Rand Stats

JSON::Fast::Hyper

zef:lizmat

Actions Status

NAME

JSON::Fast::Hyper - Read/Write hyperable JSON

SYNOPSIS

use JSON::Fast::Hyper;

my $json = to-json(@array);    # creates faster hyperable JSON

my $json = to-json(%hash);     # normal JSON

my @array = from-json($json);  # 2x faster if hyperable JSON

my %hash := from-json($json);  # normal speed

DESCRIPTION

JSON::Fast::Hyper is a drop-in wrapper for JSON::Fast. It provides the same interface as JSON::Fast, with two differences:

CONVERTING A LIST TO JSON

my $json = to-json(@array);    # creates faster hyperable JSON

When a list is converted to JSON, it will create a special hyperable version of JSON that is still completely valid JSON for all JSON parsers, but which can be interpreted in parallel when reading back. To allow this feature, the JSON will always be created as if the :pretty named argument has been specified as False.

If your data structure is not a Positional, then it will just act as the normal to-json sub from JSON::Fast.

CONVERTING JSON OF A LIST TO DATA STRUCTURE

my @array := from-json($json);  # 2x faster if hyperable JSON

When the JSON was created hyperable, then reading such JSON with from-json will return the original Positional as a List. If the JSON was not hyperable, it will call the normal from-json sub from JSON::Fast.

TECHNICAL BACKGROUND

This module makes the to-json and from-json exported subs of JSON::Fast into a multi. It adds a candidate taking a Positional to to-json, and it adds a candidate to from-json that takes the specially formatted string created by that extra to-json candidate.

The Positional candidate of to-json will create the JSON for each of the elements separately, and joins them together with an additional newline. And then adds a specially formatted header and footer to the result. The resulting string is still valid JSON, readable by any JSON decoder. But when run through the from-json sub provided by this module, will decode elements in parallel. Wallclock times are at about 50% for a 7MB JSON file, such as provided by the Raku Ecosystem Archive. While only adding 3 + number of elements bytes to the resulting string.

A similar approach could be done for handling an Associative at the top level. But this makes generally a lot less sense, as the amount of information per key/value is usually vastly different, and JSON that consists of an Associative at the top, are usually not big enough to warrant the overhead of hypering.

AUTHOR

Elizabeth Mattijsen liz@raku.rocks

Source can be located at: https://github.com/lizmat/JSON-Fast-Hyper . Comments and Pull Requests are welcome.

COPYRIGHT AND LICENSE

Copyright 2022 Elizabeth Mattijsen

This library is free software; you can redistribute it and/or modify it under the Artistic License 2.0.