Author image Ben Bullock
and 4 contributors

NAME

Image::PNG::Libpng - Perl interface to the C library "libpng".

SYNOPSIS

Libpng-like interface:

    use Image::PNG::Libpng ':all';
    my $png = create_read_struct ();
    open my $file, '<:raw', 'nice.png' or die $!;
    $png->init_io ($file);
    $png->read_png ();
    close $file;
    # Get all valid chunks
    my $valid = $png->get_valid ();
    my @valid_chunks = sort grep {$valid->{$_}} keys %$valid;
    print "Valid chunks are ", join (", ", @valid_chunks), "\n";
    # Print image information
    my $header = $png->get_IHDR ();
    for my $k (keys %$header) {
        print "$k: $header->{$k}\n";
    }

(This example is included as examples/synopsis.pl in the distribution.)

Simple interface:

    use Image::PNG::Libpng ':all';
    my $png = read_png_file ('../t/tantei-san.png');
    # Get all valid chunks
    my $valid = $png->get_valid ();
    my @valid_chunks = sort grep {$valid->{$_}} keys %$valid;
    print "Valid chunks are ", join (", ", @valid_chunks), "\n";
    # Print image information
    my $header = $png->get_IHDR ();
    for my $k (keys %$header) {
        if ($k eq 'color_type') {
            print "$k: " . color_type_name ($header->{$k}) . "\n";
        }
        else {
            print "$k: $header->{$k}\n";
        }
    }
    my $wpng = $png->copy_png ();
    $wpng->write_png_file ('new.png');
    

(This example is included as examples/synopsis-easy.pl in the distribution.)

VERSION

This document describes Image::PNG::Libpng version 0.56, corresponding to git commit 2ca3f20707e31fd38b2a3c3c658efb3f18eb12df at Wed Dec 23 07:44:31 2020 +0900.

Unless otherwise qualified, comments in this document on the libpng source code and documentation refer to libpng version 1.6.37. Libpng is not bundled with this distribution, so your installed version may vary.

DESCRIPTION

Image::PNG::Libpng is a Perl library for accessing the contents of PNG (Portable Network Graphics) images. Image::PNG::Libpng enables Perl to use the "libpng" library to read and write files in PNG format. Image::PNG::Libpng does not contain the libpng library. The libpng library must be installed on your computer prior to installing Image::PNG::Libpng.

Image::PNG::Libpng consists of Perl subroutines which mirror the C functions in libpng, plus helper subroutines to make it easier to read and write PNG data in Perl.

For those familiar with libpng, the section "Differences from libpng" explains the differences with libpng.

FUNCTIONS

The functions in the module are gathered into the following categories: "Simple input and output", which describes some convenience functions, "Libpng-style input and output", which describes functions which closely mirror libpng, "The image header", which describes functions for reading and writing the meta-information about PNG images, "Image data", which describes functions for accessing the image data itself, "Text chunks", "Private chunks", "Library version functions", "Compression and filtering", "Other chunks", "Libpng transformations", "Other libpng functions", functions from libpng which don't fit elsewhere, and "Other functions", which are functions specific to this module.

Simple input and output

These convenience functions combine common operations. They are not part of the original libpng API.

copy_png

    my $outpng = $png->copy_png ();

Copy a PNG from a read to a write structure. This function bridges two kinds of object, "read a png" objects created by "create_read_struct" and "write a png" objects created by "create_write_struct". This function copies all the valid chunks from a read structure to a write structure.

The following example demonstrates copying a PNG.

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng qw(read_png_file write_png_file) ;
    my $pngin = read_png_file ("$Bin/../t/tantei-san.png");
    my $pngout = $pngin->copy_png ();
    $pngout->set_text ([{key => 'Name', text => 'Shunsaku Kudo'}]); 
    # $pngout->write_png_file ('copy.png');

(This example is included as examples/copy-png.pl in the distribution.)

create_reader

     my $png = create_reader ('file.png');
     $png->read_png ();

This combines "create_read_struct", open, and "init_io" on the specified file name but does not read the file in. This is for the case that the user wants to apply some kind of transformation.

Setting the background

In the following example, the user sets the background with "set_background" to replace the alpha channel.

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng ':all';
    use Image::PNG::Const ':all';
    my $file = "$Bin/luv.png";
    my %color = (red => 0xC0, green => 0xFF, blue => 0xFF);
    my $png = create_reader ($file);
    $png->set_background (\%color, PNG_BACKGROUND_GAMMA_SCREEN, 0);
    $png->read_png ();
    my $wpng = copy_png ($png);
    $wpng->write_png_file ("$Bin/set-background.png");

(This example is included as examples/set-background.pl in the distribution.)

examples/luv.png is included in the distribution.

The input image has a transparent background:

examples/luv.png

examples/set-background.png is included in the distribution.

The output image has a blue background:

examples/set-background.png

This function was added to the module in version 0.53.

create_writer

     my $png = create_writer ('file.png');
     $png->set_IHDR (\%ihdr);
     $png->set_rows (\@rows);
     $png->write_png ();

This combines "create_write_struct", open, and "init_io" on the specified file name but does not read the file in. This is for the case that the user wants to apply some kind of transformation before writing.

This function was added to the module in version 0.53.

read_from_scalar

     my $png = read_from_scalar ($string);

This creates an image structure $png from the contents of a Perl scalar variable $string containing PNG image data, for example data read from a file, or data obtained from a web page. The first argument, $png, is a PNG structure created with "create_read_struct". It reads in all the data from the structure on being called.

This is useful when image data is stored in a Perl scalar. For example

    use Image::PNG::Libpng 'read_from_scalar';
    use LWP::Simple;
    use JSON::Create;
    my $image_data = get 'http://libpng.org/pub/png/img_png/libpng-88x31.png';
    # Now $image_data contains the PNG file
    my $png = read_from_scalar ($image_data);
    # Now $png contains the PNG information from the image.
    # Get the header.
    my $header = $png->get_IHDR ();
    my $jc = JSON::Create->new ();
    $jc->indent (1);
    print $jc->run ($header);

(This example is included as examples/get-www-png.pl in the distribution.)

The output looks like this:

    {
            "width":88,
            "bit_depth":4,
            "interlace_method":0,
            "color_type":3,
            "height":31
    }

See also "Input/output manipulation functions".

read_png_file

     my $png = read_png_file ('q.png');

Open q.png and read its contents into $png.

This combines "create_read_struct", open, "init_io", and "read_png". The return value is the same as that of "create_read_struct" with the entire PNG image already read in.

The optional argument to "read_png" can be specified using an optional transforms argument:

    my $png = read_png_file ('file.png', transforms => PNG_TRANSFORM_EXPAND);

"croak" in Carp is used to signal errors opening or closing the file.

set_transforms

    $png->set_transforms (PNG_TRANSFORM_BGR);

Set transforms for reading and writing. This is the same as the optional argument to "read_png" or "write_png". If both this and the optional argument are given, the optional argument overrides what is set here.

write_png_file

    $png->write_png_file ('nice.png');

This combines open, "init_io", and "write_png" to write an entire PNG image out to the file name specified by the argument. $png must be the object created by "create_write_struct", so "read_png_file" followed by a call to this does not work. See "copy_png" if you need to do that kind of operation.

The optional argument to "write_png" can be specified using "set_transforms".

"croak" in Carp is used to signal errors opening or closing the file.

write_to_scalar

    my $image_data = $png->write_to_scalar ();

This writes the PNG image data in $png into a Perl scalar. The first argument, $png, is a writeable PNG structure created with "create_write_struct". The return value of the subroutine is the Perl scalar containing the image data.

So, for example,

    # This CGI script prints a PNG in a random colour.
    
    use Image::PNG::Libpng ':all';
    use Image::PNG::Const ':all';
    my $png = create_write_struct ();
    my $size = 100;
    $png->set_IHDR ({height => $size, width => $size, bit_depth => 8,
                     color_type => PNG_COLOR_TYPE_RGB});
    my $bytes = pack "CCC", randcol (), randcol (), randcol ();
    my @rows = ($bytes x $size) x $size;
    $png->set_rows (\@rows);
    my $img = $png->write_to_scalar ();
    binmode STDOUT;
    print "Content-Type:image/png\r\n\r\n$img";
    exit;
    sub randcol
    {
        return int (rand () * 0x100);
    }

(This example is included as examples/png-cgi.pl in the distribution.)

See also "Input/output manipulation functions".

The optional argument to "write_png" can be specified using "set_transforms".

Libpng-style input and output

There are two different "new"-like functions, depending on whether you want to read or write a PNG, "create_read_struct" and "create_write_struct". These are based on the libpng API. Input uses open plus "init_io" followed by "read_png" or "write_png".

Examples

A self-pixellating program

This example demonstrates writing a monochrome PNG by creating a write structure with "create_write_struct", opening a filehandle to write it, associating the filehandle with the PNG structure with "init_io", then using the functions "write_info", "write_image", and then "write_end" to actually write the PNG data to the file.

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng ':all';
    use Image::PNG::Const ':all';
    my $outfile = "$Bin/mono.png";
    my ($height, $width, $rows) = pixelate (__FILE__, 5);
    my $png = create_write_struct ();
    open my $out, ">:raw", $outfile or die $!;
    $png->init_io ($out);
    $png->set_IHDR ({height => $height, width => $width, bit_depth => 1,
                     color_type => PNG_COLOR_TYPE_GRAY});
    $png->set_text ([{key => 'silly', text => 'finely-tuned breakfast cereal',}]);
    $png->set_tIME ({year => 1999});
    $png->write_info ();
    $png->set_invert_mono ();
    # PNG puts the leftmost pixel in the high-order part of the byte.
    $png->set_packswap ();
    $png->write_image ($rows);
    $png->write_end ();
    close $out or die $!;
    exit;
    
    sub pixelate
    {
        my ($file, $box) = @_;
        open my $in, "<", $file or die "Can't open '$file': $!";
        my $width = 0;
        my @lines;
        while (<$in>) {
            chomp;
            s/\t/        /g;
            push @lines, $_;
            if (length ($_) > $width) {
                $width = length ($_);
            }
        }
        close $in or die $!;
        $width *= $box;
        my $height = scalar (@lines) * $box;
        my $zero = pack "C", 0;
        my $bwidth = int(($width+7)/8);
        my @rows = ($zero x $bwidth) x $height;
        for my $r (0..$height-1) {
            my $y = int ($r/$box);
            my $line = $lines[$y];
            for my $x (0..length ($line) - 1) {
                if (substr ($line, $x, 1) ne ' ') {
                    for my $c (0..$box - 1) {
                        my $offset = $x*$box + $c;
                        my $byte = int ($offset / 8);
                        my $bit = $offset % 8;
                        my $octet = ord (substr ($rows[$r], $byte, 1));
                        substr ($rows[$r], $byte, 1) = chr ($octet | 1<<$bit);
                    }
                }
            }
        }
        return ($height, $width, \@rows);
    }

(This example is included as examples/libpng-write.pl in the distribution.)

examples/mono.png is included in the distribution.

The output has a bit depth of 1:

examples/mono.png

create_read_struct

     my $png = create_read_struct ();

Create a structure for reading a PNG. The return value can be used as an object with the other functions as methods. It can be copied to a write structure with "copy_png".

This function corresponds to png_create_read_struct in libpng plus create_info_struct (see "No info structure") with the error and warning handler variables set up to use Perl's error reporting.

create_write_struct

     my $png = create_write_struct ();

Create a structure for writing a PNG. This can be used as an object with the other functions as methods.

This function corresponds to png_create_write_struct in libpng plus create_info_struct (see "No info structure") with the error and warning handler variables set up to use Perl's error reporting.

init_io

     open my $file, "<", 'nice.png';
     $png->init_io ($file);

Set the file which $png reads or writes to $file. $file must be an already-opened Perl file handle. If $png was created with "create_write_struct", $file must be opened for writing. If $png was created with "create_read_struct", $file must be open for reading.

Since PNG files are binary files, it is safest to specify the "raw" pragma or use "binmode" with the file to override any default text file encoding which Perl might be using:

     open my $file, ">:raw", 'output.png';

or

     open my $file, ">", 'output.png';
     binmode $file;

This function corresponds to png_init_io in libpng, with a Perl file handle substituting for the C FILE *.

