www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - extend foreach to work on non-arrays

reply IntegratedDimensions <IntegratedDimensions gmail.com> writes:
Doesn't make any sense?

foreach(a; x)

if x is not an array then a = x and the loop reduces simply and 
function to the case it is not so their can be no harm.

It unifies the concepts so that one does not have to worry if x 
is an array or not and can offer no harm since when x is not an 
array everything simply reduces to an an alias of x.
May 24 2018
next sibling parent reply Neia Neutuladh <neia ikeran.org> writes:
On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply and 
 function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if x 
 is an array or not and can offer no harm since when x is not an 
 array everything simply reduces to an an alias of x.


You can already do this for any type you define:

class Foo
{
   auto iterate() { return only(this); }
   alias iterate this;
}

Can you give examples of code that is awkward today that would be 
simplified with your proposal? I don't expect people to give full 
cost-benefit analyses for every suggestion, but it'd be nice if 
you could at least mention some of the upsides.
May 24 2018
next sibling parent reply aliak <something something.com> writes:
On Thursday, 24 May 2018 at 23:02:00 UTC, Neia Neutuladh wrote:

 On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
 wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply 
 and function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if 
 x is an array or not and can offer no harm since when x is not 
 an array everything simply reduces to an an alias of x.


 You can already do this for any type you define:

 class Foo
 {
   auto iterate() { return only(this); }
   alias iterate this;
 }

 Can you give examples of code that is awkward today that would 
 be simplified with your proposal? I don't expect people to give 
 full cost-benefit analyses for every suggestion, but it'd be 
 nice if you could at least mention some of the upsides.


And then you can generalize this for any type

auto elseOnly(T)(T t) {
   import std.range: isInputRange;
   static if (isInputRange!T) {
     return t;
   } else  {
     import std.range: only;
     return only(t);
   }
}

foreach(i; 3.elseOnly) {
}

foreach(i; [1, 2, 3].elseOnly) {
}


Cheers
- Ali
May 24 2018
parent reply IntegratedDimensions <IntegratedDimensions gmail.com> writes:
On Thursday, 24 May 2018 at 23:08:39 UTC, aliak wrote:

 On Thursday, 24 May 2018 at 23:02:00 UTC, Neia Neutuladh wrote:

 On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
 wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply 
 and function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if 
 x is an array or not and can offer no harm since when x is 
 not an array everything simply reduces to an an alias of x.


 You can already do this for any type you define:

 class Foo
 {
   auto iterate() { return only(this); }
   alias iterate this;
 }

 Can you give examples of code that is awkward today that would 
 be simplified with your proposal? I don't expect people to 
 give full cost-benefit analyses for every suggestion, but it'd 
 be nice if you could at least mention some of the upsides.


 And then you can generalize this for any type

 auto elseOnly(T)(T t) {
   import std.range: isInputRange;
   static if (isInputRange!T) {
     return t;
   } else  {
     import std.range: only;
     return only(t);
   }
 }

 foreach(i; 3.elseOnly) {
 }

 foreach(i; [1, 2, 3].elseOnly) {
 }


 Cheers
 - Ali


cool, and this is optimal, right? non-arrays are not converted in 
to arrays, etc?
May 24 2018
parent aliak <something something.com> writes:
On Thursday, 24 May 2018 at 23:28:59 UTC, IntegratedDimensions 
wrote:

 cool, and this is optimal, right? non-arrays are not converted 
 in to arrays, etc?


Seems they are not converted no, but it loops like an extra 
variable is stored though (a bool to denote existence of a value)

https://github.com/dlang/phobos/blob/v2.080.0/std/range/package.d#L9512
May 25 2018
prev sibling parent reply IntegratedDimensions <IntegratedDimensions gmail.com> writes:
On Thursday, 24 May 2018 at 23:02:00 UTC, Neia Neutuladh wrote:

 On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
 wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply 
 and function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if 
 x is an array or not and can offer no harm since when x is not 
 an array everything simply reduces to an an alias of x.


 You can already do this for any type you define:

 class Foo
 {
   auto iterate() { return only(this); }
   alias iterate this;
 }

 Can you give examples of code that is awkward today that would 
 be simplified with your proposal? I don't expect people to give 
 full cost-benefit analyses for every suggestion, but it'd be 
 nice if you could at least mention some of the upsides.


The upside is that it is a composite pattern and does not require 
any extra code to use preexisting functionality. There is no 
downside, which is an upside.



foreach(a; x)
{
    // uses a

}

gets directly converted to

{
    // uses x
}

by a simple renaming/aliasing of a to x and so it cannot require 
any more issues than what already exists.

Your method requires one to have access to all type definitions, 
which is not the case.



I do not know why I would be required to specify a use case to 
justify the usability. Most things are not usable until they are 
invented. Most people did not know how useful the hammer would be 
until someone created it. You have to build it for them to come. 
I have my use cases which is why I came up with the idea in the 
first place. I needed a hammer but no one invented it. If I give 
you my use case then there would be two main outcomes: You 
attempt to find a workaround for the use case or claim that it is 
not applicable. These are the "I have a rock that should work as 
good as that hammer thingy you want" and "Hammers are useless".

