package Statistics::R::IO;

ABSTRACT: Perl interface to serialized R data

use 5.010; use strict; use warnings FATAL => 'all';

use Exporter 'import';

our @EXPORT = qw( ); our @EXPORT_OK = qw( readRDS readRData evalRserve );

our %EXPORT_TAGS = ( all => [ @EXPORT_OK ], );

use Statistics::R::IO::REXPFactory; use Statistics::R::IO::Rserve; use IO::Uncompress::Gunzip (); use IO::Uncompress::Bunzip2 (); use IO::Socket::INET (); use Carp;

sub readRDS { open (my $f, shift) or croak $!; binmode $f; my ($data, $rc) = ''; while ($rc = read($f, $data, 8192, length $data)) {} croak $! unless defined $rc; if (substr($data, 0, 2) eq "\x1f\x8b") {

gzip-compressed file

    seek($f, 0, 0);
    IO::Uncompress::Gunzip::gunzip $f, \$data;
}
elsif (substr($data, 0, 3) eq 'BZh') {
    ## bzip2-compressed file
    seek($f, 0, 0);
    IO::Uncompress::Bunzip2::bunzip2 $f, \$data;
}
elsif (substr($data, 0, 6) eq "\xfd7zXZ\0") {
    croak "xz-compressed RDS files are not supported";
}

my ($value, $state) = @{Statistics::R::IO::REXPFactory::unserialize($data)};
croak 'Could not parse RDS file' unless $state;
croak 'Unread data remaining in the RDS file' unless $state->eof;
$value

}

sub readRData { open (my $f, shift) or croak $!; binmode $f; my ($data, $rc) = ''; while ($rc = read($f, $data, 8192, length $data)) {} croak $! unless defined $rc; if (substr($data, 0, 2) eq "\x1f\x8b") {

gzip-compressed file

    seek($f, 0, 0);
    IO::Uncompress::Gunzip::gunzip $f, \$data;
}
elsif (substr($data, 0, 3) eq 'BZh') {
    ## bzip2-compressed file
    seek($f, 0, 0);
    IO::Uncompress::Bunzip2::bunzip2 $f, \$data;
}
elsif (substr($data, 0, 6) eq "\xfd7zXZ\0") {
    croak "xz-compressed RData files are not supported";
}

if (substr($data, 0, 5) ne "RDX2\n") {
    croak 'File does not start with the RData magic number: ' .
        unpack('H*', substr($data, 0, 5));
}

my ($value, $state) = @{Statistics::R::IO::REXPFactory::unserialize(substr($data, 5))};
croak 'Could not parse RData file' unless $state;
croak 'Unread data remaining in the RData file' unless $state->eof;
Statistics::R::IO::REXPFactory::tagged_pairlist_to_rexp_hash $value;

}

sub evalRserve { my ($rexp, $server) = (shift, shift);

my $rserve = Statistics::R::IO::Rserve->new($server // {});

$rserve->eval($rexp)

}

1; # End of Statistics::R::IO

END

=head1 SYNOPSIS

use Statistics::R::IO;

my $var = Statistics::R::IO::readRDS('file.rds');
print $var->to_pl;

my %r_workspace = Statistics::R::IO::readRData('.RData');
while (my ($var_name, $value) = each %r_workspace) {
    print $var_name, $value;
}

my $pi = Statistics::R::IO::evalRserve('pi');
print $pi->to_pl;

=head1 DESCRIPTION

This module is a pure-Perl implementation for reading native data files produced by the L<R statistical computing environment|http://www.r-project.org>)

It provides routines for reading files in the two primary file formats used in R for serializing native objects:

=over

=item RDS

RDS files store a serialization of a single R object (and, if the object contains references to other objects, such as environments, all the referenced objects as well). These files are created in R using the C function and are typically named with the C<.rds> file extension.

=item RData

RData files store a serialization of a collection of I objects, typically a workspace. These files are created in R using the C function and are typically named with the C<.RData> file extension. (Contents of the R workspace can also be saved automatically on exit to the file named F<.RData>, which is by default automatically read in on startup.)

=back

As of version 0.04, the module can also evaluate R code on a remote host that runs the L<Rserve|http://www.rforge.net/Rserve/> binary R server. This allows Perl programs to access all facilities of R without the need to have a local install of R or link to an R library.

See L for invocation and usage information on individual subroutines, and the L<R Internals manual|http://cran.r-project.org/doc/manuals/R-ints.html> for the specification of the file formats.

=head1 EXPORT

Nothing by default. Optionally, subroutines C, C, and C, or C<:all> for all three.

=head1 SUBROUTINES

=over 4

=item readRDS EXPR

Reads a file in RDS format whose filename is given by EXPR and returns a L object.

=item readRData EXPR

Reads a file in RData format whose filename is given by EXPR and returns a hash whose keys are the names of objects stored in the file with corresponding values as L instances.

=item evalRserve REXPR [ HOSTNAME [, PORT] | HANDLE]

Evaluates an R expression, given as text string in REXPR, on an L<Rserve|http://www.rforge.net/Rserve/> server and returns its result as a L object.

The server location can be specified either by its host name and (optionally) port or by a connected instance of L. The caller passing the HANDLE is responsible for reading (and checking) the server ID that is returned in the first 32-byte response when the connection was established. This allows opening the connection once and reusing it in multiple calls to 'evalRserve'.

If only REXPR is given, the function assumes that the server runs on the localhost. If PORT is not specified, it defaults to the standard Rserve port, 6311.

The function will close the connection to the Rserve host if it has opened it itself, but not if the connection was passed as a HANDLE.

=back

=head1 DEPENDENCIES

Requires perl 5.010 or newer.

=head2 Core modules

=over

=item * strict

=item * warnings

=item * overload

=item * Carp

=item * Exporter

=item * Module::Build

=item * Scalar::Util

=item * Test::More

=back

=head2 Additional CPAN modules

=over

=item * Class::Tiny

=item * Class::Tiny::Antlers

=item * Class::Method::Modifiers

=item * namespace::clean

=item * Test::Fatal

=back

=head1 BUGS AND LIMITATIONS

The module currently handles the 'version 2' serialization format, used since R 1.4.0 (released in December 2001). Only XDR and native-order binary is implemented, and since the R documentation describes the ASCII save format as "now mainly of historical interest", this is unlikely to change soon. No check is performed that a file stored in native-order binary was created on a platform that used the same order, and it is up to the caller to ensure compatibility. (Given that the default save format is XDR, and the prevalence of Intel platforms, this is unlikely to be a problem for either publicly-distributed or internal data files.)

Data files compressed with 'gzip' and 'bzip2' are supported, but not 'xz' ones. Again, given the R defaults ('gzip') and the fact that C is not production-ready, this is unlikely to change soon.

There are some R types that are not (yet) implemented, although all typical "user-facing" types -- such as vectors, lists, and environments -- are. The remaining R types will be implemented as-needed; in other words, if you come across one that you need to read a particular file, please send me the type (the id will included in the "unimplemented SEXPTYPE" error message) and, if possible, how it was generated.

There are no known bugs in this module. Please report any bugs or feature requests to C, or through the web interface at Lhttp://rt.cpan.org/NoAuth/ReportBug.html?Queue=Statistics-R-IO. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.

=head1 SUPPORT

You can find documentation for this module with the perldoc command.