# Last edited on 2004-12-30 17:48:14 by stolfi

OVERVIEW

  The input to the SELVA grammar preprocessor consists of four sections:

    MARKERS
      zero or more {MARKDEF}s
    SYMBOLS
      one or more {SYMDEF}s 
    RULES
      one or more {RULE}s.

  Comments start with "#" and extend to the end of the current line.

MARKERS SECTION

  The MARKERS section defines the valid marker types and their
  value ranges. Each {MARKDEF} has the form

    MARKDEF => MARKTYPE ":" MARKVAL |.. "\n"

    MARKTYPE => [A-Z]+
    
  A marker type is identified by one or more uppercase letters:
  "G", "P", "EO".

    MARKVAL => [a-z0-9]+
    
  Each marker type has a set of possible values ({MARKVAL}s) which 
  are strings of lowercase letters and/or numbers: "sin", "fem",
  "1", "p1".

SYMBOLS SECTION

  The SYMBOLS section defines the terminal and non-terminal 
  symbols, and their markers. Each entry has the form

    SYMDEF =>  {"*"}? SYMBOL {"(" MARKVAR ,.. ")"}? "\n"

  The names of the {MARKVAR}s (formal parameters) in the definition of
  a {SYMBOL} must be all distinct, for documentation purposes; but
  only their {MARKTYPE}s are significant. The prefix "*" specifies
  that the symbol is not important and should be "short-circuited"
  when displaying the parse trees.

    SYMBOL => NTSYMBOL | TSYMBOL

    NTSYMBOL => [A-Z][A-Za-z0-9_+]*

    TSYMBOL => [a-z][a-z0-9_]*
    
  By convention, non-terminal symbols start with an upper case letter,
  and terminal symbols with lowercase.  Both classes must be defined
  in the SYMBOLS section and may have marker parameters.
  
    MARKVAR => MARKTYPE MARKSUFF?

    MARKSUFF => [0-9]

  The type of a marker variable is determined by the letters in its
  name (upper-case only): thus "P" and "P2" have {MARKTYPE} = "P".
  Numeric suffixes allow more than one marker variable of the same
  type to appear in a single rule.

RULES 

  Grammar rules have the form

    RULE => NTSYMBOL "{" RTAG "}" {"(" MARKER,.. ")"}? "->" "\n" FACTOR "\n".. "." "\n"

  Each factor of the right-hand side must be on a separate line. The
  rule ends with a period on a line by itself.

    RTAG => [a-z][a-z0-9]*

  Each rule for the same {NTSYMBOL} must have a distinct {RTAG}.

    MARKER => MARKVAR { ":" MARKVAL |.. }? | MARKVAL

  The types of the {MARKER} parameters (values and/or variables) 
  must agree with the {MARKTYPE}s listed in the SYMBOLS section
  for the {NTSYMBOL}.

    FACTOR => { FUNCLABEL ":"} KERNEL

    FUNCLABEL => [A-Z][A-Za-z0-9_+]*

  The {FUNCLABEL} is used to label the factors according to their
  syntactic function. An occurrence of a factor like "F:X"
  automatically generates a rule "F:X -> X" in the expanded grammar.

    KERNEL => ATOM "?"
           => ATOM "^" { MARKVAR | [01] }
           => ATOM

  The "?" notation means that the {KERNEL} is either {ATOM} or
  empty. In the "^" notation, the exponent (the {MARKVAL}, or all
  possible values of the {MARKVAR}) must be either "0" (meaning that
  the {KERNEL} is empty) or "1" (meaning that the {KERNEL} is just
  the {ATOM}).

    ATOM => NTSYMBOL {"{" RTAG ,.. "}"}?  {"(" MARKER,.. ")"}? 
         => TSYMBOL {"(" MARKER,.. ")"}?

  The types of the markers (values or variables) must agree with the
  {MARKTYPE}s listed in the symbol's {SYMDEF}. The {RTAG}s must be
  all distinct, and each of them must match the {RTAG} of a {RULE}
  for the {NTSYMBOL}.