[Octave-cvsupdate] SF.net SVN: octave:[9913] trunk/octave-forge/main/strings

[<car...@us...>]

Revision: 9913
          http://octave.svn.sourceforge.net/octave/?rev=9913&view=rev
Author:   carandraug
Date:     2012-03-16 02:59:01 +0000 (Fri, 16 Mar 2012)
Log Message:
-----------
cstrcmp: complete rewrite of the function. Fix bugs and tests and implements same API as strcmp

Modified Paths:
--------------
    trunk/octave-forge/main/strings/NEWS
    trunk/octave-forge/main/strings/inst/cstrcmp.m

Modified: trunk/octave-forge/main/strings/NEWS
===================================================================
--- trunk/octave-forge/main/strings/NEWS	2012-03-15 22:39:10 UTC (rev 9912)
+++ trunk/octave-forge/main/strings/NEWS	2012-03-16 02:59:01 UTC (rev 9913)
@@ -7,3 +7,7 @@
       strtrim
 
   ** Package is no longer automatically loaded.
+
+  ** The function `cstrcmp' has been completely rewritten. It should
+     perform faster and will accept arguments exactly the same way as
+     Octave's core `strcmp'.

Modified: trunk/octave-forge/main/strings/inst/cstrcmp.m
===================================================================
--- trunk/octave-forge/main/strings/inst/cstrcmp.m	2012-03-15 22:39:10 UTC (rev 9912)
+++ trunk/octave-forge/main/strings/inst/cstrcmp.m	2012-03-16 02:59:01 UTC (rev 9913)
@@ -1,4 +1,5 @@
 ## Copyright (C) 2007 Muthiah Annamalai <mut...@ut...>
+## Copyright (C) 2012 Carnë Draug <car...@gm...>
 ##
 ## This program is free software; you can redistribute it and/or modify it under
 ## the terms of the GNU General Public License as published by the Free Software
@@ -14,68 +15,97 @@
 ## this program; if not, see <http://www.gnu.org/licenses/>.
 
 ## -*- texinfo -*-
-## @deftypefn {Function File} @var{rval}= {} cstrcmp(@var{string1},@var{string2})
-## compares the  @var{string1} and @var{string2} like the C-library
-## function, returning 0 on match, +1 if @var{string1} > @var{string2}
-## and -1 if @var{string1} < @var{string2}.
-## This operation is asymmetrical. If either @var{string1} or
-## @var{string2} are cell-arrays (not both simultaneously), then the
-## return value @var{rval} will be a similar cell-array of strings.
+## @deftypefn {Function File} {@var{rval} =} cstrcmp (@var{s1}, @var{s2})
+## Compare strings @var{s1} and @var{s2} like the C function.
 ##
+## Aside the difference to the return values, this function API is exactly the
+## same as Octave's @code{strcmp} and will accept cell arrays as well.
+##
+## @var{rval} indicates the relationship between the strings:
+## @itemize @bullet
+## @item
+## A value of 0 indicates that both strings are equal;
+## @item
+## A value of +1 indicates that the first character that does not match has a
+## greater value in @var{s1} than in @var{s2}.
+## @item
+## A value of -1 indicates that the first character that does not match has a
+## match has a smaller value in @var{s1} than in @var{s2}.
+## @end itemize
+##
 ## @example
 ## @group
-##           cstrcmp('marry','marie') 
-##           ##returns value +1
+## cstrcmp("marry","marry")
+##       @result{}  0
+## cstrcmp("marry","marri")
+##       @result{} +1
+## cstrcmp("marri","marry")
+##       @result{} -1
 ## @end group
 ## @end example
+##
+## @seealso {strcmp, strcmpi}
 ## @end deftypefn
-## @seealso {strcmp}
 
 function rval = cstrcmp (s1, s2)
 
-  if nargin < 2
+  if (nargin != 2)
     print_usage();
   endif
-  v1 = iscell (s1);
-  v2 = iscell (s2);
 
