If you are editing a data set that contains both EBCDIC and DBCS data, note these rules about CHANGE strings:

• The SO and SI characters that delimit the CHANGE string are used as part of the string only if necessary. If you specify replacement of an EBCDIC string with a DBCS string, they are used. If you specify replacement of a DBCS string with another DBCS string, they are not used.
• If you specify in a CHANGE string that an SO or SI character be changed to another character, the result is unpredictable.
• If you specify a CHANGE string that causes a field length of zero and the boundary falls between the SO and SI characters, the SO/SI or SI/SO character strings that are next to each other are replaced with a DBCS blank. If the boundary does not fall between the SO and SI characters, the SO/SI or SI/SO characters that are next to each other are removed.
• If the lengths of the two strings specified in CHANGE are different, these actions occur:
  • If string1 is shorter than string2, the data to the right of string1 is shifted to the left up to some breakpoint. Breakpoints include the border between an EBCDIC field and a DBCS field, a double or single blank, or the right boundary set by a BOUNDS command.
  • If string1 is longer than string2, blanks in the record to the right of string1 are used to make room. When blanks in a DBCS field are used, they are used in units of 2 bytes.
• If a DBCS field crosses the right boundary, CHANGE can cause an odd-length DBCS field. If this happens, the right boundary is ignored and the operation takes place.