simple example of custom tokenizer

README.md

@@ -1,10 +1,12 @@
 # datafusion-sql
 
-DataFusion SQL Parser (v2)
+This is a work-in-progress to develop a new version of the DataFusion SQL Parser.
 
-Goals:
+Goals for this version:
 
 - Support for custom SQL dialects, so other projects can implement their own parsers easily
-- Zero-copy of tokens when parsing
 - Good error reporting (e.g. show line / column numbers and descriptive messages)
+- Zero-copy of tokens when parsing
+- Concise code
+
 

examples/acme_parser.rs

@@ -1,17 +1,23 @@
 use std::str::Chars;
+use std::iter::Peekable;
 
 extern crate datafusion_sql;
 
 use datafusion_sql::tokenizer::*;
+use datafusion_sql::generic_tokenizer::*;
 use datafusion_sql::parser::*;
 
-#[derive(Debug)]
+///
+/// This example demonstrates building a custom ACME parser that extends the generic parser
+/// by adding support for a factorial operator !!
+///
+
+#[derive(Debug,PartialEq)]
 enum AcmeToken {
     /// Factorial operator `!!`
     Factorial
 }
 
-
 #[derive(Debug)]
 enum AcmeOperator {
     Factorial
@@ -19,21 +25,35 @@ enum AcmeOperator {
 
 #[derive(Debug)]
 enum AcmeTokenizerError {
-
 }
 
 struct AcmeTokenizer {
-    //chars: &'a Chars
+    generic: GenericTokenizer
 }
 
+/// The ACME tokenizer looks for the factorial operator `!!` but delegates everything else
 impl SQLTokenizer<AcmeToken, AcmeTokenizerError> for AcmeTokenizer {
 
-    fn peek_token(&mut self) -> Result<Option<SQLToken<AcmeToken>>, TokenizerError<AcmeTokenizerError>> {
-        Ok(Some(SQLToken::Custom(AcmeToken::Factorial)))
-    }
-
-    fn next_token(&mut self) -> Result<Option<SQLToken<AcmeToken>>, TokenizerError<AcmeTokenizerError>> {
-        Ok(Some(SQLToken::Custom(AcmeToken::Factorial)))
+    fn next_token(&self, chars: &mut Peekable<Chars>) -> Result<Option<SQLToken<AcmeToken>>, TokenizerError<AcmeTokenizerError>> {
+        match chars.peek() {
+            Some(&ch) => match ch {
+                '!' => {
+                    chars.next(); // consume the first `!`
+                    match chars.peek() {
+                        Some(&ch) => match ch {
+                            '!' => {
+                                chars.next(); // consume the second `!`
+                                Ok(Some(SQLToken::Custom(AcmeToken::Factorial)))
+                            },
+                            _ => Err(TokenizerError::UnexpectedChar(ch,Position::new(0,0)))
+                        },
+                        None => Ok(Some(SQLToken::Not))
+                    }
+                },
+                _ => self.generic.next_token(chars)
+            }
+            _ => self.generic.next_token(chars)
+        }
     }
 }
 
@@ -43,9 +63,13 @@ fn main() {
 
     let sql = "1 + !! 5 * 2";
 
-    let mut tokenizer = AcmeTokenizer { };
+    let mut acme_tokenizer = AcmeTokenizer {
+        generic: GenericTokenizer { }
+    };
 
-    println!("token = {:?}", tokenizer.peek_token().unwrap());
+    let tokens = tokenize(&sql, &mut acme_tokenizer).unwrap();
+
+    println!("tokens = {:?}", tokens);
 
 
 

src/generic_tokenizer.rs

@@ -0,0 +1,41 @@
+use std::cmp::PartialEq;
+use std::fmt::Debug;
+use std::iter::Peekable;
+use std::str::Chars;
+
+use super::tokenizer::*;
+
+pub struct GenericTokenizer {}
+
+impl<S,TE> SQLTokenizer<S,TE> for GenericTokenizer
+    where S: Debug + PartialEq {
+
+    fn next_token(&self, chars: &mut Peekable<Chars>) -> Result<Option<SQLToken<S>>, TokenizerError<TE>> {
+        match chars.next() {
+            Some(ch) => match ch {
+                ' ' | '\t' | '\n' => Ok(Some(SQLToken::Whitespace(ch))),
+                '0' ... '9' => {
+                    let mut s = String::new();
+                    s.push(ch);
+                    while let Some(&ch) = chars.peek() {
+                        match ch {
+                            '0' ... '9' => {
+                                chars.next(); // consume
+                                s.push(ch);
+                            },
+                            _ => break
+                        }
+                    }
+                    Ok(Some(SQLToken::Literal(s)))
+                },
+                '+' => Ok(Some(SQLToken::Plus)),
+                '-' => Ok(Some(SQLToken::Minus)),
+                '*' => Ok(Some(SQLToken::Mult)),
+                '/' => Ok(Some(SQLToken::Divide)),
+                _ => Err(TokenizerError::UnexpectedChar(ch,Position::new(0, 0)))
+            },
+            None => Ok(None)
+        }
+    }
+}
+

src/lib.rs

@@ -1,2 +1,3 @@
 pub mod tokenizer;
+pub mod generic_tokenizer;
 pub mod parser;

src/parser.rs

@@ -1,3 +1,6 @@
+use std::cmp::PartialEq;
+use std::fmt::Debug;
+
 use super::tokenizer::*;
 
