Eric J. Roode

NAME

Iterator - A general-purpose iterator class.

VERSION

This documentation describes version 0.03 of Iterator.pm, October 10, 2005.

SYNOPSIS

 use Iterator;

 # Making your own iterators from scratch:
 $iterator = Iterator->new ( sub { code } );

 # Accessing an iterator's values in turn:
 $next_value = $iterator->value();

 # Is the iterator out of values?
 $boolean = $iterator->is_exhausted();
 $boolean = $iterator->isnt_exhausted();

 # Within {code}, above:
 Iterator::is_done();    # to signal end of sequence.

DESCRIPTION

This module is meant to be the definitive implementation of iterators, as popularized by Mark Jason Dominus's lectures and recent book (Higher Order Perl, Morgan Kauffman, 2005).

An "iterator" is an object, represented as a code block that generates the "next value" of a sequence, and generally implemented as a closure. When you need a value to operate on, you pull it from the iterator. If it depends on other iterators, it pulls values from them when it needs to. Iterators can be chained together (see Iterator::Util for functions that help you do just that), queueing up work to be done but not actually doing it until a value is needed at the front end of the chain. At that time, one data value is pulled through the chain.

Contrast this with ordinary array processing, where you load or compute all of the input values at once, then loop over them in memory. It's analogous to the difference between looping over a file one line at a time, and reading the entire file into an array of lines before operating on it.

Iterator.pm provides a class that simplifies creation and use of these iterator objects. Other Iterator:: modules (see "SEE ALSO") provide many general-purpose and special-purpose iterator functions.

Some iterators are infinite (that is, they generate infinite sequences), and some are finite. When the end of a finite sequence is reached, the iterator code block should throw an exception of the type Iterator::X::Am_Now_Exhausted; this is usually done via the "is_done" function.. This will signal the Iterator class to mark the object as exhausted. The "is_exhausted" method will then return true, and the "isnt_exhausted" method will return false. Any further calls to the "value" method will throw an exception of the type Iterator::X::Exhausted. See "DIAGNOSTICS".

Note that in many, many cases, you will not need to explicitly create an iterator; there are plenty of iterator generation and manipulation functions in the other associated modules. You can just plug them together like building blocks.

METHODS

new
 $iter = Iterator->new( sub { code } );

Creates a new iterator object. The code block that you provide will be invoked by the "value" method. The code block should have some way of maintaining state, so that it knows how to return the next value of the sequence each time it is called.

If the code is called after it has generated the last value in its sequence, it should throw an exception:

    Iterator::X::Am_Now_Exhausted->throw ();

This very commonly needs to be done, so there is a convenience function for it:

    Iterator::is_done ();
value
 $next_value = $iter->value ();

Returns the next value in the iterator's sequence. If value is called on an exhausted iterator, an Iterator::X::Exhausted exception is thrown.

Note that these iterators can only return scalar values. If you need your iterator to return a list or hash, it will have to return an arrayref or hashref.

is_exhausted
 $bool = $iter->is_exhausted ();

Returns true if the iterator is exhausted. In this state, any call to the iterator's "value" method will throw an exception.

isnt_exhausted
 $bool = $iter->isnt_exhausted ();

Returns true if the iterator is not yet exhausted.

FUNCTION

is_done
 Iterator::is_done();

You call this function after your iterator code has generated its last value. See "TUTORIAL". This is simply a convenience wrapper for

 Iterator::X::Am_Now_Exhausted->throw();

THINKING IN ITERATORS

Typically, when people approach a problem that involves manipulating a bunch of data, their first thought is to load it all into memory, into an array, and work with it in-place. If you're only dealing with one element at a time, this approach usually wastes memory needlessly.

For example, one might get a list of files to operate on, and loop over it:

    my @files = fetch_file_list(....);
    foreach my $file (@files)
        ...
If C<fetch_file_list> were modified to return an iterator instead of
an array, the same code could look like this:

    my $file_iterator = fetch_file_list(...)
    while ($file_iterator->isnt_exhausted)
        ...

The advantage here is that the whole list does not take up memory while each individual element is being worked on. For a list of files, that's probably not a lot of overhead. For the contents of a file, on the other hand, it could be huge.

If a function requires a list of items as its input, the overhead is tripled:

    sub myfunc
    {
        my @things = @_;
        ...

