DN474374687

Subject: Request Tcl 8.1 Updated string command, take 2 - DN [1]

scripticsBugForm@auto.genned.post - 05 May 1999 - comp.lang.tcl

 Tcl 8.1 Request:  Generated by Scriptics' bug entry form at
     http://www.scriptics.com/support/bugForm.html
 Responses to this post are encouraged.
 ------

 Submitted by:  Jeffrey Hobbs
 OperatingSystem:  Sun Solaris
 OperatingSystemVersion:  2.5.1
 Synopsis:  Updated string command, take 2

 DesiredBehavior:
     This is related to the previous feature request for an updated string
     command, with one new method (string is) and patched against 8.1.0.

     To cover it all again, I have made the following mods:

     * string compare now takes optional length arg
       (for strncmp behavior)

     * added string equal (just a few lines of code blended
       in with string compare)

     * added string icompare/iequal for case-insensitive comparisons

     * string index's index can now be ?end[+-]?expression
         I made this change in the private TclGetIntForIndex,
         which means that the list commands also benefit, as
         well as string range, et al.

     * added string is <class> ?-strict? ?-failindex varname? string
         The following are valid classes:
         alnum, alpha, ascii, boolean, digit, double, false,
         integer, lower, space, true, upper, wordchar
         Most of the correspond to the std is*() C functions.  Those
         that are missing (like iscntrl()) are due to a lack of
         Tcl_UniCharIs*() equivalent, but could be added.  I chose
         not to mess with adding them, as that means going in and
         setting things in the stub tables and such.  However, they
         are not complicated (zB, *IsControl() is a one liner).

         Those that do not correspond to is*() (like boolean, int,
         ...) are intelligently handled to return the correct value,
         but the command is written such that no type conversion of
         the passed object occurs (a good thing, I believe).  For
         integer and double, this was tricky, but the test suite is
         full, and the C code is well-commented with the purpose.

         With the exception of int/double, which are inherently in
         Tcl not unicode safe (still calls to strto*), all of these
         are unicode safe (with tests).

     * added [string repeat string count]
         Repeats given string <count> number of times

     * added string replace, quasi string equiv to lreplace
       (quasi opposite of string range):
         string replace first last ?string?
         Example of use, replacing end of string with ...
         should the string be more than 16 chars long:
         string replace $string 16 end "..."
         This just returns the string len < 16, so it
         will only affect the long strings.

     * added optional first and last args to string to*
         This allows you to just affect certain regions of
         a string with the command (like just capping the
         first letter).  I found the original totitle to
         be too draconian to be useful.

     * added [string map charMap string]
         where charMap is a {from to from to} list that equates to
         what one might get from [array get].  Each <from> and <to>
         can be multiple chars (or none at all).  For Tcl/CGI users,
         this is a MAJOR speed booster.

 Patch:
 The patch comes to 100KB with full doc and test suite changes
 against the Tcl 8.1.0 sources.
     http://www.hobbs.wservice.com/tcl/capp/Tcl81-string.txt

 PatchFiles:
     C: tclUtil.c && tclCmdMZ.c
     doc: string.n
     tests: string.test cmd*.test l*.test error.test

 Comments:
     The implementations were done to maintain 100% backwards compatibility
     with working code (changing the options to an arg can obviously change
     what error strings arise), and reduce any loss of speed (the only
     method that actually attained a speed loss was string index, as I
     throw a Tcl_NumUtfChars call in to support end-1 like indices, and
     that is not major).