Where does `ss` set up flex input to come from command line arguments?

Question

I heard that ss uses Bison to parse command line arguments. Bison takes input from output Flex. Flex takes input from stdin by default. Where does ss set up flex input to come from command line arguments?

Answer 1

Bison calls yylex() whenever it needs a token. The return value of yylex() is a small integer, which is either one of the enumeration constants declared with %token in the bison grammar, or a character constant other than 0, or 0 to indicate EOF.

yylex() must also:

• Save the token's semantic value in the global variable yylval, if the token has a semantic value; and

• Save the token's location in the input stream in the global variable yylloc, if the parser actions make use of location information.

You can build a scanner using the flex command line tool. But you are under no obligation to do so. You could use a different tool to generate a scanner. Or you could define a yylex() function yourself. The interface for the scanner has been kept as simple as possible to simplify implementation.

Many tools, including ss, use bison with a hand-coded lexical analyzer. So the answer to your question is that ss never sets up the flex input because it doesn't use flex. (The fact that this is possible does not necessarily make it a good idea. Many hand-written scanners are significantly slower than flex, and the code is verbose and hard to reason about.)

---

Other intetfaces

The above is the default interface between yyparse and yylex. There are alternatives:

• You can define yylex to take additional arguments, if your yylex implementation requires additional data. Normally, these additional arguments will be provided to yyparse() so that the parser can use them in its call to yylex(). See the bison manual section on the %param directive.

• You can tell bison to generate a pure parser which does not rely on global variables, in which case it will call yylex with a pointer to a semantic object and, if location information is used, a pointer to a location object. See the bison manual chapter on pure parser interface.

• You can request a "push parser", which is called by the scanner every time a token becomes available. See the bison manual chapter on push parsers.