
 instead of:
 = test[0..test.length];
 
 Then you could also do negative versions like:
 
 = test[0..length-10];
 
 Since we use length in slices all the time, this would be handy syntax 
 sugar.
 
 This would also apply to UDT's.
May 10 2004
prev sibling next sibling parent J Anderson <REMOVEanderson badmama.com.au> writes:
J Anderson wrote:


 Just a thought with slices.  What about if the namespaces became 
 optional inside the square brackets (ie like a with statement).

 ie
 = test[0..length];

 instead of:
 = test[0..test.length];

 Then you could also do negative versions like:

 = test[0..length-10];

 Since we use length in slices all the time, this would be handy syntax 
 sugar.

 This would also apply to UDT's.


I should add.  It would also be useful for array indexes...

= test[length-1];



-- 
-Anderson: http://badmama.com.au/~anderson/
May 10 2004
prev sibling next sibling parent reply J C Calvarese <jcc7 cox.net> writes:
J Anderson wrote:

 Just a thought with slices.  What about if the namespaces became 
 optional inside the square brackets (ie like a with statement).
 
 ie
 = test[0..length];


That's a clever idea, but I'm afraid it'd lead to subtle errors. What if 
you meant test2.length? You won't get a compile error. If you're lucky, 
you'd get a runtime error. If you're unlucky, you'd just get the wrong 
result (and might not even realize it's the wrong result).

But using a symbol to represent this idea would be vastly "safer":
  = test[0..$length];
or
  = test[0..$];

My 2 cents.


 
 instead of:
 = test[0..test.length];
 
 Then you could also do negative versions like:
 
 = test[0..length-10];
 
 Since we use length in slices all the time, this would be handy syntax 
 sugar.
 
 This would also apply to UDT's.
 



-- 
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/
May 10 2004
parent reply A. Stevenson <A._member pathlink.com> writes:
In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...

J Anderson wrote:

 Just a thought with slices.  What about if the namespaces became 
 optional inside the square brackets (ie like a with statement).
 
 ie
 = test[0..length];


That's a clever idea, but I'm afraid it'd lead to subtle errors. What if 
you meant test2.length? You won't get a compile error. If you're lucky, 
you'd get a runtime error. If you're unlucky, you'd just get the wrong 
result (and might not even realize it's the wrong result).

But using a symbol to represent this idea would be vastly "safer":
  = test[0..$length];
or
  = test[0..$];

My 2 cents.


 
 instead of:
 = test[0..test.length];
 
 Then you could also do negative versions like:
 
 = test[0..length-10];
 
 Since we use length in slices all the time, this would be handy syntax 
 sugar.
 
 This would also apply to UDT's.
 



-- 
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/


Or an even more subtle source of confusion.
Suppose I have the following (contrived example):

int[] bob( int[] spam )
{
int length = 20;
return spam[6..length]
}

Which 'length' would be referred to? How would I specify the function-scope
variable length instead of the array property length? 

If you want to use 207 class functions/members in array indices, why not wrap it
in a with block?

That said, I'm not fond of the extra symbol - it's not obvious what it means and
it all starts to get a little Perl-esque, where your cat walking across your
keyboard can generate a valid program.
May 11 2004
next sibling parent reply Norbert Nemec <Norbert.Nemec gmx.de> writes:
A. Stevenson wrote:

 That said, I'm not fond of the extra symbol - it's not obvious what it
 means and it all starts to get a little Perl-esque, where your cat walking
 across your keyboard can generate a valid program.


Well - Perl certainly is one of the ugliest languages out there, but in
string handling, there certainly are few languages to beat it. If we want
to get powerful strings, it would be ignorant to refuse being influence
from Perl in certain -- well-considered -- respects.
May 11 2004
parent reply A. Stevenson <alexstev AT uk.ibm.com> <A._member pathlink.com> writes:
In article <c7q7mo$2vh8$2 digitaldaemon.com>, Norbert Nemec says...

A. Stevenson wrote:

 That said, I'm not fond of the extra symbol - it's not obvious what it
 means and it all starts to get a little Perl-esque, where your cat walking
 across your keyboard can generate a valid program.


Well - Perl certainly is one of the ugliest languages out there, but in
string handling, there certainly are few languages to beat it. If we want
to get powerful strings, it would be ignorant to refuse being influence
from Perl in certain -- well-considered -- respects.


True enough - I use both Perl and C in my work to achieve different tasks -
they're good for different things. 