On some versions of Perl, "init_io" may crash in some circumstances with an error like "segmentation fault", if you use code like

    my $png = create_read_struct ();
    open my $file, "<:raw", "some.png";
    $png->init_io ($file);

and you do not check whether the call to open was successful, and some.png does not exist. The crash occurs within Perl's conversion of $file into a C FILE * pointer, before this module's code runs. This bug was fixed in Perls after version 5.24.1. To avoid trouble, please check the return value of open.

read_end

     $png->read_end ();

Read the part of the PNG file after the image data.

This function corresponds to png_read_end in libpng.

read_image

     my $rows = $png->read_image ();

Read the image data of the PNG file.

This function corresponds to png_read_image in libpng.

read_info

    $png->read_info ();

Read the part of the PNG file before the image data.

This function corresponds to png_read_info in libpng.

read_png

     $png->read_png ();

Read the entire PNG file into memory.

You can provide an argument containing transformations to apply to the image:

     use Image::PNG::Const qw/PNG_TRANSFORM_STRIP_ALPHA/;
     $png->read_png (PNG_TRANSFORM_STRIP_ALPHA);

If the argument is omitted, the default value of PNG_TRANSFORM_IDENTITY (the "do nothing" value) is applied. The possible transformations which can be applied are

PNG_TRANSFORM_BGR

Flip RGB to BGR, RGBA to BGRA. See also "set_bgr".

PNG_TRANSFORM_EXPAND

Perform set_expand(). See also "set_expand".

PNG_TRANSFORM_EXPAND_16

Expand samples to 16 bits. See also "set_expand_16".

PNG_TRANSFORM_GRAY_TO_RGB

Expand grayscale samples to RGB (or GA to RGBA). See also "set_gray_to_rgb".

PNG_TRANSFORM_IDENTITY

No transformation.

PNG_TRANSFORM_INVERT_ALPHA

Change alpha from opacity to transparency. See also "set_invert_alpha".

PNG_TRANSFORM_INVERT_MONO

Invert monochrome images. See also "set_invert_mono".

PNG_TRANSFORM_PACKING

Expand 1, 2 and 4-bit samples to bytes. See also "set_packing".

PNG_TRANSFORM_PACKSWAP

Change order of packed pixels to LSB first. See also "set_packswap".

PNG_TRANSFORM_SCALE_16

Strip 16-bit samples to 8-bit accurately. See also "set_scale_16".

PNG_TRANSFORM_SHIFT

Normalize pixels to the sBIT depth.

PNG_TRANSFORM_STRIP_16

Chop 16-bit samples to 8-bit less accurately. See also "set_strip_16".

PNG_TRANSFORM_STRIP_ALPHA

Discard the alpha channel.

PNG_TRANSFORM_SWAP_ALPHA

Flip RGBA to ARGB or GA to AG. See also "set_swap_alpha".

PNG_TRANSFORM_SWAP_ENDIAN

Byte-swap 16-bit samples. See also "set_swap".

This function corresponds to png_read_png in libpng with a default value for the third argument. The fourth, unused, argument to png_read_png does not need to be supplied. See "Unused arguments omitted".

It does not take a second "info" argument. See "No info structure".

read_update_info

    $png->read_update_info ();

⛐🤪⚠ Inside Image::PNG::Libpng, the libpng function png_read_update_info is called before reading image data. According to "The libpng documentation", this function may only be called once for any particular info structure. So although the above Perl interface exists in the module, it is strongly recommended to not use this unless you know exactly what you are doing, since it will usually cause an error when the image data is read.

This function corresponds to png_read_update_info in libpng

write_end

    $png->write_end ();

Write the final part of the PNG file.

This function corresponds to png_write_end in libpng.

write_image

    $png->write_image ($rows);

Write the image of the PNG file. $rows is an array reference as per "set_rows".

This function corresponds to png_write_image in libpng.

write_info

    $png->write_info ();

Write the first part of the PNG file.

This function corresponds to png_write_info in libpng.

write_png

    $png->write_png ();

This writes the PNG to the file stream which was associated with it using "init_io". For example,

    open my $output, ">:raw", 'out.png';
    $png->init_io ($output);
    $png->write_png ();
    close $output;

An optional argument consists of transformations to apply to the PNG image before writing it:

    use Image::PNG::Const qw/PNG_TRANSFORM_STRIP_ALPHA/;
    $png->write_png (PNG_TRANSFORM_STRIP_ALPHA);

The transformations which can be applied are as follows:

PNG_TRANSFORM_BGR

Flip RGB to BGR, RGBA to BGRA. See also "set_bgr".

PNG_TRANSFORM_INVERT_ALPHA

Change alpha from opacity to transparency. See also "set_invert_alpha".

PNG_TRANSFORM_INVERT_MONO

Invert monochrome images. See also "set_invert_mono".

PNG_TRANSFORM_PACKING

Expand 1, 2 and 4-bit samples to bytes. See also "set_packing".

PNG_TRANSFORM_PACKSWAP

Change order of packed pixels to LSB first. See also "set_packswap".

PNG_TRANSFORM_SHIFT

Normalize pixels to the sBIT depth.

PNG_TRANSFORM_STRIP_FILLER_AFTER

Strip out trailing filler bytes.

PNG_TRANSFORM_STRIP_FILLER_BEFORE

Strip out leading filler bytes.

PNG_TRANSFORM_SWAP_ALPHA

Flip RGBA to ARGB or GA to AG. See also "set_swap_alpha".

PNG_TRANSFORM_SWAP_ENDIAN

Byte-swap 16-bit samples. See also "set_swap".

This function corresponds to png_write_png in libpng.

The image header

These functions handle the header part of PNG image data. See this subsection of "The PNG specification" for information on the PNG standards for the image header.

color_type_name

    $name = color_type_name ($color_type);

This is a convenience function which returns a string corresponding to the numerical color type in $color_type. The name is in upper case, with words separated by underscores, as in RGB_ALPHA.

    use Image::PNG::Libpng ':all';
    my $png = read_png_file ('tantei-san.png');
    my $name = color_type_name ($png->get_IHDR->{color_type});
    print "Your PNG has colour type $name.\n";

(This example is included as examples/color-type-name.pl in the distribution.)

This function does not correspond to anything in libpng. The names of the color types are taken from those defined in the libpng header file, png.h.

get_bit_depth

    my $bit_depth = $png->get_bit_depth ();

Get the bit depth, the number of bits for one channel of one pixel.

This function corresponds to png_get_bit_depth in libpng

get_channels

    my $channels = $png->get_channels ();

Get the number of channels, from one to four. The channels are the components of pixels, for example the red channel or the alpha (transparency) channel. The return value is 1 for color type PNG_COLOR_TYPE_GRAY and PNG_COLOR_TYPE_PALETTE, 2 for PNG_COLOR_TYPE_GRAY_ALPHA, 3 for PNG_COLOR_TYPE_RGB and 4 for PNG_COLOR_TYPE_RGB_ALPHA or PNG_COLOR_TYPE_RGB with a filler byte. Note that the number of channels does not necessarily correspond to the number of bytes, since the bit depth can also be 1, 2, 4, 8, or 16, depending on the color type. See also the convenience function "color_type_channels".

This function corresponds to png_get_channels in libpng

get_color_type

    my $color_type = $png->get_color_type ();

This returns an integer value. If you want to get a name for the color type, use "color_type_name".

This function corresponds to png_get_color_type in libpng.

get_IHDR

    my $IHDR = $png->get_IHDR ();

Read the IHDR information from the PNG file. The return value is a hash reference containing the following key/value pairs:

width

The width of the image in pixels.

height

The height of the image in pixels.

bit_depth

The bit depth of the image (the number of bits used for each color in a pixel). This can take the values 1, 2, 4, 8, 16.

color_type

The color type. This can take the values PNG_COLOR_TYPE_GRAY, PNG_COLOR_TYPE_GRAY_ALPHA, PNG_COLOR_TYPE_PALETTE, PNG_COLOR_TYPE_RGB, PNG_COLOR_TYPE_RGB_ALPHA.

interlace_method

The method of interlacing. This can take the values PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7.

So, for example, to get the width and height of an image,

    my $ihdr = $png->get_IHDR ();
    printf "Your image is %d x %d\n", $ihdr->{width}, $ihdr->{height};

This function corresponds to png_get_IHDR in libpng, with a single Perl hash reference used instead of the several pointers to integers used in libpng.

It does not return the fields filter_type and compression_type, since these are always 0. See "Unused arguments omitted".

get_image_height

    my $height = $png->get_image_height ();

Get the image's height from the header.

This function corresponds to png_get_image_height in libpng

get_image_width

    my $width = $png->get_image_width ();

Get the image's width from the header.

This function corresponds to png_get_image_width in libpng

get_interlace_type

    my $interlace_type = $png->get_interlace_type ();

Get the interlace type. This is either PNG_INTERLACE_NONE or PNG_INTERLACE_ADAM7.

This function corresponds to png_get_interlace_type in libpng

get_valid

    my $valid = $png->get_valid ();
    if ($valid->{oFFs}) {
        print "The PNG has valid screen offsets.\n";
    }

This function returns a hash with a key for each possible chunk which may or may not be valid. The chunks which you can test for are "bKGD", "cHRM", "eXIf", "gAMA", "hIST", "hIST", "iCCP", IDAT, IHDR, "iTXt", "oFFs", "pCAL", "pHYs", "PLTE", "sBIT", "sCAL", "sPLT", "sRGB", "tEXt", "tIME", "tRNS", and "zTXt".

Whereas "libpng_supports" tells you whether the installed libpng on your system supports various chunks, this tells you whether the chunks are present in a particular PNG image file.

The first argument, $png, is a PNG structure created with "create_read_struct".

This function corresponds to png_get_valid in libpng, with the difference being that the return value is a hash containing a key for each possible chunk.

height

    my $height = $png->height ();

Alias for "get_image_height". This is not exported, it's intended for use with the object only.

set_IHDR

    my $ihdr = {width => 10, height => 10, bit_depth => 8,
                color_type => PNG_COLOR_TYPE_RGB};
    $png->set_IHDR ($ihdr);

Set the IHDR chunk (the image header) of the PNG image.

The first argument, $png, is a writeable PNG structure created with "create_write_struct". The second argument is a hash with the following values:

width

The width of the image in pixels. This cannot be zero, negative, or omitted.

height

The height of the image in pixels. This cannot be zero, negative, or omitted.

bit_depth

The bit depth of the image (the number of bits used for each color in a pixel). This cannot be omitted. This can have the values 1, 2, 4, 8, 16.

color_type

The color type. This cannot be omitted. This can have the values PNG_COLOR_TYPE_GRAY, PNG_COLOR_TYPE_GRAY_ALPHA, PNG_COLOR_TYPE_PALETTE, PNG_COLOR_TYPE_RGB, PNG_COLOR_TYPE_RGB_ALPHA.

interlace_method

The method of interlacing. If this is omitted, it's set to PNG_INTERLACE_NONE. This can have the values PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7.

Other fields in the hash are ignored.

This function corresponds to png_set_IHDR in libpng, with a single Perl hash reference used instead of the seven integers.

The values compression_method, filter_method, may be supplied by the user but are ignored since they may only take the value 0. See "Unused arguments omitted".

sig_cmp

    if (sig_cmp ($should_be_png)) {
        print "Your data does not have a PNG signature.\n";
    }

This subroutine looks at $should_be_png and checks whether its first bytes correspond to a valid PNG signature. It returns a true value if they do not.

It can also take two further arguments consisting of a byte offset and a number of bytes to check respectively:

    sig_cmp ($should_be_png, 0, 8);

If these arguments are not supplied, the byte offset is assumed to be zero, and the number of bytes to check is assumed to be eight.

This function corresponds to png_sig_cmp in libpng, with default arguments of 0 and 8 if second and third arguments are not supplied.

width

    my $width = $png->width ();

Alias for "get_image_width". This is not exported, it's intended for use with the object only.

Image data

These functions deal with accessing the image data itself.

get_rowbytes

    my $bytes_in_a_row = $png->get_rowbytes ();

This returns the number of bytes needed to hold a transformed row of an image.

This function corresponds to png_get_rowbytes in libpng.

get_rows

    my $rows = $png->get_rows ();
    my $pixel = substr ($rows->[10], 20, 1);

This returns the rows of the PNG image, after uncompressing and unfiltering, as binary data. The return value, $rows in the example, is an array reference with a number of rows equal to the height of the PNG image. Each element of the array reference is a string containing the binary data making up a row of the image. The values of individual pixels can be extracted from using a function such as substr or unpack. This binary data is likely to contain bytes equal to zero.

