You're going to write a JSON parser yourself. Without a parser, it's very hard to guess whether a newline should go after a comma - it might be part of a string or key, similarly brackets might appear in strings and commas. A double quote closes a string, but not if it's preceded by a backslash, unless the backslash is preceded by another backslash, etc.

Good luck.

Since you're using JSON you can it to pretty print JSON. FWIW I use Mojo/Mojo::UserAgent to call APIs and process the results.

Note that the following does not in any way encode, decode, parse or validate JSON syntax. It just naively pretty-prints JSON-ish stuff that is assumed valid. I'm sure there are modules that would do a much better, more reliable job of this, but you specify that, for some reason, you do not want to use any modules. (Update: Note in particular choroba's comments on this.)

Win8 Strawberry 5.8.9.5 (32)  Wed 09/07/2022  7:13:21
C:\@Work\Perl\monks
>perl

use strict;
use warnings;

use Data::Dump qw(dd);  # for debug

my $j = <<'EOJ';
"device": [{"host_info"  :  { "engine_id":null   ,
"name": null   ,"host_group_info" :{"name":null,"remarks":null
,
    "id" : 0 } }}
  ]
EOJ
print ">>\n$j<< \n";

my $dent   = '    ';
my $n_dent = 0;

my %pretty = (
    '{' => sub { my $d = $dent x ++$n_dent;  return qq{$_[0]\n$d}; },
    '}' => sub { my $d = $dent x --$n_dent;  return qq{\n$d$_[0]}; },
    ',' => sub { my $d = $dent x   $n_dent;  return qq{$_[0]\n$d}; },
    ':' => sub { my $d = $dent x   $n_dent;  return qq{$_[0] };    },
    );

$j =~ s{ \s* ([{},:]) \s* }
       { $pretty{$1}->($1) }xmsge;
print ">>\n$j<< \n";

^Z
>>
"device": [{"host_info"  :  { "engine_id":null   ,
"name": null   ,"host_group_info" :{"name":null,"remarks":null
,
    "id" : 0 } }}
  ]
<<
>>
"device": [{
    "host_info": {
        "engine_id": null,
        "name": null,
        "host_group_info": {
            "name": null,
            "remarks": null,
            "id": 0
        }
    }
}]
<<
[download]

Update: Here's a variation that handles double-quoted strings a little better. All the old caveats still apply.

Win8 Strawberry 5.30.3.1 (64)  Wed 09/07/2022  8:14:56
C:\@Work\Perl\monks
>perl

use 5.010;  # needs \K regex extension

use strict;
use warnings;

# use Data::Dump qw(dd);  # for debug

my $j = <<'EOJ';
"device\"": [{"host_in}fo"  :  { "eng{ine_id,":null   ,
"name:": null   ,"host_group_info\\" :{"na\"me":null,"\"remarks":null
,
    "id" : 0 } "{[:,"  :  "weird", "crazy"  :  "[{,:" }}
  ]
EOJ
print ">>\n$j<< \n";

use constant DENT => '    ';
my $n_dent = 0;

my %pretty = (
    '{' => sub { my $d = DENT x ++$n_dent;  return "$_[0]\n$d";     },
    '}' => sub { my $d = DENT x --$n_dent;  return "\n$d$_[0]\n$d"; },
    ',' => sub { my $d = DENT x   $n_dent;  return "$_[0]\n$d";     },
    ':' => sub { my $d = DENT x   $n_dent;  return "$_[0] ";        },
    );

my $rx_d_quote = qr{ " [^\\"]* (?: \\. [^\\"]*)* " }xms;

$j =~ s{ $rx_d_quote? \K \s* ([{},:]) \s* }
       { $pretty{$1}->($1) }xmsge;
print ">>\n$j<< \n";

^Z
>>
"device\"": [{"host_in}fo"  :  { "eng{ine_id,":null   ,
"name:": null   ,"host_group_info\\" :{"na\"me":null,"\"remarks":null
,
    "id" : 0 } "{[:,"  :  "weird", "crazy"  :  "[{,:" }}
  ]
<<
>>
"device\"": [{
    "host_in}fo": {
        "eng{ine_id,": null,
        "name:": null,
        "host_group_info\\": {
            "na\"me": null,
            "\"remarks": null,
            "id": 0
        }
        "{[:,": "weird",
        "crazy": "[{,:"
    }

}
]
<<
[download]

Yes, even you can use CPAN. Nuff said.

The issue here is, I have a working script using CPAN, yet my manager said its way too easy and I should try achive the same thing by using only regex

Well it's one thing to try this as an exercise, for most things significant effort has been spent on the non-trivial task of catering for everything when parsing, same goes for HTML, XML etc...

"As a learning" exercise is about the only valid reason to eschew CPAN I can think of. As a learning exercise it is especially useful if it serves to contrast the benefit of using CPAN compared to rolling your own solution. It is even more useful if the learning exercise also involves creating even a tiny fraction of the testing that CPAN modules are subject to.

If that is the objective of your manager and you have been given time to engage in this learning exercise then you have a truly enlightened manager!

Optimising for fewest key strokes only makes sense transmitting to Pluto or beyond

I have a working script using CPAN, yet my manager said its way too easy and I should try achive the same thing by using only regex

That's plain stupid and a waste of time. If your manager regularly behaves like this, search for another job. It's just not worth wasting your time this way. Been there, done than.

Imagine a carpenter's manager telling him to put way all tools except for a rock with a sharp edge. Sure, you could build a house with that rock. But using modern tools is less risky and way faster.

Alexander

--
Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)

NIH is a cancer. Unless you picked one of the few clinkers from the CPAN, and I don’t think there are any JSON modules that fit that particular bill. What afoken and Fletch said**10.