cobol lexer issues

Bug in https://github.com/pygments/pygments/blob/master/pygments/lexers/business.py

Given the test program

       IDENTIFICATION DIVISION.
       PROGRAM-ID.             TESTFUNC.
      *AUTHOR.                 ARNOLD J. TREMBLEY.
      *DATE-WRITTEN.           2020-06-30.

       DATA DIVISION.
       WORKING-STORAGE SECTION.
       01  800-WIN-USERNAME                PIC X(24)   VALUE "UNKNOWN".
       01  800-WIN-USERPROFILE             PIC X(24)   VALUE "UNKNOWN".
       01  800-WIN-USERDOMAIN              PIC X(24)   VALUE "UNKNOWN".
       01  800-WHEN-COMPILED.
           05  800-COMPILED-DATE-YYYY      PIC X(04)   VALUE SPACES.
           05  800-COMPILED-DATE-MM        PIC X(02)   VALUE SPACES.
           05  800-COMPILED-DATE-DD        PIC X(02)   VALUE SPACES.
           05  800-COMPILED-TIME-HH        PIC X(02)   VALUE SPACES.
           05  800-COMPILED-TIME-MM        PIC X(02)   VALUE SPACES.
           05  800-COMPILED-TIME-SS        PIC X(02)   VALUE SPACES.
           05  FILLER                      PIC X(07)   VALUE SPACES.
       01  800-CURRENT-DATE.
           05  800-CURRENT-DATE-YYYY       PIC X(04)   VALUE SPACES.
           05  800-CURRENT-DATE-MM         PIC X(02)   VALUE SPACES.
           05  800-CURRENT-DATE-DD         PIC X(02)   VALUE SPACES.
           05  800-CURRENT-TIME-HH         PIC X(02)   VALUE SPACES.
           05  800-CURRENT-TIME-MM         PIC X(02)   VALUE SPACES.
           05  800-CURRENT-TIME-SS         PIC X(02)   VALUE SPACES.
           05  FILLER                      PIC X(07)   VALUE SPACES.

       PROCEDURE DIVISION.
       MAINLINE.
           DISPLAY 'TESTFUNC Start - GnuCOBOL 3.1 RC1 30JUN2020'
           MOVE FUNCTION WHEN-COMPILED TO 800-WHEN-COMPILED
           MOVE FUNCTION CURRENT-DATE  TO 800-CURRENT-DATE
           DISPLAY 'TESTFUNC Compiled = '
               800-compiled-date-yyyy '/'
               800-compiled-date-mm   '/'
               800-compiled-date-dd   space
               800-compiled-time-hh   ':'
               800-compiled-time-mm   ':'
               800-compiled-time-ss
           DISPLAY 'TESTFUNC Executed = '
               800-current-date-yyyy '/'
               800-current-date-mm   '/'
               800-current-date-dd   space
               800-current-time-hh   ':'
               800-current-time-mm   ':'
               800-current-time-ss
           accept omitted
           ACCEPT 800-WIN-USERNAME FROM ENVIRONMENT "USERNAME"
           ACCEPT 800-WIN-USERDOMAIN FROM ENVIRONMENT "USERDOMAIN"
           ACCEPT 800-WIN-USERPROFILE FROM ENVIRONMENT "USERPROFILE"
           DISPLAY "USERNAME    = " 800-WIN-USERNAME
           DISPLAY "USERPROFILE = " 800-WIN-USERPROFILE
           DISPLAY "USERDOMAIN  = " 800-WIN-USERDOMAIN
           DISPLAY 'TESTFUNC Successfully Completed'
           STOP RUN.

all variables starting with a number are partial (only the number) highlighted as keyword, both in their definition 01 800-CURRENT-DATE. and in their use MOVE FUNCTION WHEN-COMPILED TO 800-WHEN-COMPILED.

In general the hyphen is not a separator in COBOL so 123-something will always be a user-defined word.

Here are some additional updates to the reserved word lists as those were currently from OpenCOBOL 1.1 while this has gone a long way to GnuCOBOL 3.2 (I think its best to drop the error-token keywords but instead create a context-sensitive token list): gc32res.txt

Start in pygments/lexers/business.py and reproduce the COBOL highlighting with the sample program, focusing on identifiers such as 800-CURRENT-DATE and 800-WHEN-COMPILED. Compare the lexer’s reserved-word handling with gc32res.txt. Done means hyphenated words beginning with numbers are treated as user-defined words rather than highlighting only their numeric prefix as a keyword.