"get_rowbytes" gives the number of bytes in each row.

Each row is a Perl string. Perl terminates each row of data with an extra zero byte at the end.

This function corresponds to png_get_rows in libpng.

set_rows

    $png->set_rows (\@rows);

Set the rows of data to be written in to the PNG to @rows. @rows needs to contain at least the same number of rows of data as the height of the PNG image set with "set_IHDR", and the length of each entry needs to be at least the width of the image multiplied by the number of bytes required for each pixel.

This function was changed to copy the data in version 0.46.

This function corresponds to png_set_rows in libpng.

Text chunks

See this subsection of "The PNG specification" for information on the PNG standards for text information.

get_text

    my $text_chunks = $png->get_text ();

This gets all of the text chunks in the PNG image and returns them as an array reference. Each element of the array represents one text chunk. This element is a hash reference with keys such as key, lang_key, or compression taken from the PNG's information.

The text data is uncompressed by libpng. If it is international text (ITXT), it is put into Perl's internal Unicode encoding if it is found to be valid UTF-8. (PNG "international text", ITXT is required to be in the UTF-8 encoding, and non-international text is required to contain whitespace and printable ASCII characters only. See "The PNG specification" for more on the requirements of a PNG text section.)

This function corresponds to png_get_text in libpng, with a Perl array of hash references substituted for the C array of structs used by libpng. See "set_text" for details of the keys and values which may be returned.

set_text

    $png->set_text ([\%chunk1, \%chunk2]);

This sets the text chunks in a writeable image. The input value is an array reference containing one or more hash references. Each hash reference must have a key value for the text. According to the PNG specification, this should be between one and 79 bytes in length. This module enforces that restriction, so if you supply a key longer than that, the chunk cannot be added. A hash reference may also have the following:

compression

The value of compression controls the compression of the text. If compression is not supplied, a default value of PNG_TEXT_COMPRESSION_NONE is applied. The compression field can take the following values, available from Image::PNG::Const:

PNG_TEXT_COMPRESSION_NONE

TEXT = Printable ASCII and space characters.

PNG_TEXT_COMPRESSION_zTXt

TEXT = Printable ASCII and space characters.

PNG_ITXT_COMPRESSION_NONE

ITXT = International text, should be UTF-8.

PNG_ITXT_COMPRESSION_zTXt

ITXT = International text, should be UTF-8.

itxt_length

The string length of international text in bytes.

⛐🤪⚠ This is ignored by libpng when writing text chunks. When reading text chunks, if the text is marked as international text, libpng adds the length of the string in bytes in this field rather than in "text_length".

lang

This should be set to name of the language of the text, if the text chunk is iTXt. According to the PNG specification, "It is an ISO 646.IRV:1991 [ISO 646] string consisting of hyphen-separated words of 1-8 alphanumeric characters each (for example cn, en-uk, no-bok, x-klingon, x-KlInGoN)."

Support for writing lang was added in version 0.49 of this module. (Prior to that undocumented support existed via a differently-named key.)

⛐🤪⚠ This module does not attempt to check the supplied value, but merely passes it to libpng. libpng appears not to check the value either, nor to enforce restrictions on its length.

lang_key

This corresponds to the "Translated keyword" of the PNG specification. Note that the user needs to supply key and "lang" as well as lang_key.

Support for writing lang_key was added in version 0.49 of this module. (Prior to that undocumented support existed via a differently-named key.)

text

The value of "text" is added to the PNG as the text segment. This can be omitted if you just want to write a key value without any accompanying text.

text_length

The string length in bytes. The user may set this, but it is ignored when writing PNG text chunks, instead this module uses the string length obtained from Perl. This key contains the length of the string when reading text chunks via "get_text", but if the text is marked as international text, "itxt_length" is used to return its length in bytes, rather than this.

Whether or not the value of text is an ITXT field is decided by the value of compression.

People who want to fiddle with the text compression applied can do so via "set_text_compression_level" and the other functions described below that.

If set_text is called more than once, the chunks are not overwritten but appended to the existing ones. (This behaviour is from libpng itself.)

Prior to version 0.50, set_text would fail silently if the user added invalid chunks, for example hash references without a valid key, or things which were not hash references at all. From version 0.50, all invalid inputs cause fatal errors. However, unknown keys in the hash references do not cause fatal errors.

This function corresponds to png_set_text in libpng.

Example

    use utf8;
    use Image::PNG::Const ':all';
    use Image::PNG::Libpng ':all';
    my $png = create_write_struct ();
    $png->set_IHDR ({width => 1, height => 1, bit_depth => 8,
                     color_type => PNG_COLOR_TYPE_GRAY});
    $png->set_rows (['X']);
    $png->set_text ([{
        compression => PNG_TEXT_COMPRESSION_NONE,
        key => "Copyright",
        text => "Copyright (C) 2020 Fat Cat",
    }, {
        compression => PNG_ITXT_COMPRESSION_zTXt,
        key => "Copyright",
        lang_key => '著者権',
        lang => 'ja_JP',
        text => '©令和2年茉莉ニャンバーガーさん',
    }]);
    $png->write_png_file ('copyright.png');

(This example is included as examples/set-text.pl in the distribution.)

text_compression_name

    my $name = Image::PNG::Libpng::text_compression_name ($text->{compression});

Given a numerical text compression type, return the equivalent name. The name is in upper case. The possible return values are

TEXT_NONE
TEXT_zTXt
ITXT_NONE
ITXT_zTXt
an empty string

if the compression method is unknown.

The compression field is also used to store the information about whether the text is "international text" in UTF-8 or not.

This function does not correspond to anything in libpng. The names of the text compression types are based on those in png.h, but without the word "COMPRESSION", so for example the libpng constant PNG_ITXT_COMPRESSION_zTXt corresponds to a return value of ITXT_zTXt.

Private chunks

See this subsection of "The PNG specification" for information on the PNG standards for private chunks.

To test whether your version of libpng supports private chunks, use "libpng_supports" with argument '"UNKNOWN_CHUNKS"':

    if (libpng_supports ('UNKNOWN_CHUNKS')) {
        # do something
    }

get_chunk_malloc_max

    my $max = $png->get_chunk_malloc_max ();

This gets the maximum amount of memory that a chunk can use.

To test whether your version of libpng supports get_chunk_malloc_max, use "libpng_supports" with argument '"CHUNK_MALLOC_MAX"':

    if (libpng_supports ('CHUNK_MALLOC_MAX')) {
        # do something
    }

This function corresponds to png_get_chunk_malloc_max in libpng

set_chunk_malloc_max

    $png->set_chunk_malloc_max ($max);

This sets the maximum amount of memory that a chunk can use. The default value of libpng 1.6.37 is 8 megabytes.

To test whether your version of libpng supports get_chunk_malloc_max, use "libpng_supports" with argument '"CHUNK_MALLOC_MAX"':

    if (libpng_supports ('CHUNK_MALLOC_MAX')) {
        # do something
    }

This function corresponds to png_set_chunk_malloc_max in libpng

set_keep_unknown_chunks

    use Image::PNG::Const 'PNG_HANDLE_CHUNK_ALWAYS';
    $png->set_keep_unknown_chunks (PNG_HANDLE_CHUNK_ALWAYS);

Tell libpng not to discard unknown chunks when reading the file.

get_unknown_chunks

    my $private_chunks = $png->get_unknown_chunks ();
    # Get some data from a private chunk
    my $chunk_three_data = $private_chunks->[3]->{data};
    # Get the size of the data
    print length $chunk_three_data;

This gets all of the private chunks from the image. The return value is an array reference containing hash references. If there are no private chunks, this returns an undefined value. There is one element of the array for each chunk member. It is necessary to call "set_keep_unknown_chunks" with an appropriate value before reading the file, otherwise libpng discards unknown chunks when reading the file.

Each member hash reference has the following keys:

name

The name of the unknown chunk, in the PNG chunk format (four bytes).

location

The location of the unknown chunk.

data

The data of the unknown chunk

The "size" field of the PNG structure is not stored, because the "data" member of the hash contains information on its length.

This function corresponds to png_get_unknown_chunks in libpng

set_unknown_chunks

    $png->set_unknown_chunks (name => 'dUCk', data => 'abcdefg',
                              location => PNG_AFTER_IDAT);

⛐🤪⚠ This currently does not fully function.

This function corresponds to png_set_unknown_chunks in libpng

Library version functions

access_version_number

    my $libpng_version_number = Image::PNG::Libpng::access_version_number ();

This function returns the version of the libpng library which the module is using as an integer number.

This function corresponds to png_access_version_number in libpng.

get_libpng_ver

    my $libpng_version = Image::PNG::Libpng::get_libpng_ver ();

This function returns the version of the libpng library which the module is using.

This function corresponds to png_get_libpng_ver in libpng. However, it doesn't require the png_structp argument of the C function. See "Unused arguments omitted".

Compression and filtering

get_compression_buffer_size

    my $size = $png->get_compression_buffer_size ();

Returns the value of the compression buffer size, which may be altered with "set_compression_buffer_size".

set_compression_buffer_size

    $png->set_compression_buffer_size (100);

Set the size of the buffer which zlib uses to compress or decompress the image data. It takes one argument, an integer number. This cannot be less than 6.

set_compression_level

    $png->set_compression_level ($number);

Set the compression level used to make the PNG. A value of 0 (Z_NO_COMPRESSION) corresponds to no compression at all, otherwise $number may take values of 1 (Z_BEST_SPEED) to 9 (Z_BEST_COMPRESSION), with smaller values giving faster, and larger values giving better, that is with smaller output, compression. These correspond to the -1, -2, ... -9 options to the gzip utility, or the compression level parameter of zlib. Calling with -1 (Z_DEFAULT_COMPRESSION) reverts to the default compression. Calling with any other number outside the range 0 to 9 results in a fatal error.

This function was added to the module in version 0.49.

set_compression_mem_level

    $png->set_compression_mem_level ($argument);

⛐🤪⚠ Untested function corresponding to png_set_compression_mem_level. Takes one integer argument.

Sets the memLevel parameter of deflateInit2 in zlib when writing PNG image data. Argument between 1 for minimum memory and 9 for maximum speed. The default is 8. See "zlib documentation".

set_compression_window_bits

    $png->set_compression_window_bits ($argument);

⛐🤪⚠ Untested function corresponding to png_set_compression_window_bits. Takes one integer argument.

Sets the windowBits parameter of deflateInit2 in zlib when writing PNG image data. Argument value must be between 8 and 15 for libpng. The default is 15. See the "zlib documentation".

set_compression_strategy

    $png->set_compression_strategy ($argument);

⛐🤪⚠ Untested function corresponding to png_set_compression_strategy. Takes one integer argument.

Sets the strategy parameter of deflateInit2 when writing PNG image data. Setting this with libpng overrides libpng's default behaviour of changing the value depending on the filter in use. For zlib, the argument is either 0 for default behaviour, or 1 to 4. See the "zlib documentation". libpng uses the default strategy 0 (Z_DEFAULT_STRATEGY) for unfiltered image data, and 1 (Z_FILTERED) for filtered image data.

set_filter

    use Image::PNG::Const 'PNG_FILTER_NONE';
    $png->set_filter (PNG_FILTER_NONE);

This sets the filters which are allowed to be used for writing a PNG image. The possible values are

PNG_NO_FILTERS
PNG_FILTER_NONE
PNG_FILTER_SUB
PNG_FILTER_UP
PNG_FILTER_AVG
PNG_FILTER_PAETH
PNG_ALL_FILTERS

These can be combined using | (logical or):

    use Image::PNG::Const ':all';
    set_filter ($png, PNG_FILTER_UP | PNG_FILTER_AVG);

Please see this subsection of "The PNG specification" for the meanings of these filter types.

This function corresponds to png_set_filter in libpng with the second (unused) argument omitted. See "Unused arguments omitted".

set_text_compression_level

    $png->set_text_compression_level ($argument);

⛐🤪⚠ Untested function corresponding to png_set_text_compression_level. Takes one integer argument.

As "set_compression_level" but for compressed text.

set_text_compression_mem_level

    $png->set_text_compression_mem_level ($argument);

⛐🤪⚠ Untested function corresponding to png_set_text_compression_mem_level. Takes one integer argument.

As "set_compression_mem_level" but for compressed text.

set_text_compression_window_bits

    $png->set_text_compression_window_bits ($argument);

⛐🤪⚠ Untested function corresponding to png_set_text_compression_window_bits. Takes one integer argument.