 #[derive(Debug)]
@@ -35,23 +38,29 @@ pub enum SQLExpr<T> {
 }
 
 #[derive(Debug)]
-pub enum ParserError<T> {
-    WrongToken { expected: Vec<SQLToken<T>>, actual: SQLToken<T>, line: usize, col: usize },
-    Custom(T)
+pub enum ParserError<S, PE>
+    where S: Debug + PartialEq {
+
+    WrongToken { expected: Vec<SQLToken<S>>, actual: SQLToken<S>, line: usize, col: usize },
+    Custom(PE)
 }
 
-impl<T> From<TokenizerError<T>> for ParserError<T> {
-    fn from(_: TokenizerError<T>) -> Self {
+impl<S, TE> From<TokenizerError<TE>> for ParserError<S, TE>
+    where S: Debug + PartialEq {
+
+    fn from(_: TokenizerError<TE>) -> Self {
         unimplemented!()
     }
 }
 
 
-trait Parser<S, PE> {
+trait Parser<S, PE>
+    where S: Debug + PartialEq {
+
     /// parse the prefix and stop once an infix operator is reached
-    fn parse_prefix(&mut self) -> Result<Box<SQLExpr<S>>, ParserError<PE>> ;
+    fn parse_prefix(&mut self) -> Result<Box<SQLExpr<S>>, ParserError<S, PE>> ;
     /// parse the next infix expression, returning None if the precedence has changed
-    fn parse_infix(&mut self, left: SQLExpr<S>) -> Result<Option<Box<SQLExpr<S>>>, ParserError<PE>>;
+    fn parse_infix(&mut self, left: SQLExpr<S>) -> Result<Option<Box<SQLExpr<S>>>, ParserError<S, PE>>;
 }
 
 //

src/tokenizer.rs

@@ -1,24 +1,40 @@
+use std::cmp::PartialEq;
+use std::fmt::Debug;
+use std::iter::Peekable;
+use std::str::Chars;
 
 #[derive(Debug)]
 pub struct Position {
     line: usize,
     col: usize
 }
+impl Position {
+    pub fn new(line: usize, col: usize) -> Self {
+        Position { line, col }
+    }
+}
 
 #[derive(Debug)]
 pub enum TokenizerError<T> {
+    UnexpectedChar(char,Position),
     UnexpectedEof(Position),
     UnterminatedStringLiteral(Position),
     Custom(T)
 }
 
 /// SQL Tokens
-#[derive(Debug)]
-pub enum SQLToken<T> {
+#[derive(Debug,PartialEq)]
+pub enum SQLToken<T: Debug + PartialEq> {
+    Whitespace(char),
     Keyword(String), //TODO: &str ?
     Identifier(String), //TODO: &str ?
     Literal(String), //TODO: need to model different types of literal
+    Plus,
+    Minus,
+    Mult,
+    Divide,
     Eq,
+    Not,
     NotEq,
     Gt,
     GtEq,
@@ -31,9 +47,32 @@ pub enum SQLToken<T> {
     Custom(T)
 }
 
-pub trait SQLTokenizer<S, T> {
-    /// return a reference to the next token without consuming it (look ahead)
-    fn peek_token(&mut self) -> Result<Option<SQLToken<S>>, TokenizerError<T>>;
+pub trait SQLTokenizer<S, TE>
+    where S: Debug + PartialEq {
+
     /// return a reference to the next token and advance the index
-    fn next_token(&mut self) -> Result<Option<SQLToken<S>>, TokenizerError<T>>;
+    fn next_token(&self, chars: &mut Peekable<Chars>) -> Result<Option<SQLToken<S>>, TokenizerError<TE>>;
+}
+
+
+pub fn tokenize<S,TE>(sql: &str, tokenizer: &mut SQLTokenizer<S,TE>) -> Result<Vec<SQLToken<S>>, TokenizerError<TE>>
+    where S: Debug + PartialEq
+    {
+
+    let mut peekable = sql.chars().peekable();
+
+    let mut tokens : Vec<SQLToken<S>> = vec![];
+
+    loop {
+        match tokenizer.next_token(&mut peekable)? {
+            Some(SQLToken::Whitespace(_)) => { /* ignore */ },
+            Some(token) => {
+                println!("Token: {:?}", token);
+                tokens.push(token)
+            },
+            None => break
+        }
+    }
+
+    Ok(tokens)
 }