The most straightforward approach is to make every rule (or "production") a function. This is called a "recursive descent" parser.
Write two routine that will peek at and get the next character as well.
This will give you some code that looks something like this (in pseudocode):
// <sequent> ::= <lhs> # <rhs>
sequent()
lhs()
if peekchar() != '#' then error
else poundsign = nextchar()
rhs()
// <lhs> ::= <formulalist>| ε
lhs()
if peekchar() == EOF
return
else
formula()
// <rhs> ::= <formulalist>| ε
rhs()
if peekchar() == EOF
return
else
formulalist()
// <formulalist> ::= <formula>|<formula> , <formulalist>
formulalist()
formula()
if peekchar() is ','
comma = nextchar()
return formulalist()
// <formula> ::= <letter>| - <formula>| (<formula><infixop><formula>)
formula()
next = peekchar()
if next in A..Z
letter
else if next is -
minus_sign = nextchar()
return formula()
else
formula()
infixop()
formula()
// <infixop> ::= & | | | >
infixop()
c = nextchar()
if c not in &,|,> then error
// <letter> ::= A | B | ... | Z
letter()
c = nextchar()
if c not A..Z then error
and so forth, for each rule.
The general idea:
- each rule is a function
- at certain points the function peeks ahead to see what to do. for example, formula() checks to see if the first character is a minus sign.