Rand Stats



Actions Status


Hash2Class - A role to create class instances out of a Hash


  use Hash2Class;

  class UpdateInfo does Hash2Class[
    added   => Date(Str),
    changed => Date(Str),
  ] { }

  class FBB does Hash2Class[
    bar        => Int,
    baz        => UpdateInfo,
    '@bazlist' => UpdateInfo,
    '%bazmap'  => UpdateInfo,
    zap => {
      type => Str,
      name => "zippo",
      why  => "Because we can",
  ] { }

  my %hash =
    foo => "foo",
    bar => 42,
    baz => {
      added   => "2020-07-18",
      changed => "2020-07-19",
    bazlist => [
      { added => "2020-07-14", changed => "2020-07-15" },
      { added => "2020-07-16", changed => "2020-07-17" },
    bazmap => {
      first  => { added => "2020-07-01", changed => "2020-07-02" },
      second => { added => "2020-07-03", changed => "2020-07-04" },
      third  => { added => "2020-07-05", changed => "2020-07-06" },
    zap => "Groucho",

  my $fbb = FBB.new(%hash);
  dd $fbb.foo;                    # "foo"
  dd $fbb.bar;                    # 42
  dd $fbb.zippo;                  # "Groucho"
  dd $fbb.bazlist[1].added;       # Date.new("2020-07-01")
  dd $fbb.bazmap<third>.changed;  # Date.new("2020-07-06")


The Hash2Class role allows one to create a class from a parameterization of the role. The parameterization consists of a list of Pairs in which the key indicates the name of key in the hash, and the value indicates the type the value in the hash is supposed to have, or be coerced to. The key becomes the name of a method accessing that key in the hash, unless it is overriden in more extensive parameterization.

A key can be prefixed with @ to indicate an Array of values in the hash, or be prefixed with % to indicate a hash, or $ to indicate a scalar value (which is the default).

The types specified can also be classes created by the Hash2Class role, so recursive structures are possible, as long as they are defined in the correct order.

Classes made with the Hash2Class role are instantiated by calling .new with a hash as its only parameter. Such a hash is typically the result of rakufication of a JSON blob (e.g. with from-json of the JSON::Fast module). But the hash can be created in any manner.

Values are checked lazily, so no work is done on parts of the hash that are not accessed.


Hashes can be filled in many ways: JSON just being one of them. And data is not always as clean as you would hope they would be. This role allows you to add lazy typechecking to such a hash of data. It also prevents problems caused by spelling errors in keys in your code: instead of silently returning Nil, you will get a "Method not found" error on misspelled method names.

Since the type checking occurs lazily, no CPU is spent on typechecking values you do not actually need. Should you do want to have complete typechecking on all keys / values in the hash, then you can call the .invalid method on the object: this will visit all values in the hash recursively and produce a corresponding hash of any errors found, or Nil if all is ok.


There are three modes of parameterization:

Just specifying an identifier (a string of a single word), will create a method with the same name, and assume the value is a Str.

A pair consisting of an identifier and a type, will create a method with the same name as the identifier, and assume the value is constraint by the given type.

The type can also be specified as a string if necessary:

bar => "Int",

Coercing types are also supported:

bar => Int(Str),

A pair consisting of an identifier and a Hash with further parameterization values.

Four keys are recognized in such as Hash: type (the type to constrain to), name (the name to create the method with, useful in case the key conflicts with other methods, such as new), default to indicate a default value (defaults to Nil) and why (to set the contents of the .WHY function on the method object.


If you have a file with a JSON blob for which you need to create a class definition, you can call the h2c-skeleton script. You call this script with the JSON blob on standard input, and it will print a class definition on standard output.

Class names will be selected randomly, but will be consistent within the definition of the classes. The order in which classes are defined, is also correct for compilation: generally one only needs to globally modify the class names to something that makes more sense for the given data. And possibly tweak some standard types into subsets with a more limited range of values, e.g. Int to UInt, or Str to DateTime(Str).



my $foo = Foo.new(%hash);

An object of a class that does the Hash2Class role, is created by calling the new method with a hash of keys and values. Each of these values can be another hash or array: these will be handled automatically if they were so parameterized.


with $foo.invalid {
    note "Errors found:";
    dd $_;

The invalid method either returns Nil if all values in the hash where valid. Otherwise it returns a hash of error messages of which the keys are the names of the methods, and the values are the error messages. Please note that this check will access all values in the hash, so it may take some time for big hashes.


Elizabeth Mattijsen liz@raku.rocks

Source can be located at: https://github.com/lizmat/Hash2Class . Comments and Pull Requests are welcome.


Copyright 2020, 2021 Elizabeth Mattijsen

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