Now in addition to the array in the calling code, Perl must copy that array to @_, and then copy it again to @things. If you need to massage the input from somewhere, it gets even worse:

    my @data = get_things_from_somewhere();
    my @filtered_data = grep {code} @data;
    my @transformed_data = map {code} @filtered_data;
    myfunc (@transformed_data);

If myfunc is rewritten to use an Iterator instead of an array, things become much simpler:

    my $data = ilist (get_things_from_somewhere());
    $filtered_data = igrep {code} $data;
    $transformed_data = imap {code} $filtered_data;
    myfunc ($transformed_data);

(This example assumes that the get_things_from_somewhere function cannot be modified to return an Iterator. If it can, so much the better!) Now the original list is still in memory, inside the $data Iterator, but everwhere else, there is only one data element in memory at a time.

Another advantage of Iterators is that they're homogeneous. This is useful for uncoupling library code from application code. Suppose you have a library function that grabs data from a filehandle:

    sub my_lib_func
    {
        my $fh = shift;
        ...

If you need my_lib_func to get its data from a different source, you must either modify it, or make a new copy of it that gets its input differently, or you must jump through hoops to make the new input stream look like a Perl filehandle.

On the other hand, if my_lib_func accepts an iterator, then you can pass it data from a filehandle:

    my $data = ifile "my_input.txt";
    $result = my_lib_func($data);

Or a database handle:

    my $data = imap {$_->{IMPORTANT_COLUMN}}
               idb_rows($dbh, 'select IMPORTANT_COLUMN from foo');
    $result = my_lib_func($data);

If you later decide you need to transform the data, or process only every 10th data row, or whatever:

    $result = my_lib_func(imap {magic($_)} $data);
    $result = my_lib_func(inth 10, $data);

The library function doesn't care. All it needs is an iterator.

Chapter 4 of Dominus's book (See "SEE ALSO") covers this topic in some detail.

Word of Warning

When you use an iterator in separate parts of your program, or as an argument to the various iterator functions, you do not get a copy of the iterator's stream of values.

In other words, if you grab a value from an iterator, then some other part of the program grabs a value from the same iterator, you will be getting different values.

This can be confusing if you're not expecting it. For example:

    my $it_one = Iterator->new ({something});
    my $it_two = some_iterator_transformation $it_one;
    my $value  = $it_two->value();
    my $whoops = $it_one->value;

Here, some_iterator_transformation takes an iterator as an argument, and returns an iterator as a result. When a value is fetched from $it_two, it internally grabs a value from $it_one (and presumably transforms it somehow). If you then grab a value from $it_one, you'll get its second value (or third, or whatever, depending on how many values $it_two grabbed), not the first.

TUTORIAL

Let's create a date iterator. It'll take a DateTime object as a starting date, and return successive days -- that is, it'll add 1 day each iteration. It would be used as follows:

 use DateTime;

 $iter = (...something...);
 $day1 = $iter->value;           # Initial date
 $day2 = $iter->value;           # One day later
 $day3 = $iter->value;           # Two days later

The easiest way to create such an iterator is by using a closure. If you're not familiar with the concept, it's fairly simple: In Perl, the code within an anonymous block has access to all the lexical variables that were in scope at the time the block was created. After the program then leaves that lexical scope, those lexical variables remain accessible by that code block for as long as it exists.

This makes it very easy to create iterators that maintain their own state. Here we'll create a lexical scope by using a pair of braces:

 my $iter;
 {
    my $dt = DateTime->now();
    $iter = Iterator->new( sub
    {
        my $return_value = $dt->clone;
        $dt->add(days => 1);
        return $return_value;
    });
}

Because $dt is lexically scoped to the outermost block, it is not addressable from any code elsewhere in the program. But the anonymous block within the "new" method's parentheses can see $dt. So $dt does not get garbage-collected as long as $iter contains a reference to it.

The code within the anonymous block is simple. A copy of the current $dt is made, one day is added to $dt, and the copy is returned.

You'll probably want to encapsulate the above block in a subroutine, so that you could call it from anywhere in your program:

 sub date_iterator
 {
     my $dt = DateTime->now();
     return Iterator->new( sub
     {
         my $return_value = $dt->clone;
         $dt->add(days => 1);
         return $return_value;
     });
 }

If you look at the source code in Iterator::Util, you'll see that just about all of the functions that create iterators look very similar to the above date_iterator function.

Of course, you'd probably want to be able to pass arguments to date_iterator, say a starting date, maybe an increment other than "1 day". But the basic idea is the same.

The above date iterator is an infinite (well, unbounded) iterator. Let's look at how to indicate that your iterator has reached the end of its sequence of values. Let's write a scaled-down version of irange from the Iterator::Util module -- one that takes a start value and an end value and always increments by 1.

 sub irange_limited
 {
     my ($start, $end) = @_;

     return Iterator->new (sub
     {
         Iterator::is_done
             if $start > $end;

         return $start++;
     });
 }

The iterator itself is very simple (this sort of thing gets to be easy once you get the hang of it). The new element here is the signalling that the sequence has ended, and the iterator's work is done. "is_done" is how your code indicates this to the Iterator object.

