File Coverage

blib/lib/CBOR/Free.pm

Criterion	Covered	Total	%
statement	26	29	89.6
branch			n/a
condition			n/a
subroutine	11	12	91.6
pod	1	3	33.3
total	38	44	86.3

package CBOR::Free;

use strict;
use warnings;

use CBOR::Free::X;
use CBOR::Free::Tagged;

our ($VERSION);

use XSLoader ();

BEGIN {
    $VERSION = '0.32_02';
    XSLoader::load();
}

#----------------------------------------------------------------------

=encoding utf-8

=head1 NAME

CBOR::Free - Fast CBOR for everyone

=head1 SYNOPSIS

    $cbor = CBOR::Free::encode( $some_data_structure );

    $thing = CBOR::Free::decode( $cbor )

    my $tagged = CBOR::Free::tag( 1, '2019-01-02T00:01:02Z' );

Also see L for an object-oriented interface
to the decoder.

=head1 DESCRIPTION

=begin html



=end html

This library implements L
via XS under a license that permits commercial usage with no “strings
attached”.

=head1 STATUS

This distribution is an experimental effort. Its interface is still
subject to change. If you decide to use CBOR::Free in your project,
please always check the changelog before upgrading.

=head1 FUNCTIONS

=head2 $cbor = encode( $DATA, %OPTS )

Encodes a data structure or non-reference scalar to CBOR.
The encoder recognizes and encodes integers, floats, byte and character
strings, array and hash references, L instances,
L booleans, and undef (encoded as null).

The encoder currently does not handle any other blessed references.

%OPTS may be:

=over

=item * C - A boolean that makes the encoder output
CBOR in L.

=item * C - Decides the logic to use for
CBOR encoding of strings and hash keys. (The word “string”
in the below descriptions applies equally to hash keys.)

Takes one of:

=over

=item * C: The default mode of operation. If the string’s internal
UTF8 flag is set, it will become a CBOR text string; otherwise, it will be
CBOR binary. This is good for IPC with other Perl code but isn’t a very
friendly default for working with other languages that probably expect more
reliably-typed strings.

This is (currently) the only way to output text and binary strings in a
single CBOR document. Unfortunately, because Perl itself doesn’t reliably
distinguish between text and binary strings, neither can CBOR::Free. If you
want to try, though:

=over

=item * Be sure to use character-decoding logic that always
sets the string’s UTF8 flag, even if the input is plain ASCII.
(As of this writing, L and L work this way.)

=item * Whatever consumes your Perl-sourced CBOR should probably accept
“mis-typed” strings.

=back

=item * C: Treats all strings as unencoded characters.
All CBOR strings will be text.

This is probably what you want if you
follow the receive-decode-process-encode-output workflow that
L recommends (which you might be doing via C)
B if you intend for your CBOR to contain exclusively text.

Think of this option as: “All my strings are decoded.”

(Perl internals note: if !SvUTF8, the CBOR will be the UTF8-upgraded
version.)

=item * C: Treats all strings as octets of UTF-8.
Wide characters (i.e., code points above 255) are thus invalid input.
All CBOR strings will be text.

This is probably what you want if you forgo character decoding (and encoding),
treating all input as octets, B you still intend for your CBOR to
contain exclusively text.

Think of this option as: “I’ve encoded all my strings as UTF-8.”

(Perl internals note: if SvUTF8, the CBOR will be the downgraded version.)

=item * C: Like C, but outputs CBOR binary
instead of text.

This is probably what you want if your application is “all binary,
all the time”.

Think of this option as: “Just the bytes, ma’am.”

=back

=item * C - A boolean that makes the encoder encode
multi-referenced values via L. This allows encoding of shared
and circular references. It also incurs a performance penalty.

(Take care that any circular references in your application don’t cause
memory leaks!)

=item * C - A boolean that makes the encoder accept
scalar references
(rather than reject them) and encode them via
L.
Most languages don’t use references as Perl does, so this option seems of
little use outside all-Perl IPC contexts; it is arguably more useful, then,
for general use to have the encoder reject data structures that most other
languages cannot represent.

=back

Notes on mapping Perl to CBOR:

=over

=item * The internal state of a Perl scalar (e.g., whether it’s an
integer, float, string, etc.) determines its CBOR encoding.

=item * Perl doesn’t currently provide reliable binary/character string types.
The various C options (described above) provide ways to
deal with this problem.

=item * The above applies also to strings vs. numbers: whatever consumes
your Perl-sourced CBOR B account for the prospect of numbers that
are in CBOR as strings, or vice-versa.

=item * Perl hash keys are serialized as strings, either binary or text
(according to the C).

=item * L booleans are encoded as CBOR booleans.
Perl undef is encoded as CBOR null. (NB: No Perl value encodes as CBOR
undefined.)

=item * Scalar references (including references to other references) are
unhandled by default, which makes them trigger an exception. You can
optionally tell CBOR::Free to encode them via the C flag.

