Friday, October 08, 2004

Taming Tiger: Formatted output - Let there be printf

The original plan for the 1.4 release J2SE included support for formatted output. Probably due to time constraints and the fact that the feature wasn't a release driver, the capabilities were left out of the release. Now with Tiger, there is built in support for printing with format strings.

For those who grew up with Java programming and never touched C or those who haven't lived in a C world for some time, format strings are those funky text strings that specify output characteristics for a bunch of variables. Instead of just concatenating strings together with the plus sign (as in firstName + " " + lastName), you provide a single string to describe the output and provide the arguments to fill the placeholders in that string at the end of the method call: String s = String.format("%1$s %2$s", firstName, lastName).

Formatter class
First, let's look at the new java.util.Formatter class. You probably won't use this class directly much, but it provides the guts for the formatting you'll be doing. In the Javadoc for this class, you'll find a table describing the supported formatting options. These options range from something like %7.4f for specifying the precision and width of a floating point number to %tT for formatting a time to %3$s for formatting the third argument.

Using Formatter to format output involves two steps: creating an Appendable object to store the output and using the format() method to put formatted content into that object. Here's a list of implementers for the Appendable interface:

BufferedWriter
CharArrayWriter
CharBuffer
FileWriter
FilterWriter
LogStream
OutputStreamWriter
PipedWriter
PrintStream
PrintWriter
StringBuffer
StringBuilder
StringWriter
Writer
An object implementing this interface can be used as the destination when using a Formatter class by passing the object into the Formatter constructor. Most of these classes should look familiar, except for the StringBuilder class. StringBuilder is nearly identical to the StringBuffer class, with one big exception: It isn't thread safe. If you know you are going to build up a string in a single thread, use StringBuilder. If the building can cross thread bounds, use StringBuffer. Listing 1 shows how you'd typically start using Formatter:

Listing 1. Typical formatter usage

StringBuilder sb = new StringBuilder();
Formatter formatter = new Formatter(sb, Locale.US);



After creating a Formatter class, you call its format() method with format strings and arguments. If you need to use a different Locale than that sent into the constructor for part of the formatted output, you can also pass in a Locale object to the format() method. Listing 2 shows the two varieties of format():

Listing 2. format() methods of Formatter

public Formatter format(String format,
Object... args)
public Formatter format(Locale l,
String format,
Object... args)



If you want to get the value of Pi to 10 digits of precision, the code in Listing 3 will put that value into the StringBuilder and print the output. Printing the formatter object will display the content of the Appendable object.

Listing 3. Demonstrating a Formatter

import java.util.Locale;
import java.util.Formatter;

public class Build {
public static void main(String args[]) {
StringBuilder sb = new StringBuilder();
Formatter formatter = new Formatter(sb, Locale.US);
formatter.format("PI = %12.10f", Math.PI);
System.out.println(formatter);
}
}



Don't forget to compile with the -source 1.5 option or the compiler won't recognize the variable argument list. Because formatting output and sending it to the console are common tasks, there are conveniences available to this behavior. We'll look at those next.

PrintStream support
The PrintStream class contains the definition of the common System.out and System.err objects for writing to standard output and standard error, respectively. Introduced in Tiger are two new constructors (for going straight to a file) and six methods for formatting support (three sets of pairs). The first pair is different versions of the append() method. This pair implements the new java.lang.Appendable interface. You would typically not call these methods directly. The ones you would call directly are format() and printf(), where the printf() versions are just convenience wrappers for the format() versions, as shown in Listing 4:

Listing 4. PrintStream.format methods

public PrintStream format(String format,
Object... args)
public PrintStream format(Locale l,
String format,
Object... args)



Keep in mind the new variable argument support, which is designated by the ... shown in Listing 4 above.

Listing 5 demonstrates the use of the format() method of PrintStream to print today's date:

Listing 5. Example PrintStream.format usage
/
import java.util.Calendar;

public class Now {
public static void main(String args[]) {
System.out.format(
"Today is %1$tB %1$te, %1$tY.",
Calendar.getInstance()
);
}
}



The output from running this program is Today is April 2, 2004., although the actual output depends on the date you run the program. The %1$tB formatting string in the code above tells the program to use the first argument and print out the full month name for the date object. The %1$te formatting string means to display the day of the month and the %1$tY formatting string is for the four-digit year. Other options for printing dates and times are shown in the Javadoc for the Formatter object.

String support
The String class has two new static format() methods that work similarly to their printf() equivalents. Send a format string and arguments (with a possible Locale) and use what is specified in the format string to convert the arguments. In the case of the String version of the method, the results are sent back as a String object, instead of going through the stream. These methods aren't overly spectacular, but they allow you to avoid using the Formatter object directly and creating an intermediate StringBuilder.

Formatting arbitrary objects
Everything you've seen so far describes how to use the new formatting capabilities to format existing objects and primitive types. If you want to provide support for using your own objects with Formatter, that's where the Formattable interface comes into play. By implementing the single formatTo() method shown in Listing 6 in your own class, you can use your own classes with format strings:

Listing 6. Formattable interface

void formatTo(Formatter formatter,
int flags,
Integer width,
Integer precision)



Listing 7 demonstrates the use of the Formattable interface by providing a simple class with a name property. That name is displayed in the output, with support for controlling the width of the output and the justification.

Listing 7. Example formattable usage

import java.util.Locale;
import java.util.Formatter;
import java.util.Formattable;

public class MyObject implements Formattable {
String name;
public MyObject(String name) {
this.name = name;
}
public void formatTo(
Formatter fmt,
int f,
Integer width,
Integer precision) {

StringBuilder sb = new StringBuilder();
if (precision == null) {
// no max width
sb.append(name);
} else if (name.length() < precision) {
sb.append(name);
} else {
sb.append(name.substring(0, precision - 1)).append('*');
}

// apply width and justification
if ((width != null) && (sb.length() < width)) {
for (int i = 0, n=sb.length(); i < width - n; i++) {
if ((f & Formattable.LEFT_JUSTIFY) == Formattable.LEFT_JUSTIFY) {
sb.append(' ');
} else {
sb.insert(0, ' ');
}
}
}
fmt.format(sb.toString());
}

public static void main(String args[]) {
MyObject my1 = new MyObject("John");
MyObject my2 = new MyObject("Really Long Name");
// First / Using toString()
System.out.println("First Object : " + my1);
// Second / Using Formatter
System.out.format("First Object : '%s'\n", my1);
// Second / Using Formatter
System.out.format("Second Object: '%s'\n", my2);
// Second / Using Formatter with width
System.out.format("Second Object: '%10.5s'\n", my2);
// Second / Using Formatter with width and left justification
System.out.format("Second Object: '%-10.5s'\n", my2);
}
}



Running this program produces the output in Listing 8. The first two lines demonstrate the difference between using toString and Formatter. The last three show options for width and justification control.

Listing 9. Example formattable output

First Object : MyObject@10b62c9
First Object : 'John'
Second Object: 'Really Long Name'
Second Object: ' Real*'
Second Object: 'Real* '



Conclusion
Getting a grasp of all the formatting options available with Formatter will take a little time, unless you are familiar with them from the C world. Some minor differences exist, but for the most part the behaviors are very similar. One key difference with the Java platform is that when the formatting string is invalid, an exception will be thrown.

Be sure to take a long look at the formatting strings available, as shown in the Formatter class Javadoc. When creating your own custom classes, not only should you provide a toString() implementation, but implementing the Formattable interface will typically be beneficial.

No comments: