findBack: find a needle in a haystack from the back

[Timothee Cour via Digitalmars-d-learn]

Is there a more idiomatic/elegant way to achieve the following, while
remaining as efficient as possible?
Same question in the simpler case n==0?

using retro seems inefficient because of all the decodings

// returns the largest suffix of a that contains no more than n times c
string findBack(string a, char c, size_t n=0){
  auto b=cast(immutable(ubyte)[])a;
  auto val=cast(ubyte)c;
  size_t counter=0;
  for(ptrdiff_t i=cast(ptrdiff_t)b.length - 1; i=0; i--){
if(b[i]==c){
  if(counter=n)
return cast(string)a[i+1..$];
  counter++;
}
  }
  return a;
}

//unittest{
void test3(){
  auto c='\n';
  string a=abc1\nabc2\nabc3;
  assert(a.findBack(c,0)==abc3);
  assert(a.findBack(c,1)==abc2\nabc3);
  assert(a.findBack(c,2)==abc1\nabc2\nabc3);
  assert(a.findBack(c,3)==abc1\nabc2\nabc3);
}

Re: findBack: find a needle in a haystack from the back

[Tobias Pankrath via Digitalmars-d-learn]

On Monday, 9 June 2014 at 07:58:25 UTC, Timothee Cour via 
Digitalmars-d-learn wrote:
Is there a more idiomatic/elegant way to achieve the following, 
while

remaining as efficient as possible?
Same question in the simpler case n==0?

using retro seems inefficient because of all the decodings

// returns the largest suffix of a that contains no more than n 
times c

string findBack(string a, char c, size_t n=0){
  auto b=cast(immutable(ubyte)[])a;
  auto val=cast(ubyte)c;
  size_t counter=0;
  for(ptrdiff_t i=cast(ptrdiff_t)b.length - 1; i=0; i--){
if(b[i]==c){
  if(counter=n)
return cast(string)a[i+1..$];
  counter++;
}
  }
  return a;
}

//unittest{
void test3(){
  auto c='\n';
  string a=abc1\nabc2\nabc3;
  assert(a.findBack(c,0)==abc3);
  assert(a.findBack(c,1)==abc2\nabc3);
  assert(a.findBack(c,2)==abc1\nabc2\nabc3);
  assert(a.findBack(c,3)==abc1\nabc2\nabc3);
}


If you are going to cast the string to ubyte[] anyway, why not 
just do this before using retro?

Re: findBack: find a needle in a haystack from the back

[Nordlöw]

using retro seems inefficient because of all the decodings


Phobos git master just got support for next-gen string processing:

https://github.com/D-Programming-Language/phobos/pull/2043

I believe

x.byChar.retro

is what you want