But still I find Perl difficult and painful to use. Whenever I have to do
something big in Perl I find myself relieved to go back to C!

One of the things that I find most difficult in Perl (other than regex stuff) is
the various internal variables ($_, $!, $1 etc) where the symbol itself gives no
clue to what it is to the uninformed layman like myself. Shortcuts that are
useful to Perl hackers of a lot of experience cause me a headache: 

example:

while ( <SPAM> )
{
chomp;
dosomething($_);
}

It's fairly obvious to anyone with Perl experience what this is shorthand for,
but it caused me a minor headache.

Anyway, my point is that shortcuts for the experienced ought to be obvious
enough that the inexperienced can understand them without having to ask silly
questions or find an obscure reference in a document.

Of course, the above is only my opinion and does not necessarily represent
truth, beauty or freedom for all.
May 11 2004
parent Norbert Nemec <Norbert.Nemec gmx.de> writes:
A. Stevenson wrote:


 One of the things that I find most difficult in Perl (other than regex
 stuff) is the various internal variables ($_, $!, $1 etc) where the symbol
 itself gives no clue to what it is to the uninformed layman like myself.
 Shortcuts that are useful to Perl hackers of a lot of experience cause me
 a headache:


I fully agree there. I never used much of Perl, but make and bash have that
same situation. Anyhow, the situation with operators in D seems different
to me: operators have very different semantics and meanings. Whether you
introduce a keyword or an operator for a given purpose, you'll always have
to know what it means or look up the reference for it.

For example: I find the keyword "is" even more confusing than "===" because
a newby believes to know intuitively what it means, so he might not take
the time to look it up and then misinterpret it.

And for the rule of implicit namespaces within indexing expression: the
efford it takes to explain to someone what the details of this mechanism
are huge compared to the three lines it takes to explain the proposed "$".
May 11 2004
prev sibling next sibling parent J Anderson <REMOVEanderson badmama.com.au> writes:
A. Stevenson wrote:


In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...
  


J Anderson wrote:
    


Just a thought with slices.  What about if the namespaces became 
optional inside the square brackets (ie like a with statement).

ie
= test[0..length];
      


That's a clever idea, but I'm afraid it'd lead to subtle errors. What if 
you meant test2.length? You won't get a compile error. If you're lucky, 
you'd get a runtime error. If you're unlucky, you'd just get the wrong 
result (and might not even realize it's the wrong result).

But using a symbol to represent this idea would be vastly "safer":
 = test[0..$length];
or
 = test[0..$];

My 2 cents.

    


instead of:
= test[0..test.length];

Then you could also do negative versions like:

= test[0..length-10];

Since we use length in slices all the time, this would be handy syntax 
sugar.

This would also apply to UDT's.

      


-- 
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/
    


Or an even more subtle source of confusion.
Suppose I have the following (contrived example):

int[] bob( int[] spam )
{
int length = 20;
return spam[6..length]
}

Which 'length' would be referred to? How would I specify the function-scope
variable length instead of the array property length? 

If you want to use 207 class functions/members in array indices, why not wrap it
in a with block?

That said, I'm not fond of the extra symbol - it's not obvious what it means and
it all starts to get a little Perl-esque, where your cat walking across your
keyboard can generate a valid program.


  


What JC mentioned could be an issue.  This case would not be an issue, 
as the compile could easily cause a collision error.

-- 
-Anderson: http://badmama.com.au/~anderson/
May 11 2004
prev sibling parent reply "Phill" <phill pacific.net.au> writes:
"A. Stevenson" <A._member pathlink.com> wrote in message
news:c7q6nh$2ufr$1 digitaldaemon.com...

 In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...

J Anderson wrote:

 Just a thought with slices.  What about if the namespaces became
 optional inside the square brackets (ie like a with statement).

 ie
 = test[0..length];


That's a clever idea, but I'm afraid it'd lead to subtle errors. What if
you meant test2.length? You won't get a compile error. If you're lucky,
you'd get a runtime error. If you're unlucky, you'd just get the wrong
result (and might not even realize it's the wrong result).

But using a symbol to represent this idea would be vastly "safer":
  = test[0..$length];
or
  = test[0..$];

My 2 cents.


 instead of:
 = test[0..test.length];

 Then you could also do negative versions like:

 = test[0..length-10];

 Since we use length in slices all the time, this would be handy syntax
 sugar.

 This would also apply to UDT's.



--
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/


 Or an even more subtle source of confusion.
 Suppose I have the following (contrived example):

 int[] bob( int[] spam )
 {
 int length = 20;
 return spam[6..length]
 }

 Which 'length' would be referred to? How would I specify the


function-scope

 variable length instead of the array property length?


Well according to  JA's  suggestion, the length being
referred to would be the length of the spam[].
I dont know about anyone else, but I know that
I would have changed the name of the int to save
future confusion.

With JA's suggestion there would be no need for
using a with block, because the [] would automatically
act like a with block for anything that is inside the
[].

Phill.
May 11 2004
parent reply A. Stevenson <alexstev AT uk.ibm.com> <A._member pathlink.com> writes:
In article <c7qbon$4ba$1 digitaldaemon.com>, Phill says...

"A. Stevenson" <A._member pathlink.com> wrote in message
news:c7q6nh$2ufr$1 digitaldaemon.com...

 In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...

J Anderson wrote:

 Just a thought with slices.  What about if the namespaces became
 optional inside the square brackets (ie like a with statement).

 ie
 = test[0..length];


That's a clever idea, but I'm afraid it'd lead to subtle errors. What if
you meant test2.length? You won't get a compile error. If you're lucky,
you'd get a runtime error. If you're unlucky, you'd just get the wrong
result (and might not even realize it's the wrong result).

But using a symbol to represent this idea would be vastly "safer":
  = test[0..$length];
or
  = test[0..$];

My 2 cents.


 instead of:
 = test[0..test.length];

 Then you could also do negative versions like:

 = test[0..length-10];

 Since we use length in slices all the time, this would be handy syntax
 sugar.

 This would also apply to UDT's.



--
Justin (a/k/a jcc7)
http://jcc_7.tripod.com/d/


 Or an even more subtle source of confusion.
 Suppose I have the following (contrived example):

 int[] bob( int[] spam )
 {
 int length = 20;
 return spam[6..length]
 }

 Which 'length' would be referred to? How would I specify the


function-scope

 variable length instead of the array property length?


Well according to  JA's  suggestion, the length being
referred to would be the length of the spam[].
I dont know about anyone else, but I know that
I would have changed the name of the int to save
future confusion.

With JA's suggestion there would be no need for
using a with block, because the [] would automatically
act like a with block for anything that is inside the
[].

Phill.


But suppose that the array is a class that overloads opSlice - you get problems
if your local variables have the name of any of the public class methods/member
vars/attributes
May 11 2004
parent reply J Anderson <REMOVEanderson badmama.com.au> writes:
A. Stevenson <alexstev AT uk.ibm.com> wrote:


In article <c7qbon$4ba$1 digitaldaemon.com>, Phill says...
  


"A. Stevenson" <A._member pathlink.com> wrote in message
news:c7q6nh$2ufr$1 digitaldaemon.com...
    


In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...
      


Well according to  JA's  suggestion, the length being
referred to would be the length of the spam[].
I dont know about anyone else, but I know that
I would have changed the name of the int to save
future confusion.

With JA's suggestion there would be no need for
using a with block, because the [] would automatically
act like a with block for anything that is inside the
[].

Phill.



    


But suppose that the array is a class that overloads opSlice - you get problems
if your local variables have the name of any of the public class methods/member
vars/attributes
  


I've already explained that this would cause a compile error which would 
be solved by giving the full namespace.

-- 
-Anderson: http://badmama.com.au/~anderson/
May 11 2004
parent A. Stevenson <alexstev AT uk.ibm.com> <A._member pathlink.com> writes:
In article <c7qcpp$5ge$1 digitaldaemon.com>, J Anderson says...

A. Stevenson <alexstev AT uk.ibm.com> wrote:


In article <c7qbon$4ba$1 digitaldaemon.com>, Phill says...
  


"A. Stevenson" <A._member pathlink.com> wrote in message
news:c7q6nh$2ufr$1 digitaldaemon.com...
    


In article <c7p18a$17t5$1 digitaldaemon.com>, J C Calvarese says...
      


Well according to  JA's  suggestion, the length being
referred to would be the length of the spam[].
I dont know about anyone else, but I know that
I would have changed the name of the int to save
future confusion.

With JA's suggestion there would be no need for
using a with block, because the [] would automatically
act like a with block for anything that is inside the
[].

Phill.