As "set_compression_window_bits" but for compressed text.

set_text_compression_strategy

    $png->set_text_compression_strategy ($argument);

⛐🤪⚠ Untested function corresponding to png_set_text_compression_strategy. Takes one integer argument.

As "set_compression_strategy" but for compressed text.

Other chunks

These routines deal with the other possible chunks of PNGs.

The getter and setter routines for all other chunks are designed so that the return value of get_wXYZ is able to be used directly as the value for set_wXYZ, so the values of chunks can easily be copied from one PNG to another.

    my $values = $png1->get_wXYZ ();
    $png2->set_wXYZ ($values);

If the chunk is not present, or if the chunk is not supported by the user's version of libpng, the return value of get_wXYZ is the undefined value.

bKGD

The background color of the PNG image.

See this subsection of "The PNG specification" for information on the PNG standards for the background chunk.

get_bKGD

    my $bkgd = $png->get_bKGD ();

Get the bKGD (background) chunk of the image.

The return value is a hash with the following keys, depending on the color type of the image:

index

For palette color types, this is the offset into the palette.

gray

For grayscale color types.

red
green
blue

This function corresponds to png_get_bKGD in libpng with a hash function instead of a png_color struct.

set_bKGD

    $png->set_bKGD ($bkgd);

Set the bKGD (background) chunk of the image. $bkgd is a hash reference. The keys of the hash reference are as described in "get_bKGD".

This function corresponds to png_set_bKGD in libpng with a hash function instead of a png_color struct.

cHRM

See this subsection of "The PNG specification" "cHRM Primary chromaticities and white point".

get_cHRM

    my $cHRM = $png->get_cHRM ();

Get the cHRM chunk as a hash reference.

The keys of the hash are white_x, white_y, red_x, red_y, green_x, green_y, blue_x, blue_y.

The values of the hash are floating point numbers between 0 and 1.

This function corresponds to png_get_cHRM in libpng with a hash reference supplying the arguments. The hash's keys correspond to the names of the double arguments in libpng.

get_cHRM_XYZ

    my $cHRM = $png->get_cHRM_XYZ ();

Get the cHRM chunk as a hash reference for the XYZ color space.

The keys of the hash are red_x, red_y, red_z, green_x, green_y, green_z, blue_x, blue_y, blue_z.

The values of the hash are floating point numbers between 0 and 1.

This function corresponds to png_get_cHRM_XYZ in libpng with a hash reference supplying the arguments. The hash's keys correspond to the names of the double arguments in libpng.

set_cHRM

    $png->set_cHRM (\%cHRM);

Set the cHRM chunk from a hash reference.

The keys of the hash are as for "get_cHRM". The values are floating point numbers between 0 and 1.

This function corresponds to png_set_cHRM in libpng with a hash reference instead of the double arguments.

set_cHRM_XYZ

    $png->set_cHRM_XYZ (\%cHRM);

Set the cHRM chunk from a hash reference for the XYZ color space.

The keys of the hash are as for "get_cHRM_XYZ". The values are floating point numbers between 0 and 1. The Y values red_y, green_y, and blue_y should sum to 1. If you supply values outside the allowed range, libpng corrects them silently on writing rather than producing an error.

This function corresponds to png_set_cHRM_XYZ in libpng with a hash reference instead of the double arguments.

eXIf

The eXIf chunk is an extension to the PNG specification. See http://www.simplesystems.org/png-group/proposals/eXIf/. Support for this chunk was added in version 0.50 of this module.

get_eXIf

    my $exif = $png->get_eXIf ();

This retrieves the eXIf chunk from $png but does not process the internal information.

set_eXIf

    $png->set_eXIf ($exif);

libpng checks whether the chunk's first two bytes are either II for little-endian (from Intel) or MM for big-endian (from Motorola) then adds the entire chunk, including the first two bytes, to the PNG.

⛐🤪⚠ As of December 2020, there appears to be a bug in libpng in which the eXIf chunk is added twice, causing a warning of the form libpng warning: eXIf: duplicate on reading a PNG file back in. See https://github.com/glennrp/libpng/pull/351.

gAMA

See this subsection of "The PNG specification".

get_gAMA

    my $gamma = $png->get_gAMA ();

Get the gamma value or gAMA chunk. The return value is a floating-point number.

This function corresponds to png_get_gAMA in libpng

set_gAMA

    $png->set_gAMA (0.2);

Set the gamma value or gAMA chunk.

This function corresponds to png_set_gAMA in libpng

hIST

See this subsection of "The PNG specification".

get_hIST

    my $hist = $png->get_hIST ();

If the PNG file contains a histogram, the return value is array reference, otherwise it is the undefined value. The number of entries in the array reference is the same as in the palette.

This function corresponds to png_get_hIST in libpng

set_hIST

    $png->set_hIST (\@hist);

Set the histogram of frequencies of the colors of a paletted ("PLTE") image. The entries of the histogram are 16 bit unsigned integers, so the maximum value that can be entered is 65535 = 2^16 - 1. Larger numbers and floating point numbers will cause a warning to be printed and the value to be set to zero. The histogram must have exactly the same number of entries as the palette or the call will fail with a warning. A histogram cannot be added to an image without a palette. A call to set_hIST for an image without a palette will cause a warning and return without setting the value.

(I'm not sure of the best form of error handling for this function so it may change in future versions to have errors for a bad histogram rather than warnings.)

This function corresponds to png_set_hIST in libpng

iCCP

See this subsection of "The PNG specification".

get_iCCP

    my $iccp = $png->get_iCCP ();

The return value is a hash with two keys,

name

The name of the profile.

profile

The color profile.

The compression_type value, which is always 0, is not returned. See "Unused arguments omitted".

This function corresponds to png_get_iCCP in libpng.

set_iCCP

    $png->set_iCCP ({name => 'name', profile => 'profile'});

This function corresponds to png_set_iCCP in libpng. For details of the arguments, see "get_iCCP". A compression_type argument, which must always be 0 anyway, is ignored. See "Unused arguments omitted".

oFFs

This is an extension to the PNG specification. See http://www.libpng.org/pub/png/spec/1.1/pngext-1.1.0-pdg.html#C.oFFs.

get_oFFs

    my $phys = $png->get_oFFs ();

Get oFFs chunk. Return value is a hash reference

This function corresponds to png_get_oFFs in libpng

set_oFFs

    $png->set_oFFs ({x_offset => 1, y_offset => 1, unit_type => 0});

Set oFFs chunk. See the specification linked above for the meanings of the parameters. If unit_type is not 0 or 1, a warning is produced.

This function corresponds to png_set_oFFs in libpng

pCAL

pCAL is an extension of the PNG specification which allows one to associate pixels in the PNG image with non-image data such as a heat map. See http://www.libpng.org/pub/png/spec/1.1/pngext-1.1.0-pdg.html#C.pCAL.

To test whether your version of libpng supports the pCAL extension, use "libpng_supports" with argument '"pCAL"':

    if (libpng_supports ('pCAL')) {
        # do something
    }

get_pCAL

    my $pcal = $png->get_pCAL ();

The return value is a hash reference with the following keys:

params

If this exists, its value is a reference to an array containing the parameter list of the pCAL chunk, which are floating point numbers. Within the PNG file chunk, these parameters are stored as strings representing floating point numbers, but you can pass in Perl floating point numbers rather than strings. The number of parameters you should set is fixed by the type parameter. Refer to the PNG specification for full details.

purpose

The purpose string of the pCAL chunk.

type

The equation type as a number, from 0 to 3. See the PNG specification for the meanings. Other numbers cause an error.

units

The units as a string. See the PNG specification for details.

x0

The zero value for the equation, an integer.

x1

The max value for the equation, an integer.

This function corresponds to png_get_pCAL in libpng

set_pCAL

    $png->set_pCAL (\%values);

See "get_pCAL" for the parameters of %values.

This function corresponds to png_set_pCAL in libpng

pHYs

See this subsection of "The PNG specification".

get_pHYs

    my $phys = $png->get_pHYs ();

The return value is a hash reference with the keys

res_x
res_y
unit_type

Is either 0 or 1.

This function corresponds to png_get_pHYs in libpng

set_pHYs

    $png->set_pHYs ({res_x => 1, res_y => 1, unit_type => 1});

This function corresponds to png_set_pHYs in libpng

PLTE

See this subsection of "The PNG specification" for information on the PNG standards for the palette chunk.

get_palette_max

    my $pmax = $png->get_palette_max ();

If your libpng supports it, this will return the maximum palette index found in the image, or "-1" if the palette was not checked, or "0" if no palette was found.

To test whether your version of libpng supports get_palette_max, use "libpng_supports" with argument '"GET_PALETTE_MAX"':

    if (libpng_supports ('GET_PALETTE_MAX')) {
        # do something
    }

⛐🤪⚠ This function is implemented but so far it is not very clear to me what it does, since, for example, it returns zero for the test image t/tantei-san.png which is an image with a fully-used 256-color palette. I've asked about it on the libpng mailing list, but so far nobody has responded.

get_PLTE

     my $colors = $png->get_PLTE ();
     # Get the green value of the twentieth entry in the palette.
     my $green = $colors->[20]->{green};

This function gets the palette from the PNG. The return value is an array reference containing the palette. This array contains hash references with the values "green", "blue" and "red" for the color of each pixel in the palette. If the PNG has no palette, it returns an undefined value.

A PNG image may or may not contain a palette. To check whether the image contains a palette, use something of the following form:

     use Image::PNG::Const ':all';
     my $color_type = $png->get_color_type ();
     if ($color_type == PNG_COLOR_TYPE_PALETTE) {
         # The PNG uses a palette.
     }

A PNG image may also contain a palette even when the "color_type" does not indicate that. To check for that case, use "get_valid".

This function corresponds to png_get_PLTE in libpng.

set_PLTE

    $png->set_PLTE ($palette);

Set the palette of $png. The argument is an array reference containing hash references. There is one hash reference for each palette entry. The hash references contain three fields, red, green, and blue, corresponding to the pixel value for that palette entry. Other values in the hash references are ignored. For example,

    $png->set_PLTE ([{red => 1, green => 99, blue => 0x10},
                     {red => 0xFF, green => 0xFF, blue => 0xFF}]);

creates a palette with two entries in $png.

This function corresponds to png_set_PLTE in libpng.

sBIT

See this subsection of "The PNG specification".

get_sBIT

    my $sbit = $png->get_sBIT ();

The return value is a hash reference containing integer values for the keys red, blue, green, gray and alpha, depending on the color_type of $png.

⛐🤪⚠ Prior to version 0.50 of this module, get_sBIT wrote values of 0 for all of the keys, regardless of the color_type, causing errors in some circumstances. From version 0.50, hash keys are not entered if the color_type of the image makes them invalid.

This function corresponds to png_get_sBIT in libpng

set_sBIT

    $png->set_sBIT ({red => 1, blue => 2, green => 3});

The argument is a hash reference containing integer values for the keys red, blue, green, gray, and alpha, as required by the color_type of $png.

This function corresponds to png_set_sBIT in libpng

sCAL

This is an extension to the PNG specification. See http://www.libpng.org/pub/png/spec/1.1/pngext-1.1.0-pdg.html#C.sCAL.

To test whether your version of libpng supports the sCAL chunk, use "libpng_supports" with argument '"sCAL"':

    if (libpng_supports ('sCAL')) {
        # do something
    }

get_sCAL

    my $scal = $png->get_sCAL ();

The returned hash value contains the following keys:

unit

The unit type, which is either PNG_SCALE_UNKNOWN, PNG_SCALE_METER, or PNG_SCALE_RADIAN.

width

The width, as a string.

height

The height, as a string.

This function corresponds to png_get_sCAL_s in libpng. Note that this uses the sCAL_s function rather than the get_sCAL and the returned values are the strings themselves rather than parsed numbers.

set_sCAL

    $png->set_sCAL ($scal);

$scal is a hash reference containing the keys described in "get_sCAL".

