Re: OUString replace(sal_Unicode, sal_Unicode)

[Stephan Bergmann]

On 10/01/2013 05:05 PM, Matteo Casalin wrote:

 OUString provides the metod replace(sal_Unicode, sal_Unicode), which seems to be not widely 
used, while there are for sure some replaceAll(a, b) calls here and there.
Would it be fine to rename the former to replaceAll(sal_Unicode, sal_Unicode) 
for consistency, fix the current calls and then slowly convert all of the 
call-places of the latter (which I think to be be less efficient)?
If any backward compatibility is needed, replace could be kept and just call 
the related replaceAll.


Not too sure that would really be worth it:

* Single-character replacement is different from multi-character 
replacement in that it doesn't need to specify how to handle cases where 
one replacement gives rise to further replacement opportunities. 
(That's why there originally only was a replaceFirst for the 
multi-character case, to avoid having to make a decision.)


* There is no support for fromIndex in replace.

* replace is an inline function, so it could be replaced (no pun 
intended) with only becoming build-time, not runtime incompatible. 
Keeping it around and having two inline functions doing the same trivial 
thing to avoid that looks like a bit too much overhead to me.


* The underlying C function is called rtl_uString_newReplace, not 
..._rewReplaceAll (and /that/ couldn't be changed without breaking ABI).


Stephan
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice

Re: OUString replace(sal_Unicode, sal_Unicode)

[Matteo Casalin]

Hi Stephan,

On Tue, 01 Oct 2013 18:13:39 +0200
Stephan Bergmann sberg...@redhat.com wrote:

 On 10/01/2013 05:05 PM, Matteo Casalin wrote:
   OUString provides the metod replace(sal_Unicode, sal_Unicode), which 
  seems to be not widely used, while there are for sure some replaceAll(a, 
  b) calls here and there.
  Would it be fine to rename the former to replaceAll(sal_Unicode, 
  sal_Unicode) for consistency, fix the current calls and then slowly convert 
  all of the call-places of the latter (which I think to be be less 
  efficient)?
  If any backward compatibility is needed, replace could be kept and just 
  call the related replaceAll.
 
 Not too sure that would really be worth it:
 
 * Single-character replacement is different from multi-character 
 replacement in that it doesn't need to specify how to handle cases where 
 one replacement gives rise to further replacement opportunities. 
 (That's why there originally only was a replaceFirst for the 
 multi-character case, to avoid having to make a decision.)
 
 * There is no support for fromIndex in replace.
 
 * replace is an inline function, so it could be replaced (no pun 
 intended) with only becoming build-time, not runtime incompatible. 
 Keeping it around and having two inline functions doing the same trivial 
 thing to avoid that looks like a bit too much overhead to me.
 
 * The underlying C function is called rtl_uString_newReplace, not 
 ..._rewReplaceAll (and /that/ couldn't be changed without breaking ABI).
 
 Stephan

Your rationale really overcomes my simple it changes All of the occurrences :)
I will reduce the task to just not introducing any new replaceAll(a, b) and 
by replacing the existing ones with replace('a', 'b'), when I find any.

Many thanks for the quick and detailed reply!
Matteo


 ___
 LibreOffice mailing list
 LibreOffice@lists.freedesktop.org
 http://lists.freedesktop.org/mailman/listinfo/libreoffice
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice