A Parsing Expression Grammar for parsing an SQL subset. Written as a start to parsing SQL and generating Veritas relations (see: https://github.com/dkubb/veritas )

sql.citrus

grammar SQL
	rule statement
			select_statement <Veritas::SQL::Select> 
		| set_operation_statement <Veritas::SQL::SetOperation>
		| "(" [\s]* statement [\s]* ")" 
	end

	rule set_operation_statement
		"(" [\s]* left:select_statement <Veritas::SQL::Select> [\s]* ")" 
		[\s]* set_operator ([\s]+ distinct:`DISTINCT`)? [\s]* 
		"(" [\s]* right:select_statement <Veritas::SQL::Select> [\s]* ")"
	end
	
	rule select_statement
		`SELECT` ([\s]+ distinct:`DISTINCT`)? [\s]+ target_list <Veritas::SQL::Header>
				[\s]+ `FROM` [\s]+ relation <Veritas::SQL::Relation>
			(	[\s]+ `WHERE` [\s]+ predicate <Veritas::SQL::Predicate>)? 
			(	[\s]+ order_statement <Veritas::SQL::Order> )?
			(	[\s]+ limit_statement <Veritas::SQL::Limit> ([\s]+ offset_statement <Veritas::SQL::Offset>)?)? 
	end

	rule target_list
		"*" | (target <Veritas::SQL::HeaderTarget> ([\s]* "," [\s]* target <Veritas::SQL::HeaderTarget>)*)
	end

	rule target
		value ([\s]+ `AS` [\s]+ identifier)? 
	end

	rule relation
		( "(" [\s]* (select_statement | relation) [\s]* ")" ([\s]* `AS` [\s]+ identifier)? | identifier )
		( [\s]+ `INNER` [\s]+ `JOIN` [\s]+ relation ([\s]+ `ON` [\s]+ predicate)? )*
	end

	rule order_statement
		`ORDER BY` [\s]+ order_target ([\s]* "," [\s]* order_target)*
	end
	
	rule order_target
		qualified_identifier ([\s]+ (`DESC` | `ASC`))?
	end

	rule predicate
			simple_predicate [\s]+ boolean_infix [\s]+ predicate
		| negation [\s]+ predicate
		| simple_predicate
		| "("[\s]* predicate [\s]*")" ([\s]+ boolean_infix [\s]+ predicate)*
	end
	
	rule simple_predicate
			value [\s]* comparator [\s]* value 
		| value [\s]+ (negation [\s]+)? (between | in_statement)
		| value
	end
	
	rule between
		`BETWEEN` [\s]+ value [\s]+ `AND` [\s]+ value
	end
	
	rule in_statement
		`IN` [\s]* value_list
	end
	
	rule value_list
		"(" value ([\s]*","[\s]* value)* ")"
	end
	
	rule value
			"(" [\s]* value [\s]* ")" ([\s]* operator [\s]* value)?
		| scalar_function
		|	simple_value [\s]* operator [\s]* value
		| simple_value
		| qualified_identifier
		| identifier
	end

	rule scalar_function
		function_name "(" value ([\s]* "," [\s]* value)* ")"
	end
	
	rule simple_value
		identifier | literal_value
	end
	
	rule literal_value
		numeric | string
	end
	
	rule qualified_identifier
		( identifier | ("\"" identifier "\"") )("." (identifier | "\"" identifier "\""))?
	end
	
	######################################
	# Lexical rules
	######################################

	rule limit_statement
		`LIMIT` [\s]+ [0-9]+
	end
	
	rule offset_statement
		`OFFSET` [\s]+ [0-9]+
	end
	
	rule function_name
		[A-Z]+
	end
	
	rule boolean_infix
		`AND` | `OR` | `XOR`
	end
	
	rule negation
		`NOT`
	end

	rule comparator
		( "=" | "<>" | ">=" | ">" | "<=" | "<" )
	end
	
	rule operator
		( "+" | "-" | "*" | "/" | "%" | "sqrt" | "**" )
	end
	
	rule set_operator
		`UNION` | `INTERSECT` | `EXCEPT`
	end
	
	rule identifier
		[A-Za-z_][A-Za-z0-9_]*
	end
	
	rule numeric
		[-+]?[0-9]+ ("." [0-9]+)? (`E` [-+]? [0-9]+)?
	end
	
	rule string
		"'" ('\\\'' | /[^']+/)+ "'"
	end
end

@knowtheory: is this updated from the original? I wonder how it would do against the current generated SQL, which is 99% complete.

No, sorry this isn't, i posted this for someone else to see. I should update more though : )

Yeah, I'd love to see that. I've verified all the generated SQL against PostgreSQL, so I'm sure it's valid. The only things I need to do is some cleanup on my summarization branch (it's functional).