If you can't find validity in the suggestion based on it's own 
inherent usefulness than the only way I can convince you is to 
provide examples that you actually find useful... that makes it 
difficult on my part and is not fair to me.
May 24 2018
next sibling parent reply Jordan Wilson <wilsonjord gmail.com> writes:
On Thursday, 24 May 2018 at 23:22:56 UTC, IntegratedDimensions 
wrote:

 the idea in the first place. I needed a hammer but no one 
 invented it. If I give you my use case then there would be two 
 main outcomes: You attempt to find a workaround for the use 
 case or claim that it is not applicable. These are the "I have 
 a rock that should work as good as that hammer thingy you want" 
 and "Hammers are useless".


3rd outcome: noobs like me who read the forums who benefit from 
such discussion.

Of course, you could be as disinterested in the 3rd outcome as 
you are the 1st and 2nd, and that's completely fair.

Jordan
May 24 2018
parent meppl <mephisto nordhoff-online.de> writes:
On Friday, 25 May 2018 at 03:34:43 UTC, IntegratedDimensions 
wrote:

 On Friday, 25 May 2018 at 02:43:39 UTC, Jordan Wilson wrote:

 On Thursday, 24 May 2018 at 23:22:56 UTC, IntegratedDimensions 
 wrote:

 [...]


 3rd outcome: noobs like me who read the forums who benefit 
 from such discussion.

 Of course, you could be as disinterested in the 3rd outcome as 
 you are the 1st and 2nd, and that's completely fair.

 Jordan





 So, I will solve the homework problem for you:

 auto max(T)(T t)
 {
   baseTypeOf(T) a;
   foreach(a; t)
      a = max(a,t)
   return a;
 }

 max(3) = 3
 max([3,4,5]) = 5



you didnt take your time to give a proper example. That code is 
wrong in multiple ways. People would have to reconstruct your code
May 25 2018
prev sibling parent Neia Neutuladh <neia ikeran.org> writes:
On Thursday, 24 May 2018 at 23:22:56 UTC, IntegratedDimensions 
wrote:

 If you can't find validity in the suggestion based on it's own 
 inherent usefulness than the only way I can convince you is to 
 provide examples that you actually find useful... that makes it 
 difficult on my part and is not fair to me.


You are asking us all to do work for you. You're asking everyone 
to learn about this new language element. (And yes, we would have 
to learn, since it means the compiler is less able to find errors 
in our code, because more code is valid.) You're asking Walter 
Bright or another compiler developer to implement this feature 
for you, and write the tests, and write the documentation, and 
support this new code for years.

This is not an impossibly huge request, but it isn't trivial by 
any means. There are two socially acceptable ways to get people 
to implement something you want: convince them it's worthwhile, 
or pay them.

Compare: there's no socket.io client or server library for D. You 
should implement one. Does my saying so make you inclined in any 
way to write a socket.io library?
May 24 2018
prev sibling next sibling parent reply meppl <mephisto nordhoff-online.de> writes:
On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply and 
 function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if x 
 is an array or not and can offer no harm since when x is not an 
 array everything simply reduces to an an alias of x.


on the otherhand some programmer might want to get informed buy 
an error-msg, if he accidentally used a non-iteratable variable 
for `foreach`-iteration. To avoid a silent bug
May 24 2018
parent IntegratedDimensions <IntegratedDimensions gmail.com> writes:
On Thursday, 24 May 2018 at 23:03:34 UTC, meppl wrote:

 On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
 wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply 
 and function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if 
 x is an array or not and can offer no harm since when x is not 
 an array everything simply reduces to an an alias of x.


 on the otherhand some programmer might want to get informed buy 
 an error-msg, if he accidentally used a non-iteratable variable 
 for `foreach`-iteration. To avoid a silent bug


In this case it cannot be a bug. The foreach is simply 
ignored/reduced. It would be impossible for any bug to creep 
in(assuming no compiler bugs) in such cases.

foreach(a; x)
{
    x[]
}

would be some type of potential bug... but that bug would also 
exist without the loop. If x is not an array then the the foreach 
effectively is removed and a is just auto a = x;

Any code that uses x would fail just as much as using a and no 
more except.
May 24 2018
prev sibling parent Stefan Koch <uplink.coder googlemail.com> writes:
On Thursday, 24 May 2018 at 22:43:00 UTC, IntegratedDimensions 
wrote:

 Doesn't make any sense?

 foreach(a; x)

 if x is not an array then a = x and the loop reduces simply and 
 function to the case it is not so their can be no harm.

 It unifies the concepts so that one does not have to worry if x 
 is an array or not and can offer no harm since when x is not an 
 array everything simply reduces to an an alias of x.


please look at opApply.
May 25 2018