Dispatch is not working in StringReplace

Question

I tried to improove a function I wrote to modify "castellano" the proper name for spanish to coherent (so that there's no exception rules to phonetics). I made this with StringReplace with a long list of rules.

euzk[cad_] := 
 StringReplace[
  cad, {"á" -> "a", "é" -> "e", "í" -> "i", "ó" -> "o", "ú" -> "u", 
   "ca" -> "ka", "ce" -> "ze", "ci" -> "zi", "co" -> "ko", 
   "cu" -> "ku", "ch" -> "c", "cl" -> "kl", "cr" -> "kr", 
   "cc" -> "kz", "cg" -> "kg", "cn" -> "kn", "cp" -> "kp", 
   "ct" -> "kt", "ge" -> "je", "gi" -> "ji", "gue" -> "ge", 
   "gui" -> "gi", "h" -> "", "qu" -> "k", "v" -> "b", "w" -> "u", 
   "y" -> "i"}]

As I read about Dispatch and how can make functions run faster I made

eusk = {"á" -> "a", "é" -> "e", "í" -> "i", "ó" -> "o", "ú" -> "u", 
   "ca" -> "ka", "ce" -> "ze", "ci" -> "zi", "co" -> "ko", 
   "cu" -> "ku", "ch" -> "c", "cl" -> "kl", "cr" -> "kr", 
   "cc" -> "kz", "cg" -> "kg", "cn" -> "kn", "cp" -> "kp", 
   "ct" -> "kt", "ge" -> "je", "gi" -> "ji", "gue" -> "ge", 
   "gui" -> "gi", "h" -> "", "qu" -> "k", "v" -> "b", "w" -> "u", 
   "y" -> "i"};

then

deusk = Dispatch[eusk]

and finally

euzk[cad_] := StringReplace[cad, deusk]

But when try to use it... receive

StringReplace::srep: Dispatch[Length: 27] is not a valid string replacement rule.

I can of course leave the function as list of rules, but Dispatch looks like so efficient. Can I use it otherwise? Some advice ?

I believe StringReplace compiles to RegularExpression generally, making Dispatch is unnecessary. — b3m2a1, Jan 16 '18 at 09:58
@-b3m2a1 so I must consider that this faster capacity is already got otherwise ? — Anxon Pués, Jan 16 '18 at 10:12
@AnxonPués Yes, exactly. But writing your rules directly in terms of RegularExpression sometimes can give you a performance gain (and sometimes opposite). — Alexey Popkov, Jan 16 '18 at 10:21

Alexey Popkov · Answer 1 · 2018-01-16T18:19:07.017

Dispatch is designed to optimize lists of replacement rules for the purposes of such functions as Replace, ReplaceAll, ReplaceRepeated and ReplaceList, it is not intended to be used with string patterns (StringReplace, StringReplaceList, StringCases). Mathematica translates string patterns to regular expressions which will be further compiled by the PCRE library and cached (see more info here and here), what already gives huge speed-up.

In your particular case you should just use your replacement rules "as is" for achieving the best possible performance:

euzk[cad_] := StringReplace[cad, eusk]

In some other cases when you need to replace not a literal substring, but a string pattern, rewriting your string pattern (or a part of the pattern) in terms of RegularExpression can give a performance gain, but not always.

Dispatch is not working in StringReplace

1 Answers1