This function corresponds to png_set_sCAL_s in libpng. Note that this uses the set_sCAL_s function rather than set_sCAL and the input values are the strings themselves rather than numbers.

sPLT

See this subsection of "The PNG specification".

get_sPLT

⛐🤪⚠ Provisional. See "set_sPLT" for documentation, the return value is like the input of that.

set_sPLT

    $png->set_sPLT ([{ name => 'palette', depth => 8, entries => [{red => 1, blue => 2},]}]);

Set suggested palettes. The input is an array reference containing hash references with the following fields:

name

The name of the suggested palette.

depth

The depth of the suggested palette.

entries

The entries of the palette. This is an array reference containing hash references with keys as follows:

red
blue
green
frequency
alpha

The field nentries which is returned by "get_sPLT" does not need to be specified, it is calculated from the length of entries.

sRGB

See this subsection of "The PNG specification".

get_sRGB

    my $sRGB = $png->get_sRGB ();

The return value is an integer number corresponding to one of the following:

PNG_sRGB_INTENT_SATURATION
PNG_sRGB_INTENT_PERCEPTUAL
PNG_sRGB_INTENT_ABSOLUTE
PNG_sRGB_INTENT_RELATIVE

This function corresponds to png_get_sRGB in libpng

set_sRGB

    $png->set_sRGB ($srgb);

$srgb is one of the values described in "get_sRGB".

This function corresponds to png_set_sRGB in libpng

tIME

See this subsection of "The PNG specification" for information on the PNG standards for time stamp information.

get_tIME

    my $time = $png->get_tIME ();
    if ($time && $time->{year} < 2005) {
        warn "Your PNG is now getting old. Don't forget to oil it to prevent rust.";
    }

The return value is either the undefined value, if no tIME chunk exists in the PNG, or a hash reference containing fields year, month, day, hour, minute and second. month and day start from 1 rather than 0.

The "modification time value" of the PNG image is a chunk written into the PNG file itself, and may not have the same value as the operating system's modification time for the file. The tIME chunk is not a compulsory requirement for PNG files, and most PNG image files do not contain this chunk. PNG tIME chunks do not contain a time zone. According to this subsection of "The PNG specification", "Universal Time (UTC) should be specified rather than local time."

This function corresponds to png_get_tIME in libpng, with a Perl hash reference substituted for the C struct png_timep used in libpng.

set_tIME

    # Set the time to "now"
    $png->set_tIME ();
    # Set the time
    $png->set_tIME ({year => 1999, month => 12});

Set the modification time of the PNG to the values given by the argument, a hash reference containing the fields year, month, day for the day of the month, hour, minute, and second. The numbering for month and day is from 1, not 0. If any of year, hour, minute or second is omitted from the hash reference, these are set to zero. If month or day are omitted, these are set to 1. PNG tIME chunks do not contain a time zone. According to this subsection of "The PNG specification", "Universal Time (UTC) should be specified rather than local time." If the entire argument is omitted or contains an invalid value, the time is set to the current time.

This function corresponds to png_set_tIME in libpng, with a Perl hash reference substituted for the C struct png_timep used in libpng.

tRNS

See this subsection of "The PNG specification".

get_tRNS

    my $trns = $png->get_tRNS ();

Get the tRNS chunk. If the image is a palette type, this is an array reference. If the image is a non-palette type, this is a hash containing values for the keys red, green, blue, and gray.

set_tRNS

    $png->set_tRNS ($trns);

Set the tRNS chunk. If the image is a palette type, $trns is a reference to an array of positive or zero values between 0 and 255 of the same size as the palette. It must not contain more than 256 values. If the image is not a palette type, $trns is a hash reference containing values for the keys red, green, blue and gray.

Libpng transformations

set_bgr

    $png->set_bgr ();

⛐🤪⚠ Untested. Flips RGB to BGR. See the libpng manual for details.

To test whether your version of libpng supports set_bgr, for a read object, use "libpng_supports" with argument '"READ_BGR"'.

    if (libpng_supports ('READ_BGR')) {
        # do something
    }

For a write object, use argument '"WRITE_BGR"'.

This function corresponds to png_set_bgr in libpng

set_expand

    $png->set_expand ();

Set transformation in $png such that paletted images are expanded to RGB, grayscale images of bit-depth less than 8 are expanded to 8-bit images, and tRNS chunks are expanded to alpha channels.

To test whether your version of libpng supports set_expand, use "libpng_supports" with argument '"READ_EXPAND"':

    if (libpng_supports ('READ_EXPAND')) {
        # do something
    }

This function corresponds to png_set_expand in libpng

set_expand_16

    $png->set_expand_16 ();

⛐🤪⚠ Untested.

To test whether your version of libpng supports set_expand_16, use "libpng_supports" with argument '"READ_EXPAND_16"':

    if (libpng_supports ('READ_EXPAND_16')) {
        # do something
    }

This function corresponds to png_set_expand_16 in libpng

set_expand_gray_1_2_4_to_8

    $png->set_expand_gray_1_2_4_to_8 ();

⛐🤪⚠ Untested.

To test whether your version of libpng supports set_expand_gray_1_2_4_to_8, use "libpng_supports" with argument '"READ_EXPAND"':

    if (libpng_supports ('READ_EXPAND')) {
        # do something
    }

This function corresponds to png_set_expand_gray_1_2_4_to_8 in libpng

set_gray_to_rgb

    $png->set_gray_to_rgb ();

Convert a grayscale PNG file to RGB format on reading.

To test whether your version of libpng supports set_gray_to_rgb, use "libpng_supports" with argument '"READ_GRAY_TO_RGB"':

    if (libpng_supports ('READ_GRAY_TO_RGB')) {
        # do something
    }

This function corresponds to png_set_gray_to_rgb in libpng

set_invert_alpha

    $png->set_invert_alpha ();

⛐🤪⚠ Untested. See the libpng manual for details.

To test whether your version of libpng supports set_invert_alpha, use "libpng_supports" with argument '"READ_INVERT_ALPHA"':

    if (libpng_supports ('READ_INVERT_ALPHA')) {
        # do something
    }

This function corresponds to png_set_invert_alpha in libpng

set_invert_mono

    $png->set_invert_mono ();

Invert monochrome images. See the libpng manual for details.

To test whether your version of libpng supports set_invert_mono, use "libpng_supports" with argument '"READ_INVERT"':

    if (libpng_supports ('READ_INVERT')) {
        # do something
    }

This function corresponds to png_set_invert_mono in libpng

set_packing

    $png->set_packing ();

When reading a PNG, expand the image to 1 pixel per byte for bit-depths 1, 2 and 4 without changing the order of the pixels. If this is not called, the pixels of bit_depths 1, 2 and 4 are packed into bytes as small as possible, for example, 8 pixels per byte for 1-bit files. See the libpng manual for details. See also "set_packswap".

To test whether your version of libpng supports set_packing, for a read object, use "libpng_supports" with argument '"READ_PACK"'.

    if (libpng_supports ('READ_PACK')) {
        # do something
    }

For a write object, use argument '"WRITE_PACK"'.

This function corresponds to png_set_packing in libpng

set_packswap

    $png->set_packswap ();

Swaps the bits of 1, 2, or 4 bits/pixel images. The default for PNG image is to put the left pixel into the highest part of the byte, so for example if the bit depth is 1, the pixels from left to right go into a byte as values 128 for the leftmost, 64, 32, 16, 8, 4, 2, then 1 for the rightmost pixel. This call reverses that behavior so that the pixels in a byte go 1 for the leftmost, 2, ..., then 128 for the rightmost. See the libpng manual for details.

To test whether your version of libpng supports set_packswap, for a read object, use "libpng_supports" with argument '"READ_PACKSWAP"'.

    if (libpng_supports ('READ_PACKSWAP')) {
        # do something
    }

For a write object, use argument '"WRITE_PACKSWAP"'.

This function corresponds to png_set_packswap in libpng

set_palette_to_rgb

    $png->set_palette_to_rgb ();

⛐🤪⚠ Untested. See the libpng manual for details.

To test whether your version of libpng supports set_palette_to_rgb, use "libpng_supports" with argument '"READ_EXPAND"':

    if (libpng_supports ('READ_EXPAND')) {
        # do something
    }

This function corresponds to png_set_palette_to_rgb in libpng

set_scale_16

    $png->set_scale_16 ()

⛐🤪⚠ Untested.

To test whether your version of libpng supports set_scale_16, use "libpng_supports" with argument '"READ_SCALE_16_TO_8"':

    if (libpng_supports ('READ_SCALE_16_TO_8')) {
        # do something
    }

This function corresponds to png_set_scale_16 in libpng

set_strip_16

    $png->set_strip_16 ();

Strip the pixels of a PNG stream with 16 bits per channel to 8 bits per channel.

To test whether your version of libpng supports set_strip_16, use "libpng_supports" with argument '"READ_STRIP_16_TO_8"':

    if (libpng_supports ('READ_STRIP_16_TO_8')) {
        # do something
    }

This function corresponds to png_set_strip_16 in libpng

set_strip_alpha

    $png->set_strip_alpha ();

Strip the alpha channel of a PNG stream.

To test whether your version of libpng supports set_strip_alpha, use "libpng_supports" with argument '"READ_STRIP_ALPHA"':

    if (libpng_supports ('READ_STRIP_ALPHA')) {
        # do something
    }

This function corresponds to png_set_strip_alpha in libpng

set_swap

    $png->set_swap ();

⛐🤪⚠ Untested. Swaps around the bytes of 16 bits/pixel images. See the libpng manual for details.

To test whether your version of libpng supports set_swap, use "libpng_supports" with argument '"READ_SWAP"':

    if (libpng_supports ('READ_SWAP')) {
        # do something
    }

This function corresponds to png_set_swap in libpng

set_swap_alpha

    $png->set_swap_alpha ();

⛐🤪⚠ Untested. See the libpng manual for details.

To test whether your version of libpng supports set_swap_alpha, use "libpng_supports" with argument '"READ_SWAP_ALPHA"':

    if (libpng_supports ('READ_SWAP_ALPHA')) {
        # do something
    }

This function corresponds to png_set_swap_alpha in libpng

set_tRNS_to_alpha

    $png->set_tRNS_to_alpha

⛐🤪⚠ Untested. See the libpng manual for details.

To test whether your version of libpng supports set_tRNS_to_alpha, use "libpng_supports" with argument '"READ_EXPAND"':

    if (libpng_supports ('READ_EXPAND')) {
        # do something
    }

This function corresponds to png_set_tRNS_to_alpha in libpng

Other libpng functions

get_chunk_cache_max

    my $max = $png->get_chunk_cache_max ();

Get the maximum number of ancilliary chunks allowed. See also "set_chunk_cache_max".

To test whether your version of libpng supports get_chunk_cache_max, use "libpng_supports" with argument '"CHUNK_CACHE_MAX"':

    if (libpng_supports ('CHUNK_CACHE_MAX')) {
        # do something
    }

This function corresponds to png_png_get_chunk_cache_max in libpng

get_rgb_to_gray_status

    my $badpixels = $png->get_rgb_to_gray_status ();

Returns a true value if there were some non-gray pixels in an RGB image after calling "set_rgb_to_gray".

Saving bandwidth

Try this one weird old get_rgb_to_gray_status trick to check whether an image marked as RGB is really monochrome or not.

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng ':all';
    use Image::PNG::Const ':all';
    
    my $file = 'life.png';
    print "Size before: ", -s $file, "\n";
    my $png = create_reader ($file);
    $png->read_info ();
    $png->set_rgb_to_gray ();
    if ($png->get_rgb_to_gray_status ()) {
        print "The image contained non-gray pixels.\n";
    }
    else {
        print "The image was grayscale already.\n";
    }
    $png->read_image ();
    $png->read_end ();
    my $wpng = $png->copy_png ();
    my $ihdr = $wpng->get_IHDR ();
    $ihdr->{color_type}  = PNG_COLOR_TYPE_GRAY;
    $wpng->set_IHDR ($ihdr);
    my $after = "life-gray.png";
    $wpng->write_png_file ($after);
    print "Size after: ", -s $after, "\n";
    
    if (! png_compare ($file, $after)) {
        print "The two files contain exactly the same image data.\n";
    }
    else {
        print "The two files contain different image data.\n";
    }

(This example is included as examples/xkcd.pl in the distribution.)

The output looks like this:

    Size before: 97299
    The image was grayscale already.
    Size after: 46956
    The two files contain exactly the same image data.

examples/life-gray.png is included in the distribution.

The resulting image looks like this:

examples/life-gray.png

Correct use of Image::PNG::Libpng could have spared as many as 50,000 bytes from the indignity of being sent around the internet.

This also illustrates the use of "png_compare".

Original image, licence statement and copyright notice here.

get_user_height_max

    $png->get_user_height_max ();

Get the maximum height allowed for a PNG file. The default is 1 million pixels. These values can be changed by "set_user_limits".

