Question

How can I, in a c program, perform a glob by using the function provided by the z shell?

I have created a README of my explorations so far. It is for use in an open source library.

https://bitbucket.org/sentimental/zsh_source_experimentation/src/master/README

I copy it here:

---

Start

Lets get the sources

apt-get source zsh
apt-get source zsh-dev

I’ve discovered by using ldd that zsh does not produce any library files::

#ldd /bin/zsh4

        linux-gate.so.1 =>  (0xb7775000)
        libcap.so.2 => /lib/i386-linux-gnu/libcap.so.2 (0xb7751000)
        libdl.so.2 => /lib/i386-linux-gnu/libdl.so.2 (0xb774c000)
        libtinfo.so.5 => /lib/i386-linux-gnu/libtinfo.so.5 (0xb772c000)
        libm.so.6 => /lib/i386-linux-gnu/libm.so.6 (0xb7700000)
        libc.so.6 => /lib/i386-linux-gnu/libc.so.6 (0xb755b000)

I think I will have to use the source files directly.

Lets locate the files containing references to extended globbling
(I’m using zsh as my shell).::

grep -ir EXTENDEDGLOB . |  egrep "\.(c|h):"  | cut -d: -f1 | sort -u

./README
./zsh-4.3.17/Etc/ChangeLog-3.0
./zsh-4.3.17/Src/glob.c
./zsh-4.3.17/Src/Modules/zutil.c
./zsh-4.3.17/Src/options.c
./zsh-4.3.17/Src/pattern.c
./zsh-4.3.17/Src/utils.c
./zsh-4.3.17/Src/Zle/complist.c
./zsh-4.3.17/Src/Zle/zle_tricky.c
./zsh-4.3.17/Src/zsh.h

Lets consider a couple of those files

zsh.h

In here EXTENDEDGLOB is defined as part of an anonymous enum

There are publications

here
http://publications.gbdirect.co.uk/c_book/chapter6/enums.html

and here
http://bytes.com/topic/c/answers/63891-enum-within-function-standard

detailing the use of enum in c

An example of its use is probably the method arguments for the function
…………….

 static int
    bin_zregexparse(char *nam, char **args, Options ops, UNUSED(int func))

Found in the file
……………..

./zsh-4.3.17/Src/Modules/zutil.c

Lets see what’s calling that function. Hmm only one call. The only reference to that call is in that file.

grep -r  bin_zregexparse . |  egrep "\.(c|h):"  | cut -d: -f1 | sort -u

./zsh-4.3.17/Src/Modules/zutil.c

Hmm…. How does it work if nothing calls this function?

Ok lets see if there is some conditional configuration that sets up or aliases this code somehow?

 grep -i regex ./**/conf*

/zsh-4.3.17/config.h.in :/* Define to 1 if you have the `regexec' function. */
./zsh-4.3.17/config.h.in :#undef HAVE_REGEXEC
./zsh-4.3.17/config.h.in :/* Define to 1 if you have the `regexec' function. */
./zsh-4.3.17/config.h.in :#undef HAVE_REGEXEC
./zsh-4.3.17/configure :       regcomp regexecc regerror regfree \
./zsh-4.3.17/configure :       regcomp regexec regexecerror regfree \
./zsh-4.3.17/configure.ac :       regcomp regexec regerrorror regfree \

Lets investigate these files.

 config.h.in

Doesn’t seem to exist, perhaps it is generated?

There seems to be a block in

 configure


 8148 for ac_func in strftime strptime mktime timelocal \     
 8149          difftime gettimeofday clock_gettime \          
 8150          select poll \                                  
 8151          readlink faccessx fchdir ftruncate \           

 etc etc etc ..

 8178          htons ntohs \                                   
 8179          regcomp regexec regerror regfree \              
 8180          gdbm_open getxattr \                            
 8181          realpath canonicalize_file_name \               
 8182          symlink getcwd                                  
 8183 do :                                                     
 8184   as_ac_var=`$as_echo "ac_cv_func_$ac_func" | $as_tr_sh` 
 8185 ac_fn_c_check_func "$LINENO" "$ac_func" "$as_ac_var"     
 8186 if eval test \"x\$"$as_ac_var"\" = x"yes"; then :        
 8187   cat >>confdefs.h <<_ACEOF                              
 8188 #define `$as_echo "HAVE_$ac_func" | $as_tr_cpp` 1        
 8189 _ACEOF                                                   
 8190                                                          
 8191 fi                                                       
 8192 done                                                     

}}}

I’ve no idea what that is doing. TODO: Investigate!

In

  ./zsh-4.3.17/configure.ac 

  1167 dnl ---------------
  1168 dnl CHECK FUNCTIONS
  1169 dnl ---------------
  1170                                                                                                                                                                                  |     rglobdata.gd_gf_noglobdot
  1171 dnl need to integrate this function
  1172 dnl AC_FUNC_STRFTIME
  1173                                                                                                                                                                                  |     rglobdata.gd_gf_listtypes
  1174 AC_CHECK_FUNCS(strftime strptime mktime timelocal \
  1175          difftime gettimeofday clock_gettime \
  1176          select poll \

  etc etc etc 


  1205          regcomp regexec regerror regfree \                                 
  1206          gdbm_open getxattr \                                               
  1207          realpath canonicalize_file_name \                                  
  1208          symlink getcwd)                                                    
  1209 AC_FUNC_STRCOLL

Not really sure at this stage how I can do this, consider that I perhaps want to unit-test that function, how might I do so?

Answer 1

The function bin_zregexparse is the implementation of the zregexparse builtin provided by the zsh/zutil module. It is used below its definition in zutil.c:

static struct builtin bintab[] = {
    …
    BUILTIN("zregexparse", 0, bin_zregexparse, 3, -1, 0, "c", NULL),
    …
};

zregexparse is intended to be used in the implementation of _regex_arguments. This isn’t the most promising entry point.

If you want to implement zsh’s globbing features, you’ll have to pull in almost all of the zsh code into your program, since glob patterns can contain arbitrary embedded code. You can exclude the line editor during build, but that’s about it.

I would recommend using a separate zsh binary and feeding it requests through a pair of pipes.

setopt extended_glob null_glob
print -Nr -- **/*(.Om+my_predicate) ''