Context-free grammar

A grammar is context-free (CFG) if the rules of the grammar do not depend on the context in which the tokens appear. Context-free grammars consist of:

• A set of terminal symbols $\Sigma$.
  • i.e., symbols that appear in the input stream (tokens).
  • By convention, these are in lower-case letters, operators, digits, brackets.
• A set of non-terminal symbols $\mathcal{N}$.
  • i.e., a set of “intermediate” symbols that help us express the grammar. We substitute these out for tokens that eventually become terminals.
  • By convention, these start with upper-case letters.
• A start symbol $S\in \mathcal{N}$. It’s denoted $S$.
  • i.e., a special non-terminal symbol that refers to the whole language.
• A set of productions $\mathcal{P}$.
  • These express what the language is. We can re-write non-terminals in terms of non-terminals and terminals.
• We also define $$$ as the end-of-input token.

We also define a derivation of a sentence as a possible conversion from a sentence to a production rule (also called a sentential form) used to describe it.

Productions

Productions can be described as followed. The start production is often the first one.

$$\mathcal{P}:S\to A$$

There could be a recursive definition in our grammar. The below is a sentential form of $A$.

$$A=aAb$$

The arrow means one can re-write or express the LHS of the arrow as the RHS of the arrow. Only non-terminals appear on the LHS (and only one for CFGs). Note that we can use the vertical pipe $|$ to express alternatives.

$$A\to ϵ$$

or alternatively:

$$A\to aAb|ϵ$$

These lines and this convention allows us to specify a grammar only with its productions.

Derivation

Given a more complicated grammar, we have two key questions: which production do we use? Which non-terminal in a production do we re-write?

The leftmost (rightmost) derivation is where we consistently re-write the leftmost (rightmost) non-terminal in a sentential form. We try to be consistent with being leftmost or rightmost. Ideally our grammar must be defined in a way that it’s possible to use a single consistent approach. We also define some important sets. Given a sentential form $s$ of grammar symbols (terminals and non-terminals), the first set $\mathcal{F}irst(s)$ is the set of terminals that can begin any sentential from derived from $s$.

Given a non-terminal symbol $s$, the follow set $\mathcal{F}ollow(s)$ is the set of terminals that can appear immediately after $s$ in some derivation. We can compute the follow set by repeating these steps until nothing can be added to any $\mathcal{F}ollow$ set:

• Add \$$ to \mathcal{Follow}(S)$,where$S$isthestartsymboland$$$ is the end-of-input symbol.
• For a production $\mathcal{A}\to \alpha B\beta$, add all elements of $\mathcal{F}irst(\beta )$, except for $ϵ\in \mathcal{F}irst(\beta )$, add $\mathcal{F}ollow(A)$ to $\mathcal{F}ollow(B)$.
• If productions $A,B$ occur where $B$ follows $A$ in a production rule, and if $ϵ\in \mathcal{F}irst(B)$, then we add the elements of $\mathcal{F}ollow(B)$ into $\mathcal{F}ollow(A)$.