as a demo:

(NB: I changed your indentation levels to make it identical, otherwise please uncomment the whitespace eraser prior to testing. )

use warnings;
use strict;
use autodie;
BEGIN { $ENV{PERL_JSON_BACKEND}='JSON::PP'; }
use JSON;
use Test::More;

use 5.010;

my @order = qw/id disp ver auth/;



my $dat = [ ({ id => 'a', disp => 'a', ver => 'a', auth => 'a' })x2 ];
+ # dummy data

local $\ = "\n";

#print 'Unsorted => ', JSON->new->indent->space_after->encode( $dat );
#print 'Alpha Sorted => ', JSON->new->indent->space_after->canonical->
+encode( $dat );
print "Pryrt's Order => ",
    my $pryrt = manual_ordered_json( $dat, \@order );

# --- Sorting with JSON::PP
my %ORDER;
@ORDER{@order} = 1..@order;
my $sort = sub {
    package JSON::PP;
    ($ORDER{$a} // 1e999) <=> ($ORDER{$b} // 1e999)
        or
    $a cmp $b
};
print "LanX' Order => ",
    my $lanx = JSON->new->sort_by($sort)->indent->space_after->encode(
+ $dat );

# # --- erase whitespace diffs
# s/\s//g for $pryrt, $lanx;


is($lanx,$pryrt,"same JSON");

done_testing();

use Data::Dumper;

sub manual_ordered_json{
    my @list = @{$_[0]};
    my @ordr = @{$_[1]};
    my $out = "[\n";
    for my $i ( 0 .. $#list ) {
        my $h = $list[$i];
        $out .= "   {\n";
        for my $j ( 0 .. $#ordr ) {
            my $k = $ordr[$j];
            next unless defined $k;
            next unless exists $h->{$k};
            $out .= sprintf qq|      "%s": "%s"|, $k, $h->{$k} // '<un
+def>';
            $out .= ',' if $j < $#ordr;
            $out .= "\n";
        }
        $out .= "   }";
        $out .= "," if $i < $#list;
        $out .= "\n";
    }
    $out .= "]\n";
    return $out;
}
[download]

--->

Pryrt's Order => [
   {
      "id": "a",
      "disp": "a",
      "ver": "a",
      "auth": "a"
   },
   {
      "id": "a",
      "disp": "a",
      "ver": "a",
      "auth": "a"
   }
]

LanX' Order => [
   {
      "id": "a",
      "disp": "a",
      "ver": "a",
      "auth": "a"
   },
   {
      "id": "a",
      "disp": "a",
      "ver": "a",
      "auth": "a"
   }
]

ok 1 - same JSON
1..1
[download]

>> Are you aware of JSON::PP#sort_by?

Thanks! I knew that JSON used JSON::PP or JSON::XS under the hood, but I hadn't thought to go searching through the backend modules for features not advertised (edit: "...not advertised at the top level"). That's just what I was hoping would exist somewhere. And thanks for the example to go with it: works great for me.

> Thanks! ... And thanks for the example to go with it: works great for me.

Your welcome, I'm learning a lot from interesting questions.

The code is mostly adopted from the perldoc though and could be written in a more generic way.

I also like the fact that JSON::PP automatically detects hashes ordered via tie interface.

> I knew that JSON used JSON::PP or JSON::XS under the hood,

Not only under the hood, you can even use both modules in the same code, as long as you are restricting one of them to pure OO. =)

use warnings;
use strict;
use autodie;

use JSON::PP qw//;                      # don't export anything
use JSON;

use Test::More;

use 5.010;

my @order = qw/id disp ver auth/;



my $dat = [ ({ id => 'a', disp => 'a', ver => 'a', auth => 'a' })x2 ];
+ # dummy data

local $\ = "\n";

#print 'Unsorted => ', JSON->new->indent->space_after->encode( $dat );
#print 'Alpha Sorted => ', JSON->new->indent->space_after->canonical->
+encode( $dat );
print "Pryrt's Order => ",
    my $pryrt = manual_ordered_json( $dat, \@order );

# --- Sorting with JSON::PP
my %ORDER;
@ORDER{@order} = 1 .. @order;
my $sort = sub {
    package JSON::PP;
    ($ORDER{$a} // 1e999) <=> ($ORDER{$b} // 1e999)
        or
    $a cmp $b
    };
print "LanX' Order => ",
    my $lanx = JSON::PP->new->sort_by($sort)->indent->space_after->enc
+ode( $dat );

# # --- erase whitespace diffs
# s/\s//g for $pryrt, $lanx;

is($lanx,$pryrt,"same JSON");

done_testing();

use Data::Dumper;

sub manual_ordered_json{
    my @list = @{$_[0]};
    my @ordr = @{$_[1]};
    my $out = "[\n";
    for my $i ( 0 .. $#list ) {
        my $h = $list[$i];
        $out .= "   {\n";
        for my $j ( 0 .. $#ordr ) {
            my $k = $ordr[$j];
            next unless defined $k;
            next unless exists $h->{$k};
            $out .= sprintf qq|      "%s": "%s"|, $k, $h->{$k} // '<un
+def>';
            $out .= ',' if $j < $#ordr;
            $out .= "\n";
        }
        $out .= "   }";
        $out .= "," if $i < $#list;
        $out .= "\n";
    }
    $out .= "]\n";
    return $out;
}
[download]

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery FootballPerl is like chess, only without the dice}

Rest assured: JSON::XS supports the same feature.

Interesting. Since you resurrected this thread, my curiosity has been piqued, and I couldn't remember whether I'd tried to do the same with JSON::XS three years ago or not.

Searching through the JSON::XS codebase finds 0 instances of sort_by , so I am not sure what line of code in the source of JSON::XS could be implementing that function (though maybe there's an inheritance in the XS source, because I'm not sure what the equivalent of use/require/parent/base/@ISA are in XS). Because of that uncertainty, I took the working code that LanX had posted three years ago, and changed every instance of JSON::PP to JSON::XS . When I ran that, I got the message sort_by is not supported by JSON::XS. at C:\....\11111902.pl line 32. and the is() test failed, because the $lanx version was not sorted. Based on that experiment, I cannot see how to use sort_by with JSON::XS in an equivalent manner to how it's used with JSON::PP.

Could you show an example of JSON::XS using sort_by (using the same data and structure as in the working example that LanX posted earlier)? And if it's just called something other than sort_by in the JSON::XS version, please let me know what the right name is (and whether or not it's documented in JSON::XS's POD). Because if it is possible, I'd like to see how. Thanks.

I found this thread through Google because I recently had this problem.

When I looked in the docs of JSON::XS I found pretty much the the same paragraph as in the docs of JSON::PP

$json = $json->canonical ($enable)
$enabled = $json->get_canonical
If $enable is true (or missing), then the encode method will output JSON objects by sorting their keys. This is adding a comparatively high overhead.

Use of canonical fixed my problem... and it will work in both.