There are at least two models for navigating data sources. I've named these models "Walkable" and "Streamable". While Raku has leaned strongly towards the Streamable model (due to the benefits in concurrency), it uses the Walkable model for grammars -- that's what backtracking is all about.
| Item | Walkable | Streamable |
|---|---|---|
| Movement through data source | Forwards, backwards, absolute, relative | Forwards only |
| Advantages | More flexible in coding; makes the easy things easy | Better for large data sources, and for concurrency |
| Disadvantages | Doesn't scale as well | Comes with restrictions that make it more difficult to use |
| Pointer | Cursor | Iterator/Supply |
| Comparable XSLT mode | Non-streaming | Streaming |
| Where does Raku use it? | Grammars | Most other things, including Iterators and Supplies |
Two new roles will need to be added to Raku to support the generic sources (rather than just the grammar-specific Str). These are:
Walkable. This indicates that the data structure is
walkable, and has a method that provides a Cursor. The
data structure in question might be a Str, a Tree,
or a Table.Walkable has one or more Cursors that
indicates the current point in a walkthrough of the data structure. The
difference between a Cursor and a Iterator
is that a Cursor can be walked backwards, and can skip forwards/backwards
through the Walkable. This is less parallelisable than an
Iterator, but is necessary for backtracking (which is
important both in Regexes and in Logic Programming). In this case, it's useful to compare the sources by type:
| does Walkable | Parent Type | Indicator Type | Walks by | Comments |
|---|---|---|---|---|
| Str or Stream | StreamCursor (?) | Character | ||
| Relation | TOP::Cursor | Tuple | ||
| XML Tree | Tree | JOP::Cursor | Node (or Element?) | |
| File Tree | Tree | JOP::Cursor | File |