5.4.1 Dict types and tags

The PIP implementors forum is working on three PIPs related to dicts and their syntax: 0102, 0104 and 0109. These PIPs are converging to the following consensus regarding the Tag{...} syntax:

• The ... in Tag{...} is a set of Key:Value, where Key is an atom (and possibly (small) integer) and Value is any term. This is compatible to SWI-Prolog.
• The Tag is either
  • The atom #. This denotes a dynamic dict, which is what our dicts are.
  • Any other atom. This denotes a named arguments or struct. This syntax is mapped to a compound term whose arity and allowed keys is defined by a declaration.
  • A variable. This is parsed into an attributed variable (see section 8.1).

Our aim is to support these PIPs. Next to mapping “any other atom” to a named argument structure, we will support mapping these to a dynamic dict, optionally with user-defined functions (see section 5.4.2.1). The transition is guided by the Prolog flag var_tag. Its current default (dict) is compatible with the original dict implementation of SWI-Prolog.

The use of unbound tags is deprecated. It is recommended to use # for anonymous dynamic dicts. The main reason for using unbound tags is in using them with the predicates :</2 and >:</2 to access a set of keys in a dict with unknown tag. In the current version the # tag matches any tag in these predicates. For example, to extract the x and y from a dict, we can now use

    ...,
    #{x:X, y:Y} :< Dict,