


                DIFF - A Text-file Differencing Program

     This is a utility to compare two text files (e.g.,  two versions of 
     a  program  source  file)  and print the lines which differ between 
     them.  The algorithm  used  is  that  of  J.  W.  Hunt  and  M.  D. 
     McIlroy,  "An  algorithm  for  differential file comparison",  Bell 
     Telephone Laboratories Computing Science Technical Report #41. 

     Usage is 

                     diff [option ...] file1 file2 

     The options, or switches, are in the UNIX style of -s -t etc.  They 
     cannot be concatenated (e.g.,  -st ),  but they  may  be  used  any 
     number  of times on the command line in any order.  If incompatible 
     switches, such as -s and -l, are both specified, the last-occurring 
     switch is valid. 

     The following options are available: 

          -l  -  "line" mode, i.e.,  compare  files  record  by  record,
          where a record is terminated by a newline.  Only the first 256
          characters of the records are compared and printed.   This  is
          the default.

          -s  -  "sentence" mode.  Here, records  are  terminated  by  a
          period,   colon,   or  semicolon.   Newlines  and  white-space
          characters are converted to spaces, and consecutive spaces are
          converted  to  a  single  space.   Records  also  end  at  256
          characters, but the remainder  of  the  record  is  not  lost;
          rather,  the  next  record  starts where the previous one left
          off.

          -c[n1[:[n2]]]  -  column restriction.  The comparison  between
          two records is restricted to columns n1 through n2.  Either or
          both numbers are optional;  if not present, each has a default
          (if  n2  is present, the colon must be included).  n1 defaults
          to 1, n2 to 256.

          -m[n1]  -  specifies the  number  of  records  constituting  a
          match;    that   is,  if  a  mismatch  is  detected,  then  n1
          consecutive pairs of matching lines must  be  found  before  a
          match is declared again.  n1 defaults to 1.

     A typical invocation might be

                    diff -c10:20 -m3 prog.v1 prog.v2

     This command would cause  the  files  prog.v1  and  prog.v2  to  be
     compared line by line, and those lines which differed in columns 10
     through 20 would be listed in sections.  A difference section would
     end  only  when  a  sequence  of  three  or more matching lines (in
     columns 10 through 20) were found.
                                                                  Page 2


     The output of DIFF is similar to the default output of the  VAX/VMS
     DIFFERENCES command.  However, the algorithms are not the same, and
     in some unusual circumstances their sets of  difference  lines  can
     differ  for  the  same values of the match record option.  However,
     the difference sets will always include all unmatched lines.


     Bugs (features) in V1.30 -

     The  filenames must be in the current directory.  This is a problem 
     caused by the Microsoft C compiler.  A later version of Lattice  C, 
     or  somebody  else's  compiler,  will  doubtless  permit  filespecs 
     including pathnames. Watch this space. 

     DIFF currently has no options for output format.

     DIFF will probably not be sensible about handling internal  formats
     used by word processors such as Wordmarc or Multi-mate.
     
     If you have any suggestions,  please  contact  C. K. Miller,  7613,
     4-4079.
 a‚ "  åV      EXE           x¼.
€3  