    


But suppose that the array is a class that overloads opSlice - you get problems
if your local variables have the name of any of the public class methods/member
vars/attributes
  


I've already explained that this would cause a compile error which would 
be solved by giving the full namespace.

-- 
-Anderson: http://badmama.com.au/~anderson/


My apologies - that's what I get for trying to work and read newsgroups at the
same time :)
May 11 2004
prev sibling parent reply "Phill" <phill pacific.net.au> writes:
Im not with you, you can already do this:

 test[0..test.length - 10];



"J Anderson" <REMOVEanderson badmama.com.au> wrote in message
news:c7npt6$2c59$1 digitaldaemon.com...

 Just a thought with slices.  What about if the namespaces became
 optional inside the square brackets (ie like a with statement).

 ie
 = test[0..length];

 instead of:
 = test[0..test.length];

 Then you could also do negative versions like:

 = test[0..length-10];

 Since we use length in slices all the time, this would be handy syntax
 sugar.

 This would also apply to UDT's.

 --
 -Anderson: http://badmama.com.au/~anderson/
May 10 2004
parent reply J Anderson <REMOVEanderson badmama.com.au> writes:
Phill wrote:


Im not with you, you can already do this:

 test[0..test.length - 10];
  


People have been requesting a $ symbol to be used as length.  I believe 
it would be better to have a more generic idea that  uses words instead 
of ugly symbols.

test[0..length - 10];

Is shorter and quite readable.  It's like doing:

with (test) { test[0..length-10]; }

I would imagine it could save a lot more typing in UDT's ie:

class Array
{
...    stride(int, int) {...}
...    opIndex {...}
}

Array obj(10, 10);

= obj[obj.stride(0, obj.find(10)) .. obj.stride(2, 10)][obj.stride(0, 
10) .. obj.stride(1, 3)] ;

As opposed to:

= obj[stride(0, find(10)) .. stride(2, 10)][stride(0, 10) .. stride(1, 3)] ;

or with opCall

= obj[(0, find(10)) .. (2, 10)][(0, 10) .. (1, 3)] ;

Actually you could even do this:

= obj[(10, 10)]; //Almost rectangular array form (make the brackets 
optional and you'd have it)

Ok a bit overboard but you get the idea.

Another advantage of this technique is that its generic.  It's easily to 
modify one name at the start rather then 3 or 4 entries, which is a good 
maintenance helper.


"J Anderson" <REMOVEanderson badmama.com.au> wrote in message
news:c7npt6$2c59$1 digitaldaemon.com...
  


Just a thought with slices.  What about if the namespaces became
optional inside the square brackets (ie like a with statement).

ie
= test[0..length];

instead of:
= test[0..test.length];

Then you could also do negative versions like:

= test[0..length-10];

Since we use length in slices all the time, this would be handy syntax
sugar.

This would also apply to UDT's.

--
-Anderson: http://badmama.com.au/~anderson/
    



  



-- 
-Anderson: http://badmama.com.au/~anderson/
May 10 2004
next sibling parent Derek Parnell <derek psych.ward> writes:
On Tue, 11 May 2004 08:20:09 +0800, J Anderson wrote:


 Phill wrote:
 

Im not with you, you can already do this:

 test[0..test.length - 10];
  


 People have been requesting a $ symbol to be used as length.  I believe 
 it would be better to have a more generic idea that  uses words instead 
 of ugly symbols.


Hmmmm....ugly symbols....
 

 test[0..length - 10];


test begin slice 0 to length minus 10 end slice end statement

Starts to look a bit COBOL-like now ;-)


-- 
Derek
11/May/04 11:37:39 AM
May 10 2004
prev sibling parent "Phill" <phill pacific.net.au> writes:
"J Anderson" <REMOVEanderson badmama.com.au> wrote in message
news:c7p67p$1ejh$1 digitaldaemon.com...

 Phill wrote:


Im not with you, you can already do this:

 test[0..test.length - 10];


 People have been requesting a $ symbol to be used as length.  I believe
 it would be better to have a more generic idea that  uses words instead
 of ugly symbols.


I agree with you, I would much rather the word length.
There is no "now what did '$' mean again?"
Plus I think that most people are already used to
the word length and what it means.


 test[0..length - 10];

 Is shorter and quite readable.  It's like doing:

 with (test) { test[0..length-10]; }


Ok I get the picture now.
This could be very handy.

I just love "with", its really fantastic, and (OT)also foreach.

Thanks
         Phill
May 11 2004