From looking at the code, the endDelimiter regular expression in the formatted SQL changelog appears to be looking for word characters (\w+). Do you see any problems with changing this to zero or more non-whitespace characters (\S*)? This would allow you to set the end delimiter to be empty or to utilize non-alphanumeric characters.
- In FormattedSqlChangeLogParser.parse:
- Pattern endDelimiterPattern = Pattern.compile(".*endDelimiter:(\\S*).*", Pattern.CASE_INSENSITIVE);