You may also want to throw an exception if the user specified bad input parameters. There are a couple ways you can do this.

     ...
     die "Too few parameters to irange_limited"  if @_ < 2;
     die "Too many parameters to irange_limited" if @_ > 2;
     my ($start, $end) = @_;
     ...

This is the simplest way; you just use die (or croak). You may choose to throw an Iterator parameter error, though; this will make your function work more like one of Iterator.pm's built in functions:

     ...
     Iterator::X::Parameter_Error->throw(
         "Too few parameters to irange_limited")
         if @_ < 2;
     Iterator::X::Parameter_Error->throw(
         "Too many parameters to irange_limited")
         if @_ > 2;
     my ($start, $end) = @_;
     ...

EXPORTS

No symbols are exported to the caller's namespace.

DIAGNOSTICS

Iterator uses Exception::Class objects for throwing exceptions. If you're not familiar with Exception::Class, don't worry; these exception objects work just like $@ does with die and croak, but they are easier to work with if you are trapping errors.

All exceptions thrown by Iterator have a base class of Iterator::X. You can trap errors with an eval block:

 eval { $foo = $iterator->value(); };

and then check for errors as follows:

 if (Iterator::X->caught())  {...

You can look for more specific errors by looking at a more specific class:

 if (Iterator::X::Exhausted->caught())  {...

Some exceptions may provide further information, which may be useful for your exception handling:

 if (my $ex = Iterator::X::User_Code_Error->caught())
 {
     my $exception = $ex->eval_error();
     ...

If you choose not to (or cannot) handle a particular type of exception (for example, there's not much to be done about a parameter error), you should rethrow the error:

 if (my $ex = Iterator::X->caught())
 {
     if ($ex->isa('Iterator::X::Something_Useful'))
     {
         ...
     }
     else
     {
         $ex->rethrow();
     }
 }
  • Parameter Errors

    Class: Iterator::X::Parameter_Error

    You called an Iterator method with one or more bad parameters. Since this is almost certainly a coding error, there is probably not much use in handling this sort of exception.

    As a string, this exception provides a human-readable message about what the problem was.

  • Exhausted Iterators

    Class: Iterator::X::Exhausted

    You called "value" on an iterator that is exhausted; that is, there are no more values in the sequence to return.

    As a string, this exception is "Iterator is exhausted."

  • End of Sequence

    Class: Iterator::X::Am_Now_Exhausted

    This exception is not thrown directly by any Iterator.pm methods, but is to be thrown by iterator sequence generation code; that is, the code that you pass to the "new" constructor. Your code won't catch an Am_Now_Exhausted exception, because the Iterator object will catch it internally and set its "is_exhausted" flag.

    The simplest way to throw this exception is to use the "is_done" function:

     Iterator::is_done() if $something;
  • User Code Exceptions

    Class: Iterator::X::User_Code_Error

    This exception is thrown when the sequence generation code throws any sort of error besides Am_Now_Exhausted. This could be because your code explicitly threw an error (that is, died), or because it otherwise encountered an exception (any runtime error).

    This exception has one method, eval_error, which returns the original $@ that was trapped by the Iterator object. This may be a string or an object, depending on how die was invoked.

    As a string, this exception evaluates to the stringified $@.

  • I/O Errors

    Class: Iterator::X::IO_Error

    This exception is thrown when any sort of I/O error occurs; this only happens with the filesystem iterators.

    This exception has one method, os_error, which returns the original $! that was trapped by the Iterator object.

    As a string, this exception provides some human-readable information along with $!.

  • Internal Errors

    Class: Iterator::X::Internal_Error

    Something happened that I thought couldn't possibly happen. I would appreciate it if you could send me an email message detailing the circumstances of the error.

REQUIREMENTS

Requires the following additional module:

Exception::Class, v1.21 or later.

SEE ALSO

THANKS

Much thanks to Will Coleda and Paul Lalli (and the RPI lily crowd in general) for suggestions for the pre-release version.

AUTHOR / COPYRIGHT

Eric J. Roode, roode@cpan.org

Copyright (c) 2005 by Eric J. Roode. All Rights Reserved. This module is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

To avoid my spam filter, please include "Perl", "module", or this module's name in the message's subject line, and/or GPG-sign your message.




Hosting generously
sponsored by Bytemark