This function corresponds to png_get_user_height_max in libpng

get_user_width_max

    $png->get_user_width_max ();

Get the maximum width allowed for a PNG file. The default is 1 million pixels. These values can be changed by "set_user_limits".

This function corresponds to png_get_user_width_max in libpng

permit_mng_features

    $png->permit_mng_features ($mask);

⛐🤪⚠ Untested. $mask is an integer containing flags. See the libpng manual for details.

(MNG, 'Multiple-image Network Graphics', is an image animation format related to PNG. It has not been widely adopted.)

To test whether your version of libpng supports permit_mng_features, use "libpng_supports" with argument '"MNG_FEATURES"':

    if (libpng_supports ('MNG_FEATURES')) {
        # do something
    }

This function corresponds to png_permit_mng_features in libpng

set_add_alpha

    $png->set_add_alpha ($filler, $filler_loc);

⛐🤪⚠ Untested. Change $png to add an alpha channel. This only works for grayscale or RGB images with bit depth of 8 or 16. $filler contains the alpha value to assign to each pixel, and $filler_loc is either PNG_FILLER_BEFORE or PNG_FILLER_AFTER. The function of set_add_alpha is "set_filler" and a change of the color type to add an alpha channel. See the libpng manual for details.

To test whether your version of libpng supports set_add_alpha, use "libpng_supports" with argument '"READ_FILLER"':

    if (libpng_supports ('READ_FILLER')) {
        # do something
    }

This function corresponds to png_set_add_alpha in libpng

set_alpha_mode

    $png->set_alpha_mode ($mode, $screen_gamma);

⛐🤪⚠ Untested. $mode is an integer, one of the PNG_ALPHA_* constants. $screen_gamma is a floating point number. See the libpng manual for details.

To test whether your version of libpng supports set_alpha_mode, use "libpng_supports" with argument '"READ_ALPHA_MODE"':

    if (libpng_supports ('READ_ALPHA_MODE')) {
        # do something
    }

This function corresponds to png_set_alpha_mode in libpng

set_background

    $png->set_background (\%color, $bkgd_gamma_code, $need_expand, $gamma);

set_background sets the background of an image with an alpha channel or simple transparency (a "tRNS" chunk) with the color specified by %color. If $bkgd_gamma_code is set to PNG_BACKGROUND_GAMMA_SCREEN, it indicates that the supplied background color is in the gamma space of the display, else if it is set to PNG_BACKGROUND_GAMMA_FILE, the color is in the gamma space of the file. If $bkgd_gamma_code is set to PNG_BACKGROUND_GAMMA_UNIQUE, the value of $gamma is used, otherwise $gamma appears to be ignored and can be omitted. (This is not documented in the libpng manual or elsewhere. See https://github.com/glennrp/libpng/issues/358.)

If the background color is supplied at the original bit-depth for a grayscale image that is expanded to truecolor or to a higher bit-depth, $need_expand must be set to a true value, but if the background color is supplied at the expanded bit-depth, $need_expand must be set to a false value. Similarly for paletted images, if %color is supplied as a palette index ($color{index}), $need_expand must be set to a true value, but if %color is an RGB triplet, $need_expand must be set to a false value.

See "Setting the background" for an example.

To test whether your version of libpng supports set_background, use "libpng_supports" with argument '"READ_BACKGROUND"':

    if (libpng_supports ('READ_BACKGROUND')) {
        # do something
    }

set_chunk_cache_max

    $png->set_chunk_cache_max (42);

Set the maximum number of ancilliary chunks allowed. The default is 1000. See also "get_chunk_cache_max".

To test whether your version of libpng supports set_chunk_cache_max, use "libpng_supports" with argument '"CHUNK_CACHE_MAX"':

    if (libpng_supports ('CHUNK_CACHE_MAX')) {
        # do something
    }

This function corresponds to png_png_set_chunk_cache_max in libpng

set_filler

    $png->set_filler ($filler, $flags);

Set transformations in $png such that a filler byte $filler is added when an 8-bit grayscale image or 24-bit RGB image is read, and a filler byte is deleted when an 8-bit grayscale image or 24-bit RGB image is written. $flags may be PNG_FILLER_BEFORE or PNG_FILLER_AFTER. An error is produced if $png has bit depth less than 8.

This does not change the color type of the image. See the related function "set_add_alpha" if you want to add an alpha channel.

To test whether your version of libpng supports set_filler, use "libpng_supports" with argument '"READ_FILLER"':

    if (libpng_supports ('READ_FILLER')) {
        # do something
    }

This function corresponds to png_set_filler in libpng

set_gamma

    $png->set_gamma ($screen_gamma, $output_gamma);

⛐🤪⚠ Untested. $screen_gamma and $output_gamma should contain floating-point arguments. See the libpng manual for details.

To test whether your version of libpng supports set_gamma, use "libpng_supports" with argument '"READ_GAMMA"':

    if (libpng_supports ('READ_GAMMA')) {
        # do something
    }

This function corresponds to png_set_gamma in libpng

set_quantize

    $png->set_quantize (\@palette, $ncolors, \@histogram, $full_quantize);

The @palette part must be set to a palette similar to "set_PLTE". $ncolors must be the length of @palette or shorter. @histogram can be an empty array or if not it needs to be an array of integers of exactly the same length as @palette. The integers represent the frequency of the colors in @palette. These integers can range from 0 to 65355. Larger values or negative values cause an error. The final argument, $full_quantize, should be 1 for RGB images or 0 for paletted images.

Experiments in quantization

This example program experiments with two ways to quantize a PNG image, first of all using a completely random palette, and second using colors picked from the image at random:

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng ':all';
    my $file = "wave.png";
    my $ncolors = 40;
    my $palette = randompalette ($ncolors);
    write_with_palette ($file, $palette, $ncolors, [], "random");
    my $picked = points ($file, $ncolors);
    my @hist = (1) x $ncolors;
    write_with_palette ($file, $picked, $ncolors, \@hist, "picked");
    exit;
    
    sub write_with_palette
    {
        my ($file, $palette, $ncolors, $hist, $name) = @_;
        my $rpng = create_reader ($file);
        $rpng->set_quantize ($palette, $ncolors, $hist, 1);
        $rpng->read_png ();
        my $wpng = copy_png ($rpng);
        $wpng->set_PLTE ($palette);
        $wpng->write_png_file ("$name-$file");
    }
    
    sub points
    {
        my ($pngfile, $ncolors) = @_;
        my $png = read_png_file ($pngfile);
        my $rows = $png->get_rows ();
        my $h = $png->height ();
        my $w = $png->width ();
        my $ch = $png->get_channels ();
        my @p;
        for (0..$ncolors-1) {
            my $x = int (rand ($w));
            my $y = int (rand ($h));
            my $row = $rows->[$y];
            my @pix = split ('', substr ($row, $x*$ch, $ch));
            push @p, {
                red => ord ($pix[0]),
                green => ord ($pix[1]),
                blue => ord ($pix[2]),
            };
        }
        return \@p;
    }
    
    sub randompalette
    {
        my ($n) = @_;
        my @p;
        for (0..$n-1) {
            my %color;
            for my $c (qw!red green blue!) {
                $color{$c} = int (rand (256))
            }
            push @p, \%color;
        }
        return \@p;
    }

(This example is included as examples/q.pl in the distribution.)

examples/wave.png is included in the distribution.

The input image is this:

examples/wave.png

examples/random-wave.png is included in the distribution.

Completely random colors gives this kind of output:

examples/random-wave.png

examples/picked-wave.png is included in the distribution.

Picked colors gives this kind of output:

examples/picked-wave.png

set_rgb_to_gray

    $png->set_rgb_to_gray ();

    $png->set_rgb_to_gray ($error_action, $red_weight, $green_weight);

Convert a PNG image from RGB (colored) to gray. See the libpng manual for details.

Without arguments, the libpng default values are used for $red_weight and $green_weight, and an error action of "none" is chosen. If the error action is set to a value of PNG_ERROR_ACTION_WARN or PNG_ERROR_ACTION_ERROR, pixels in the RGB image where the R, G and B values are not equal will cause a warning or error respectively. To get the default values for $red_weight or $green_weight, use any negative value. See the libpng manual for details.

If you choose PNG_ERROR_ACTION_NONE you can find out whether there were non-RGB pixels in your image using "get_rgb_to_gray_status".

This function corresponds to png_png_set_rgb_to_gray in libpng with $error_action having a default value of PNG_ERROR_ACTION_NONE, and $red_weight and $green_weight both set to negative values.

Convert a colored image to gray

    use utf8;
    use FindBin '$Bin';
    use Image::PNG::Libpng ':all';
    use Image::PNG::Const ':all';
    
    my $png = create_reader ("$Bin/luv.png");
    $png->read_info ();
    $png->set_rgb_to_gray ();
    $png->read_image ();
    $png->read_end ();
    my $wpng = $png->copy_png ();
    my $ihdr = $wpng->get_IHDR ();
    $ihdr->{color_type}  = PNG_COLOR_TYPE_GRAY_ALPHA;
    $wpng->set_IHDR ($ihdr);
    $wpng->write_png_file ("$Bin/grayface.png");

(This example is included as examples/rgb-to-gray.pl in the distribution.)

examples/grayface.png is included in the distribution.

This is the output:

examples/grayface.png

set_user_limits

    $png->set_user_limits ($width, $height);

This enables a user to override libpng's restrictions to one million pixels in width and one million pixels in height for a PNG image.

This function corresponds to png_set_user_limits in libpng

Other functions

This section contains other functions which don't correspond to anything in libpng itself, but which have been added to the module for utility.

any2gray8

    my $wpng = any2gray8 ("any.png");

Convert any type of PNG file whatsoever into a bit-depth 8 grayscale image without an alpha channel (PNG_COLOR_TYPE_GRAY). The return value is a writing structure as made by "create_write_struct".

Currently the "tRNS" chunk is ignored by this.

By default the alpha channel is set to either the value of the "bKGD" chunk if there is one, or white if not. You can set another background using the option bkgd:

    my $wpng = any2gray8 ('any.png', bkgd => {red => 255, green => 0, blue => 99});

This will of course be ignored if any.png does not contain an alpha channel. RGB to gray conversion is done using "set_rgb_to_gray" with the default values.

This function is not supported for versions of libpng earlier than 1.6.14.

color_type_channels

    my $channels = color_type_channels ('PNG_COLOR_TYPE_RGB_ALPHA');

Returns the number of channels (the number of components of one pixel) of the specified color type.

copy_row_pointers

    $png->copy_row_pointers ($row_pointers);

This allows XS routines to directly set the value of the row pointers for the PNG image. The memory is copied, so whatever is in the row pointers can be freed immediately after calling this. However, the image data it points to is not copied, so this needs to be valid until the PNG is written.

The Perl scalar $row_pointers should be set up something like the following (where rp is the C pointer):

    RETVAL = newSViv (PTR2IV (rp));

It's extracted from the Perl scalar using

    rp = INT2PTR (png_byte **, SvIV (row_pointers));

where row_pointers is the SV * corresponding to $row_pointers in the Perl script. See also "set_row_pointers" which does the same thing but takes ownership of the memory.

get_chunk

    my $timechunk = $png->get_chunk ('tIME');

Get the specified chunk. This produces a fatal error if asked for the IDAT (image data) chunk, use "get_rows" instead. Returns undef if asked for an unknown chunk name.

This is used by pnginspect to get the valid chunks. See also "set_chunk".

get_internals

    my ($png_struct, $png_info) = get_internals ($png);

This function returns the png_structp and png_infop contained in $png. The return value is a list containing the png_structp as the first argument and the png_infop as the second argument, wrapped up as references to objects of type Image::PNG::Libpng::png_struct and Image::PNG::Libpng::png_info.

To access the values of the pointers, use something like this:

    void access_png_internals (png, info)
        SV * png;
        SV * info;
    PREINIT:
        png_struct * cpng;
        png_info * cinfo;
    CODE:
        cpng = INT2PTR (png_struct *, SvIV (png));
        cinfo = INT2PTR (png_info *, SvIV (info));

An example exists in Image::PNG::Cairo.

get_tRNS_palette

    $png->get_tRNS_palette ();

This is not a libpng-equivalent function.

image_data_diff

    my $diff = image_data_diff ('a.png', 'b.png');

This returns the undefined value if the image data in a.png is the same as the image data in b.png. If the image data is different, it returns a readable text message describing the first difference found, for example the height is different, or row 0 of the image data is different, etc.

This function is used in testing this module. See also "png_compare".

libpng_supports

    if (libpng_supports ('iTXt')) {
        print "Your libpng supports international text.\n";
    }

This function returns true or false depending on whether the version of libpng which this was compiled with supports or does not support a particular facility. For the most part, these are taken directly from the C macros of libpng.

The possible arguments to libpng_supports are:

16BIT
ALIGNED_MEMORY
ARM_NEON_API
BENIGN_ERRORS
BENIGN_READ_ERRORS
BENIGN_WRITE_ERRORS
bKGD

Does the libpng support the "bKGD" chunk?

BUILD_GRAYSCALE_PALETTE
BUILTIN_BSWAP16
CHECK_FOR_INVALID_INDEX
cHRM

Does the libpng support the "cHRM" chunk?

cHRM_XYZ

This is local to Image::PNG::Libpng and not a part of libpng itself.

It was necessary to extend libpng because the conditional compilation macro for "set_cHRM_XYZ" and "get_cHRM_XYZ", PNG_cHRM_SUPPORTED is defined (true) for old libpngs which do not actually contain these functions.

CHUNK_CACHE_MAX

This is local to Image::PNG::Libpng and not a part of libpng itself.

It was necessary to extend libpng because the conditional compilation macro for "set_chunk_cache_max" and "get_chunk_cache_max", PNG_USER_LIMITS_SUPPORTED is defined (true) for old libpngs which do not actually contain these functions.

CHUNK_MALLOC_MAX

This is local to Image::PNG::Libpng and not a part of libpng itself.

The reasons for this are identical to those for "CHUNK_CACHE_MAX".

COLORSPACE
CONSOLE_IO
CONVERT_tIME

This is related to two deprecated functions, "png_convert_from_time_t" and "png_convert_from_struct_tm".

CONVERT_tIME
EASY_ACCESS
ERROR_NUMBERS
ERROR_TEXT
eXIf

Does the libpng support the "eXIf" chunk?

FIXED_POINT
FIXED_POINT_MACRO
FLOATING_ARITHMETIC
FLOATING_POINT
FORMAT_AFIRST
FORMAT_BGR
gAMA

Does the libpng support the "gAMA" chunk?

GAMMA
GET_PALETTE_MAX

Does your libpng support "get_palette_max"?

HANDLE_AS_UNKNOWN
HANDLE_AS_UNKNOWN
hIST

Does the libpng support the "hIST" chunk?

iCCP

Does the libpng support the "iCCP" chunk?

INCH_CONVERSIONS
INFO_IMAGE
IO_STATE
iTXt

Does the libpng support international text?

MIPS_MSA_API
MNG_FEATURES
oFFs

Does the libpng support the "oFFs" chunk?

pCAL

Does the libpng support the "pCAL" extension?

PEDANTIC_WARNINGS
pHYs

Does the libpng support the "pHYs" chunk?

POINTER_INDEXING
POWERPC_VSX_API
PROGRESSIVE_READ
READ

Can libpng read PNGs?

READ_16_TO_8
READ_ALPHA_MODE

See "set_alpha_mode".

READ_BACKGROUND

See "set_background".

READ_BGR

See "set_bgr".

READ_COMPOSITE_NODIV
READ_COMPRESSED_TEXT
READ_EXPAND
READ_EXPAND_16
READ_FILLER
READ_GAMMA

See "set_gamma".

READ_GRAY_TO_RGB
READ_INTERLACING
READ_INT_FUNCTIONS
READ_INVERT
READ_INVERT_ALPHA
READ_OPT_PLTE
READ_PACK
READ_PACKSWAP
READ_QUANTIZE
READ_RGB_TO_GRAY
READ_SCALE_16_TO_8
READ_SHIFT
READ_STRIP_16_TO_8
READ_STRIP_ALPHA
READ_SWAP
READ_SWAP_ALPHA
READ_tEXt
READ_TRANSFORMS
READ_USER_TRANSFORM
READ_zTXt
SAVE_INT_32
SAVE_UNKNOWN_CHUNKS
sBIT

Does the libpng support the "sBIT" chunk?

sCAL

Does the libpng support the "sCAL" extension? This actually tests for the presence of the get_sCAL_s/set_sCAL_s functions, so its behaviour is dependent on other factors for versions 1.2 and 1.4 of libpng.

SEQUENTIAL_READ
SETJMP
SET_OPTION
SET_UNKNOWN_CHUNKS
SET_USER_LIMITS
SIMPLIFIED_READ
SIMPLIFIED_READ_AFIRST
SIMPLIFIED_WRITE
SIMPLIFIED_WRITE_AFIRST
SIMPLIFIED_WRITE_BGR
SIMPLIFIED_WRITE_STDIO
sPLT

Does the libpng support "sPLT" chunks?

sRGB

Does the libpng support the "sRGB" chunk?

STDIO
STORE_UNKNOWN_CHUNKS
TEXT

Does the libpng support text?

tEXt

Does the libpng support tEXt chunks?

tIME

Does the libpng support the "tIME" chunk?

TIME_RFC1123
tRNS

Does the libpng support the "tRNS" chunk?

UNKNOWN_CHUNKS

Does the libpng support unknown chunks (see "Private chunks")?

USER_CHUNKS
USER_LIMITS

Does the libpng support "set_user_limits" and the related functions "get_user_width_max", and "get_user_height_max"?

USER_LIMITS
USER_MEM
USER_TRANSFORM_INFO
USER_TRANSFORM_PTR
WARNINGS
WRITE

Can libpng write pngs?

WRITE_BGR
WRITE_COMPRESSED_TEXT
WRITE_CUSTOMIZE_COMPRESSION

Does the libpng support "set_compression_level" and similar functions?

⛐🤪⚠ It's not very clear that this returns a useful value, since "set_compression_level" seems to be in libpngs from at least as far back as 1.5.1, and yet this macro was only added to libpng in version 1.6.13.

WRITE_CUSTOMIZE_ZTXT_COMPRESSION

Does the libpng support "set_text_compression_level" and similar functions?

WRITE_FILLER
WRITE_FILTER
WRITE_FLUSH
WRITE_FLUSH_AFTER_IEND
WRITE_INTERLACING
WRITE_INT_FUNCTIONS
WRITE_INVERT
WRITE_INVERT_ALPHA
WRITE_OPTIMIZE_CMF
WRITE_PACK
WRITE_PACKSWAP
WRITE_SHIFT
WRITE_SWAP
WRITE_SWAP_ALPHA
WRITE_TRANSFORMS
WRITE_USER_TRANSFORM
WRITE_WEIGHTED_FILTER
zTXt

Does the libpng support zTXt chunks?

png_compare

    if (png_compare ('a.png', 'b.png') == 0) {
        print "The PNGs are the same.\n";
    }

This compares the image data in two PNGs and returns 0 if they contain exactly the same image data, or 1 if they contain different image data. For a more detailed comparison, see "image_data_diff". This does not compare to see if the PNG files "look like" each other, but whether each pixel contains exactly the same values. Please see Image::Similar for a looser comparison of images.

This uses the "expand" transform of libpng to read both the images, so it is able to compare images of different color types. It compares the alpha values as well as the color pixels. See "Saving bandwidth" for an example of comparing an RGB and a grayscale image.

read_struct

    if ($png->read_struct ()) {
         warn "Can't write this, use copy_png to copy it";
    }

This returns a true value if $png was created with "create_read_struct" or allied functions like "read_from_scalar", and a false value if $png was created with "create_write_struct" or allied functions like "copy_png".

set_chunk

    $png->set_chunk ('tIME', $timechunk);

Set the specified chunk. This produces a fatal error if given an IDAT (image data) chunk, use "set_rows" instead. This produces a fatal error if given an unknown chunk name.

The first argument is the chunk name and then the second argument is a scalar containing whatever the chunk requires, for example a hash reference for the tIME chunk as described under "set_tIME" and "get_tIME".

This is essentially a convenience function for the benefit of "copy_png" which, together with "get_valid", enables the PNG chunks to be copied in a loop rather than one-by-one.

set_image_data

    $png->set_image_data ($image_data);

Set the internal image data pointer to $image_data. $image_data should contain a pointer to memory stored as an SvIV allocated with Newx or a similar function. This transfers ownership of the memory to $png, which will free it with Safefree when $png is destroyed. Calling this function with any value does not actually change the content of the PNG image itself.

set_row_pointers

    $png->set_row_pointers ($row_pointers);

This sets the rows of the PNG image to $row_pointers using png_set_rows. $row_pointers must contain a pointer to memory stored as an SvIV allocated with a Perl memory allocator like Newx or a similar function. This also transfers ownership of the memory to $png, which will free it with Safefree when $png is destroyed. See also "copy_row_pointers", which does the same thing except for the freeing of the memory.

split_alpha

    my $split = $png->split_alpha ();
    my $alpha = $split->{alpha};
    my $color = $split->{data};

Split the alpha channel away from the other data. This only works for images with color types PNG_COLOR_TYPE_RGB_ALPHA and PNG_COLOR_TYPE_GRAY_ALPHA. This is not part of the libpng API. This function was added to this module to assist the author of PDF::Builder due to problems with very slow access to PNG images using Perl.

To remove the alpha channel from an image, use "set_background". See "Setting the background" for an example. You can also use the transform PNG_TRANSFORM_STRIP_ALPHA in "read_png_file", or "set_strip_alpha", but this may leave spurious pixel values in transparent parts of the image.

EXPORTS

Nothing is exported by default, but all the functions in this module, including the object methods, can be exported on request. The export tag 'all' exports everything in the module:

    use Image::PNG::Libpng ':all';

This includes all the methods, which can then be used with the $png argument as the first argument.

Differences from libpng

The functions in Image::PNG::Libpng are closely based on those of libpng, with the following differences.

No info structure

This module does not use the "info" structure of libpng. Almost all libpng functions require two initial arguments, a png_structp and a png_infop. However, in Image::PNG::Libpng, both the "png" and the "info" are contained in the object. People who need access to the internals of an Image::PNG::Libpng object (C programmers or XS programmers) can use "get_internals" to access them.

This module does not support the "end info" structure of PNGs when writing, although these are handled when reading.

Unused arguments omitted

This module eliminates unevaluated arguments of libpng. For example, libpng requires the user to pass a pointer to a png_struct to call the libpng version number function, (see "get_libpng_ver"), but it actually ignores this structure. There are many similar instances of unevaluated arguments, which have all been eliminated from this module.

If you are interested in exactly which libpng arguments are omitted, you can find each instance in the file perl-libpng.c in the top directory of the distribution in the macro UNUSED_ZERO_ARG.

Useless functions omitted

The functions from libpng which don't do anything have been omitted from this module.

Function return values are used to return values

libpng is very inconsistent in its calling conventions. Some functions return results using references, and some return results using the function's return value. For example png_get_rows (see "get_rows") uses the return value of the function to return an array of pointers, but png_get_PLTE (see "get_PLTE") uses a pointer reference to return an array of pointers, and the return value to indicate errors.

Image::PNG::Libpng uses only the return value. Errors and non-existence are indicated by a return value of the undefined value.

Further to this, libpng's error handling is also very inconsistent. Some functions use the return value to indicate errors, and some of the functions don't indicate errors at all, but just fail silently. Even more inconsistently, some of the functions which use the return value to indicate an error use a non-zero value, and some use a zero value, to indicate an error.

No destructors

Freeing the memory allocated by "create_read_struct" and "create_write_struct" is automatically handled by Perl.

Older versions of this module (pre-0.18) had functions called destroy_read_struct and destroy_write_struct corresponding to the functions with similar names in libpng. From version 0.18, these functions still exist, but they no longer do anything. The memory freeing is now handled by Perl automatically.

Other unimplemented parts of libpng

Memory management functions

The management of memory for libpng objects is handled automatically by this module, and so it does not offer an interface to png_malloc and png_free or other libpng memory handling functions.

Error handling functions

This module does not offer an interface to png_error and png_get_error_ptr or any of the other error handling functions of libpng. It redirects the error and warning handlers to Perl's error stream.

Input/output manipulation functions

This module does not offer a direct interface to png_set_write_fn and png_set_read_fn. However, it is possible to use their functionality to access Perl data via "read_from_scalar" and "write_to_scalar".

Fixed point functions

There is currently no support for the PNG fixed point functions in this Perl module, the functions with suffix _fixed.

List of unsupported functions

The following functions from the libpng API are unsupported by this module.

png_benign_error

💔 See "Error handling functions".

png_build_grayscale_palette

⛐🤪⚠ Undocumented function, but it is part of the public API. See https://github.com/glennrp/libpng/issues/353.

This function builds an evenly-spaced grayscale palette at a specified bit depth into a user-supplied array. It is not used elsewhere in libpng.

png_chunk_benign_error

💔 See "Error handling functions".

png_chunk_warning

💔 See "Error handling functions".

png_convert_from_struct_tm

👎 Deprecated in libpng. This function related to the tIME chunk is deprecated in libpng 1.7.

png_convert_from_time_t

👎 Deprecated in libpng. This function related to the tIME chunk is deprecated in libpng 1.7.

png_convert_to_rfc1123

👎 Deprecated in libpng. This function related to the tIME chunk is deprecated in libpng 1.7.

png_convert_to_rfc1123_buffer

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

This function is related to the tIME chunk; it seems fairly superfluous with the many other ways to manipulate time strings already in Perl.

png_data_freer

🐘 See "Memory management functions".

png_free

🐘 See "Memory management functions".

png_free_data

🐘 See "Memory management functions".

png_get_cHRM_XYZ_fixed

🍓 See "Fixed point functions".

png_get_cHRM_fixed

🍓 See "Fixed point functions".

png_get_compression_type

🙃🤡 See "Useless functions omitted". Useless function which returns 0 since there is only one type of compression.

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

This function gives you the copyright string for libpng.

png_get_current_pass_number
png_get_current_row_number
png_get_error_ptr

💔 See "Error handling functions".

png_get_filter_type

🙃🤡 See "Useless functions omitted". This function returns the value of the unused field filter_type in the PNG header, which is always 0. See "Unused arguments omitted".

png_get_gAMA_fixed

🍓 See "Fixed point functions".

png_get_header_ver

This function was omitted from this module because it merely duplicates another function.

Identical to "get_libpng_ver".

png_get_header_version

This function was omitted from this module because it merely duplicates another function.

Identical to "get_libpng_ver".

png_get_io_chunk_type
png_get_io_ptr
png_get_io_state
png_get_mem_ptr
png_get_pHYs_dpi
png_get_pixel_aspect_ratio
png_get_pixel_aspect_ratio_fixed

🍓 See "Fixed point functions".

png_get_pixels_per_inch
png_get_pixels_per_meter
png_get_progressive_ptr

Incremental reading of PNGs is not handled yet.

png_get_sCAL_fixed

🍓 See "Fixed point functions".

png_get_signature

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

Returns the PNG signature of a PNG you read in.

png_get_uint_31

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

This produces an error if the unsigned integer argument is too big for a 31 bit number.

png_get_user_chunk_ptr
png_get_user_height_max
png_get_user_transform_ptr
png_get_x_offset_inches
png_get_x_offset_inches_fixed

🍓 See "Fixed point functions".

png_get_x_offset_microns
png_get_x_offset_pixels

Duplicates "get_oFFs".

png_get_x_pixels_per_inch
png_get_x_pixels_per_meter

Duplicates "get_oFFs".

png_get_y_offset_inches
png_get_y_offset_inches_fixed

🍓 See "Fixed point functions".

png_get_y_offset_microns
png_get_y_offset_pixels

Duplicates "get_oFFs".

png_get_y_pixels_per_inch
png_get_y_pixels_per_meter

Duplicates "get_oFFs".

png_handle_as_unknown
png_image_begin_read_from_file
png_image_begin_read_from_stdio
png_image_finish_read
png_image_free

🐘 See "Memory management functions".

png_image_write_to_file
png_image_write_to_memory

🐘 See "Memory management functions".

png_image_write_to_stdio
png_process_data
png_process_data_pause
png_process_data_skip
png_progressive_combine_row
png_read_row
png_read_rows
png_reset_zstream

👎 Deprecated in libpng. Resets the zstream of the zlib instance used for the image data.

png_save_int_32

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

Writes a 32 bit signed number into an octet buffer. Perl programmers will probably use pack for this.

png_save_uint_16

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

Writes a 16 bit number into an octet buffer. Perl programmers will probably use pack for this.

png_save_uint_32

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

Writes a 32 bit unsigned number into an octet buffer. Perl programmers will probably use pack for this.

png_set_alpha_mode_fixed

🍓 See "Fixed point functions".

png_set_background_fixed

🍓 See "Fixed point functions".

png_set_benign_errors

💔 See "Error handling functions".

png_set_cHRM_XYZ_fixed

🍓 See "Fixed point functions".

png_set_cHRM_fixed

🍓 See "Fixed point functions".

png_set_check_for_invalid_index
png_set_compression_method

🙃🤡 See "Useless functions omitted". This function in libpng corresponds to the unused method parameter of zlib functions like deflateInit2. See "zlib documentation". The libpng function just produces a warning if the user sets the value to anything but 8, the value of the macro Z_DEFLATED, and then lets zlib produce an error.

png_set_crc_action

💔 See "Error handling functions".

png_set_error_fn

💔 See "Error handling functions".

png_set_filter_heuristics

👎 Deprecated in libpng.

png_set_filter_heuristics_fixed

🍓 See "Fixed point functions".

png_set_flush
png_set_gAMA_fixed

🍓 See "Fixed point functions".

png_set_gamma_fixed

🍓 See "Fixed point functions".

png_set_interlace_handling

Incremental writing is not handled.

png_set_invalid

🐘 See "Memory management functions".

png_set_longjmp_fn

⛐🤪⚠ Undocumented function, but it is part of the public API.

png_set_mem_fn

🐘 See "Memory management functions".

png_set_option
png_set_progressive_read_fn
png_set_read_status_fn
png_set_read_user_chunk_fn
png_set_read_user_transform_fn
png_set_sCAL_fixed

🍓 See "Fixed point functions".

png_set_sRGB_gAMA_and_cHRM
png_set_shift
png_set_sig_bytes

👎🐪 This function was deliberately omitted from this module because it doesn't seem useful for Perl programmers.

png_set_strip_error_numbers

🙃🤡 See "Useless functions omitted". According to the libpng manual, "we never got around to actually numbering the error messages", so I assume this is not very useful.

png_set_text_compression_method

🙃🤡 See "Useless functions omitted". Unsupported for the same reasons as "png_set_compression_method".

png_set_unknown_chunk_location

This is related to some kind of bug in version 1.5 and previous of libpng.

png_set_user_transform_info
png_set_write_status_fn

💔 See "Error handling functions".

png_set_write_user_transform_fn
png_start_read_image
png_write_chunk
png_write_chunk_data
png_write_chunk_end
png_write_chunk_start
png_write_flush
png_write_info_before_PLTE
png_write_row
png_write_rows
png_write_sig

This is an exhaustive list of unsupported libpng API functions, extracted from the libpng source code using util/api.pl in this module's repository.

Conditional compilation and old libpng versions

It is possible to compile a version of the libpng library without support for various things. For example, a libpng without support for text chunks may be created by undefining the C macro PNG_TEXT_SUPPORTED. This module supports the conditional compilation choices which we're aware of, but if you encounter problems using this module because of a conditionally-compiled libpng, then let us know and we'll add the necessary facility.

Currently we test this module against vanilla libpng versions 1.2.59, 1.4.22, 1.5.30, and 1.6.37 before releases. There doesn't seem to have been a 1.3 or a 1.1 version of libpng. See http://www.libpng.org/pub/png/pngnews.html. Versions 1.0 and earlier of libpng are not tested against.

The following issues with older libpng versions may affect users of this module:

cHRM chunk

The "get_cHRM_XYZ" and "set_cHRM_XYZ" functions for the "cHRM" chunk are not present in pre-1.5.5 versions of libpng, but the conditional compilation macro of libpng for these newer functions is the same as for "get_cHRM" and "set_cHRM" functions, which are present on older versions of libpng. Because of this, this module includes its own protection macro, accessible via "libpng_supports" as "cHRM_XYZ". These functions are disabled for libpng versions earlier than 1.5.5.

chunk_cache_max

The functions "set_chunk_cache_max" and "get_chunk_cache_max" are not protected by a libpng macro, so this module includes its own protection macro, accessible via "libpng_supports" as "CHUNK_CACHE_MAX". These functions are disabled for libpng versions earlier than 1.4.0.

Compression level

Versions of libpng up to 1.5 behave erratically when "set_compression_level" or other set_compression_* functions are used to alter the compression of the image data. Testing of compression-related functions is disabled by this Perl module for pre-1.6.13 versions of libpng.

sCAL chunk

Support for the sCAL chunk is disabled within this module for pre-1.6 versions of libpng.

Text chunk handling

Versions of libpng up to 1.6.3 produce erratic results with iTXt (international text) chunks. Testing of text-related chunks is disabled by this Perl module for pre-1.6.13 versions of libpng.

Not all constants are available

Some of the libpng constants are defined in pnglibconf.h or pngconf.h but Image::PNG::Const only looks at png.h to make its constants. Because of this, some constant values like PNG_Z_DEFAULT_COMPRESSION aren't currently available in Image::PNG::Const. This probably should be fixed to extract the constants from the other files in a future version.

STANDALONE SCRIPT

A standalone script, pnginspect, is installed with the distribution. It prints out the contents of the chunks of the PNG file on the command line.

SEE ALSO

The PNG specification

The PNG specification (link to W3 consortium) explains the details of the PNG format.

PNG The Definitive Guide by Greg Roelofs

The book "PNG - The Definitive Guide" by Greg Roelofs, published in 1999 by O'Reilly is available online at http://www.faqs.org/docs/png/ or http://www.libpng.org/pub/png/book/.

The libpng documentation

Official documentation

The starting point is the plain text libpng manual at http://libpng.org/pub/png/libpng-manual.txt and the manual page libpng.3, which you can read using "man 3 libpng".

⛐🤪⚠ The documentation which comes with libpng is rather sketchy. See "Differences from libpng". It doesn't contain full specifications (prototypes, return values) for all of the functions in the library. For programming in C using libpng, look at the header file png.h. In some cases, you need to look at the source code of the library.

Other documentation

There is a collection of function definitions under the title "Interface Definitions for libpng12" at https://refspecs.linuxbase.org/LSB_4.0.0/LSB-Desktop-generic/LSB-Desktop-generic/libpng12man.html as part of the "Linux Standard Base Desktop Specification". These contain extensive information on the prototypes and return values for the libpng routines, something which is often only available elsewhere by actually looking at the libpng source code. These pages are usually the first hits on search engines if you search for a function name in libpng.

zlib documentation

See https://zlib.net/manual.html for the zlib documentation.

Other Perl modules on CPAN

These other modules may also be useful.

Alien::PNG

Alien::PNG claims to be a way of "building, finding and using PNG binaries". It may help in installing libpng. We didn't use it as a dependency for this module because it seems not to work in batch mode, but stop and prompt the user. We're interested in hearing feedback from users whether this works or not on various platforms.

Imager

Imager, Imager::Files and Imager::Files::PNG contain support for reading and writing PNGs via libpng, as well as support for reading and writing various other kinds of image files, changing the images, converting, and more.

Image::ExifTool

Image::ExifTool is a pure Perl (doesn't require a C compiler) solution for accessing the text segments of images. It supports PNG text segments.

Image::Info

Image::Info gets information out of images. It supports PNG and is written in pure Perl, so it doesn't require a C compiler. As well as basics such as height, width, and colour type, it can get text chunks, modification time, palette, gamma (gAMA chunk), resolution (pHYs chunk), and significant bits (sBIT chunk). At the time of writing (version 1.31) it doesn't support other chunks.

Image::PNG::Rewriter

Image::PNG::Rewriter is a utility for unpacking and recompressing the IDAT (image data) part of a PNG image. The main purpose seems to be to recompress the image data with the module author's other module Compress::Deflate7. At the time of writing, that only works with Perl versions 5.12 or later.

Image::Pngslimmer

Image::Pngslimmer reduces the size of dynamically created PNG images. It's very, very slow at reading PNG data, but seems to work OK.

Image::PNG::Write::BW

Image::PNG::Write::BW writes black and white PNGs from strings.

Image::Size

If you only need to read the sizes of images, Image::Size works with PNG and other image formats.

AUTHOR

Ben Bullock, <bkb@cpan.org>

COPYRIGHT & LICENCE

This package and associated files are copyright (C) 2011-2020 Ben Bullock.

You can use, copy, modify and redistribute this package and associated files under the Perl Artistic Licence or the GNU General Public Licence.