home | O'Reilly's CD bookshelfs | FreeBSD | Linux | Cisco | Cisco Exam  

Java Language Reference

Previous Chapter 2 Next

2. Lexical Analysis


When the Java compiler compiles a program, the first thing it does is determine the structure of the program. The compiler reads the characters in the program source and then applies rules to recognize progressively larger chunks of the file, such as identifiers, expressions, statements, and classes. The process of discovering the organization of the program is divided into two components:

  • The lexical analyzer. This component looks for sequences of characters called tokens that form identifiers, literals, operators, and the like.

  • The parser. This component is responsible for discovering higher levels of organization in the sequences of tokens discovered by lexical analysis.

This chapter describes the rules governing the lexical analysis of Java programs. The rules governing the parsing of Java programs are described over the course of subsequent chapters.

The lexical analysis rules for Java can appear slightly ambiguous. Where ambiguity occurs, the rules for interpreting character sequences specify that conflicts are resolved in favor of the interpretation that matches the most characters. That's a bit confusing, so an example should help. Take the character sequence:


The ambiguity is that the sequence could potentially be interpreted as either + followed by ++ or the other way around; both are valid tokens. But according to the lexical analysis rules that insist that tokenization favor the longest character match, Java interprets the character sequence as:

++ +

Because ++ is longer than +, Java first recognizes the token ++ and then the +.

These rules can produce undesired results when character sequences are not separated by white space. For example, the following sequence is ambiguous:


The programmer probably intended this sequence to mean "x + (+y)", but the lexical analyzer always produces the token sequence "x ++ y". This sequence is syntactically incorrect.

Java lexical analysis consists of two phases: pre-processing and tokenization. The pre-processing phase is discussed in the following section. The tokenization phase is responsible for recognizing the tokens in the pre-processed input and is discussed later in this chapter.

2.1 Pre-Processing

A Java program is a sequence of characters. These characters are represented using 16-bit numeric codes defined by the Unicode standard.[1] Unicode is a 16-bit character encoding standard that includes representations for all of the characters needed to write all major natural languages, as well as special symbols for mathematics. Unicode defines the codes 0 through 127 to be consistent with ASCII. Because of that consistency, Java programs can be written in ASCII without any need for programmers to be aware of Unicode.

[1] Unicode is defined by an organization called the Unicode Consortium. The defining document for Unicode is The Unicode Standard, Version 2.0 (published by Addison-Wesley, ISBN 0-201-48345-9). More recent information about Unicode is available at http://unicode.org/.

Java is based on Unicode to allow Java programs to be useful in as many parts of the world as possible. Internally, Java programs store characters as 16-bit Unicode characters. The benefits of using Unicode are currently difficult to realize, however, because most operating environments do not support Unicode. And those environments that do support Unicode generally do not include fonts that cover more than a small subset of the Unicode character set.

Since most operating environments do not support Unicode, Java uses a pre-processing phase to make sure that all of the characters of a program are in Unicode. This pre-processing comprises two steps:

  • Translate the program source into Unicode characters if it is in an encoding other than Unicode. Java defines escape sequences that allow all characters that can be represented in Unicode to be represented in other character encodings, such as ASCII or EBCDIC. The escape sequences are recognized by the compiler, even if the program is already represented in Unicode.

  • Divide the stream of Unicode characters into lines.

Conversion to Unicode

The first thing a Java compiler does is translate its input from the source character encoding (e.g., ASCII or EBCDIC) into Unicode. During the conversion process, Java translates escape sequences of the form \u followed by four hexadecimal digits into the Unicode characters indicated by the given hexadecimal values. These escape sequences let you represent Unicode characters in whatever character set you are using for your source code, even if it is not Unicode. For example, \u0000 is a way of representing the NUL character.

More formally, the compiler input is converted from a stream of EscapedSourceCharacters into a stream of Unicode characters. EscapedSourceCharacter is defined as:

[Graphic: Figure from the text]

HexDigit is either a Digit or one of the following letters: A, a, B, b, C, c, D, d, E, e, F, or f.

A Digit is one of the following characters: 0, 1, 2, 3, 4, 5, 6, 7, 8, or 9.

As you can see, the definition of EscapedSourceCharacter specifies that the `u' in the escape sequence can occur multiple times. Multiple occurrences have the same meaning as a single occurrence of `u'.

If the program source is already in Unicode, this conversion step is still performed in order to process these \u escapes.

The Java language specification recommends, but does not require, that the classes that come with Java use the \uxxxx escapes when called upon to display a character that would not otherwise be displayable.

Division of the Input Stream into Lines

The second step of pre-processing is responsible for recognizing sequences of characters that terminate lines. The character sequence that indicates the end of a line varies with the operating environment. By recognizing end-of-line character sequences during pre-processing, Java makes sure that subsequent compilation steps do not need to be concerned with multiple representations for the end of a line.

In this step, the lexical analyzer recognizes the combinations of carriage return (\u000D) and line feed (\u000A) characters that are in widespread use as end-of-line indicators:

[Graphic: Figure from the text]

As always, ambiguities in lexical rules are resolved by matching the longest possible sequence of characters. That means that the sequence of a carriage return character followed by a linefeed character is always recognized as a one-line terminator, never as two.

Previous Home Next
Notational Conventions Book Index Tokenization

Java in a Nutshell Java Language Reference Java AWT Java Fundamental Classes Exploring Java