The term that has been read is classified into one of the following categories (Kind):
    :- export struct(source_term(
	term,		% the read term itself
	vars,		% list of [VarName|Var] pairs (as in readvar/3)
	annotated       % the read term with type and source annotations 
	...
    )).
    
    For category 'comment', the term is a string containing the comment.
    For category 'end' and 'end_include', the term is the atom end_of_file.
    In all these cases, vars is the empty list, and annotated is
    uninstantiated.
    Note that either the vars-field or the annotated field is valid, depending on the setting of the with_annotations-option. If the option is set, the vars field contains an empty list, and the annotated term is in the same format as that returned by read_annotated/2, i.e. a recursive wrapper structure around the read term of the following format:
    :- export struct(annotated_term(
        term,                   % var, atomic or compound
                                % args of compound terms are annotated
        type,                   % term type (see below)
        file,                   % file name (atom)
        line,                   % line number (integer >= 1)
        from, to                % source position (integers >= 0)
        ...
    )).
    If the with_annotations-option is not set, the annotated-field remains
    uninstantiated, and the vars-field is a list as detailed in readvar/3.
    Notes on module handling: When source_read/4 encounters a module-directive (which is a handled_directive), the corresponding module is implicitly created (unless it exists already, in which case it is either reused or erased and re-created, depending on the setting of the recreate_modules option), and that module becomes the context module for any subsequently read clauses or directives. By default, source_close/2 removes these modules again in order to restore the original state.