Chapter 2: Differences between flex and flexc++

Although flexc++ tries to be as compatible as possible with flex(1), there are some noteworthy differences. This chapter provides a quick overview for the users already familiar with flex.

2.1: Format of the input file

In flex(1) initializating code can be provided in the definition section (see section 3.1), and pre-match code can be provided as the first lines in the rules section.

Flexc++ does not support code blocks. Since flexc++ generates a class with appropriate header files, initialization code can be activated bij flexc++'s constructor, code to be executed when lex is called can be placed in the member lex's body, just before calling lex_, and Scanner::preCode can be provided with any required pre-match code. See also generated files 2.3 below.

Flexc++ also does not support a trailing `user code' section, where additional code can be placed to be copied verbatim to the source file. Again, the Scanner class approach offers preferable means for adding user code to the scanner.

Sections 2.1.1, 2.1.2 and 2.1.3 cover items which are no longer supported in flexc++, offering alternatives.

2.1.1: Definition section

2.1.2: Rules section

2.1.3: User code section

2.2: Patterns

Not all patterns supported by flex(1) are supported by flexc++. Notably, flexc++ does not yet support certain flags in regular expressions, like the flag to define case-insensitive regular expressions, or the flag allowing white space in regular expressions.

Another minor difference is that named patterns, defined in the definion section, cannot be used if they contain the lookahead operator (`/'). This is the result of the way name expansions are handled by flexc++. Flexc++ handles name expansions as a parenthesized regular expression (a group). Since groups may occur any number of times in a regular expression but a lookahead operator only once, the look-ahead operator is not accepted in a named pattern.

2.3: Generated files

While flex(1) only generates a file lex.yy.cc, flexc++ generates several files: several header files and a source file. By default flexc++ generates a class header file (Scanner.h), and internal header file (Scanner.ih), a base class header file (Scannerbase.h), and the file lex.cc containing the implementation of the required members of the class Scanner. Scannerbase.h and lex.cc should not be edited: they are overwritten whenever flexc++ is invoked. The other files (Scanner.h and Scanner.ih) are generated only once, and can thereafter be modified by the user (e.g., to add members to the Scanner class).

2.4: Comment

Flexc++ supports traditional C and C++ style end-of-line comment. Flexc++ handles comment more flexible than flex. Cf. section 3.3 for details.

2.5: Members and macros

As C++ supports namespaces, the yy-prefix is no longer needed. In addition, as flexc++ generates a scanner class, member functions rather than macros can (should) be used. See the conversion table below.

flex flexc++ flexc++ alternative
yylex() lex()
YYText() matched()
YYLeng() length()
ECHO echo()
yymore() more()
yyless() redo() accept()
BEGIN startcondition begin(StartCondition_::startcondition)
YY_AT_BOL n.a.
yy_set_bol(at_bol) n.a.

The member functions in the flexc++ column above are either members of Scanner or of its base class. Flexc++ does not use macros. All member functions can be used from within actions or by other member functions.

2.6: Multiple input streams

Multiple input files can easily be handled by flexc++. See section 4.1 for details.