-  if (v1+v2) == 2;
-    error("only one argument can be a cell-array");
-  endif
+  ## this function is just like Octave's strcmp but the 0 and 1 need to be
+  ## inverted. Once is done, if there are 1, we need to decide if they will
+  ## be positive or negative. Also, since it's possible that the value needs
+  ## to be negative, class must be double (strcmp returns logical)
+  rval = double(!strcmp (s1, s2));
 
-  if (v1+v2) == 0
-    rval = do_cstrcmp (s1, s2);
+  if (!any (rval))
+    ## all zeros, no need to do anything else
     return
   endif
 
-  if(v2)
-    [s1, s2] = deal (s1, s2);
+  ## get index of the ones we have to "fix"
+  idx = find (rval == 1);
+  ## if any is not a cell, this simplifies the code that follows
+  if (!iscell (s1)), s1 = {s1}; endif
+  if (!iscell (s2)), s2 = {s2}; endif
+  ## there's 2 hypothesis:
+  ##  - arrays have same length (even if it's only one cell)
+  ##  - arrays have different lengths (in which case, one will have a single cell)
+  if (numel (s1) == numel (s2))
+    rval(idx) = cellfun (@get_sign, s1(idx), s2(idx));
+  elseif (numel (s1) > 1)
+    rval(idx) = cellfun (@get_sign, s1(idx), s2(1));
+  elseif (numel (s2) > 1)
+    rval(idx) = cellfun (@get_sign, s1(1), s2(idx));
   endif
-
-  L     = length (s1);
-  rval  = zeros (1, L);
-  for idx=1:L
-    rval(idx )= do_cstrcmp (s1{idx}, s2);
-  endfor
 endfunction
 
-function v = do_cstrcmp (s1, s2)
-  L2 = length (s2);
-  L1 = length (s1);
-  L  = min (L1, L2);
-  for idx=1:L
-    p = s1(idx);
-    q = s2(idx);
-    if (p != q)
-      v = sign (p-q);
-     return
-    endif
-  endfor
-  v = sign (L1-L2);
+function r = get_sign (s1, s2)
+  ## strings may have different lengths which kinda complicates things
+  ## in case the strings are of different size, we need to make them equal
+  ## If once "trimmed", the strings are equal, the "shortest" string is
+  ## considered smaller since the comparison is made by filling it with null
+
+  ns1  = numel (s1);
+  ns2  = numel (s2);
+  nmin = min (ns1, ns2);
+
+  ## if one of the strings is empty, we are already done
+  if (nmin == 0), r = sign (ns1 - ns2); return endif
+
+  s  = sign (s1(1:nmin) - s2(1:nmin));
+  if (any (s))
+    ## if there's any difference between this part of the two strings, get the
+    ## index of the first occurence and return its value
+    r = s(find (s != 0, 1));
+  else
+    r = sign (ns1 - ns2);
+  endif
 endfunction
 
-%!assert(cstrcmp("hello","hello"),0,0)
-%!assert(cstrcmp('marry','marie'),+1,0)
-%!assert(cstrcmp('Matlab','Octave'),-1,0)
-%!assert(cstrcmp('Matlab',{'Octave','Scilab','Lush','Yorick'}),[-1,-1,+1,-1],[])
-%!assert(cstrcmp({'Octave','Scilab','Lush','Yorick'},'Matlab'),[+1,+1,-1,+1],[])
+%!assert(cstrcmp("hello","hello"),0);
+%!assert(cstrcmp("marry","marie"),+1);
+%!assert(cstrcmp("Matlab","Octave"),-1);
+%!assert(cstrcmp("Matlab",{"Octave","Scilab","Lush"}), [-1 -1 +1]);
+%!assert(cstrcmp({"Octave","Scilab","Lush"},"Matlab"), [+1 +1 -1]);

This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.