Take me over?
NAME
Data::Stream::Bulk - N at a time iteration API
SYNOPSIS
# get a bulk stream from somewere
my $s = Data::Stream::Bulk::Foo->new( ... );
# can be used like this:
until ( $s->is_done ) {
foreach my $item ( $s->items ) {
process($item);
}
}
# or like this:
while( my $block = $s->next ) {
foreach my $item ( @$block ) {
process($item);
}
}
DESCRIPTION
This module tries to find middle ground between one at a time and all at once processing of data sets.
The purpose of this module is to avoid the overhead of implementing an iterative api when this isn't necessary, without breaking forward compatibility in case that becomes necessary later on.
The API optimizes for when a data set typically fits in memory and is returned as an array, but the consumer cannot assume that the data set is bounded.
The API is destructive in order to minimize the chance that resultsets are leaked due to improper usage.
API
Required Methods
The API requires two methods to be implemented:
- is_done
-
Should return true if the stream is exhausted.
As long as this method returns a false value (not done)
next
could potentially return another block. - next
-
Returns the next block.
Note that
next
is not guaranteed to return an array reference, even ifis_done
returned false prior to calling it.
Convenience Methods
- items
-
This method calls
next
and dereferences the result if there are pending items. - all
-
Force evaluation of the entire resultset.
Note that for large data sets this might cause swap thrashing of various other undesired effects. Use with caution.
- cat @streams
-
Concatenates this stream with @streams, returning a single stream.
- list_cat @tail
-
Returns a possibly cleaned up list of streams.
Used by
cat
.Overridden by Data::Stream::Bulk::Array, Data::Stream::Bulk::Cat and Data::Stream::Bulk::Nil to implement some simple short circuiting.
- filter $filter
-
Applies a per-block block filter to the stream.
Returns a possibly new stream with the filtering layered.
$filter
is invoked once per block and should return an array reference to the filtered block. - loaded
-
Should be overridden to return true if all the items are already realized (e.g. in the case of Data::Stream::Bulk::Array).
Returns false by default.
When true calling
all
is supposed to be safe (memory usage should be in the same order of magnitude as stream's own usage).This is typically useful when tranforming an array is easier than transorming a stream (e.g. optional duplicate filtering).
CLASSES
- Data::Stream::Bulk::Array
-
This class is not a stream at all, but just one block. When the data set easily fits in memory this class can be used, while retaining forward compatibility with larger data sets.
- Data::Stream::Bulk::Callback
-
Callback driven iteration.
- Data::Stream::Bulk::DBI
-
Bulk fetching of data from DBI statement handles.
- Data::Stream::Bulk::DBIC
-
DBIx::Class::ResultSet iteration.
- Data::Stream::Bulk::Nil
-
An empty result set.
- Data::Stream::Bulk::Cat
-
A concatenation of several streams.
- Data::Stream::Bulk::Filter
-
A filter wrapping a stream.
SEE ALSO
HOP::Stream, Iterator, Class::Iterator etc for one by one iteration
http://en.wikipedia.org/wiki/MapReduce, LISP, and all that other kool aid
TODO
- Sorted streams
-
Add a hint for sorted streams (like
loaded
but as an attribute in the base role).Introduce a
merge
operation for merging of sorted streams.Optimize
unique
to make use of sorting hints for constant space uniquing. - More utility functions
-
To assist in proccessing and creating streams.
- Coercion tables
VERSION CONTROL
This module is maintained using git. You can get the latest version from http://github.com/nothingmuch/data-stream-bulk/.
AUTHOR
Yuval Kogman <nothingmuch@woobling.org>
COPYRIGHT
Copyright (c) 2008 Yuval Kogman. All rights reserved
This program is free software; you can redistribute
it and/or modify it under the same terms as Perl itself.