=item * Via the optional C flag, circular and shared
references may be preserved. Without this flag, circular references cause an
exception, and other shared references are not preserved.

=item * Instances of L are encoded as tagged values.

=back

An error is thrown on excess recursion or an unrecognized object.

=head2 $data = decode( $CBOR )

Decodes a data structure from CBOR. Errors are thrown to indicate
invalid CBOR. A warning is thrown if $CBOR is longer than is needed
for $data.

Notes on mapping CBOR to Perl:

=over

=item * C decodes CBOR text strings as UTF-8-decoded Perl strings.
CBOR binary strings become undecoded Perl strings.

(See L and L for more
character-decoding options.)

Notes:

=over

=item * Invalid UTF-8 in a CBOR text string is usually considered
invalid input and will thus prompt a thrown exception. (See
L and L if you want
to tolerate invalid UTF-8.)

=item * You can reliably use C to determine if a given Perl
string came from CBOR text or binary, but B if you test the scalar as
it appears in the newly-decoded data structure itself. Generally Perl code
should avoid C, but with CBOR::Free-created strings this limited
use case is legitimate and potentially gainful.

=back

=item * The only map keys that C accepts are integers and strings.
An exception is thrown if the decoder finds anything else as a map key.
Note that, because Perl does not distinguish between binary and text strings,
if two keys of the same map contain the same bytes, Perl will consider these
a duplicate key and prefer the latter.

=item * CBOR booleans become the corresponding L values.
Both CBOR null and undefined become Perl undef.

=item * L is interpreted as a scalar reference. This behavior is always
active; unlike with the encoder, there is no need to enable it manually.

=item * C mode complements the same flag
given to the encoder.

=item * This function does not interpret any other tags. If you need to
decode other tags, look at L. Any unhandled tags that
this function sees prompt a warning but are otherwise ignored.

=back

=head2 $obj = tag( $NUMBER, $DATA )

Tags an item for encoding so that its CBOR encoding will preserve the
tag number. (Include $obj, not $DATA, in the data structure that
C receives.)

=head1 BOOLEANS

C and C are defined as
convenience aliases for the equivalent L functions.
(Note that there are no equivalent scalar aliases.)

=head1 FRACTIONAL (FLOATING-POINT) NUMBERS

Floating-point numbers are encoded in CBOR as IEEE 754 half-, single-,
or double-precision. If your Perl is compiled to use anything besides
IEEE 754 double-precision to represent floating-point values (e.g.,
“long double” or “quadmath” compilation options), you may see rounding
errors when converting to/from CBOR. If that’s a problem for you, append
an empty string to your floating-point numbers, which will cause CBOR::Free
to encode them as strings.

=head1 INTEGER LIMITS

CBOR handles up to 64-bit positive and negative integers. Most Perls
nowadays can handle 64-bit integers, but if yours can’t then you’ll
get an exception whenever trying to parse an integer that can’t be
represented with 32 bits. This means:

=over

=item * Anything greater than 0xffff_ffff (4,294,967,295)

=item * Anything less than -0x8000_0000 (2,147,483,648)

=back

Note that even 64-bit Perls can’t parse negatives that are less than
-0x8000_0000_0000_0000 (-9,223,372,036,854,775,808); these also prompt an
exception since Perl can’t handle them. (It would be possible to load
L to handle these; if that’s desirable for you,
file a feature request.)

=head1 ERROR HANDLING

Most errors are represented via instances of subclasses of
L, which subclasses L.

=head1 SPEED

CBOR::Free is pretty snappy. I find that it keeps pace with or
surpasses L, L, L, L,
and L.

It’s also quite light. Its only “heavy” dependency is
L, which is only loaded when you actually need it.
This keeps memory usage low for when, e.g., you’re using CBOR for
IPC between Perl processes and have no need for true booleans.

=head1 AUTHOR

L (FELIPE)

=head1 LICENSE

This code is licensed under the same license as Perl itself.

=head1 SEE ALSO

L is a pure-Perl CBOR library.

L is an older CBOR module on CPAN. It’s got more bells and
whistles, so check it out if CBOR::Free lacks a feature you’d like.
Note that L
onward|http://blog.schmorp.de/2015-06-06-stableperl-faq.html>, though,
and its GPL license limits its usefulness in
commercial L
applications.

=cut

#----------------------------------------------------------------------

sub true {
    require Types::Serialiser;
    *true = *Types::Serialiser::true;
    goto &true;
}

sub false {
    require Types::Serialiser;
    *false = *Types::Serialiser::false;
    goto &false;
}

sub tag {
    return CBOR::Free::Tagged->new(@_);
}

#----------------------------------------------------------------------

sub _die_recursion {
    die CBOR::Free::X->create( 'Recursion', _MAX_RECURSION());
}

sub _die {
    my ($subclass, @args) = @_;

    die CBOR::Free::X->create($subclass, @args);
}

sub _warn_decode_leftover {
    my ($count) = @_;

    warn "CBOR buffer contained $count excess bytes";
}

1;