The map iterator implementation was only handling buckets in the
Element unions. This meant that if the length of HAMT is sufficiently
large, such that Elements could contain a link to a node, the map
iterator would panic.
To avoid this, load the Element link when it is encountered and
populate the iterator steps appropriately so that the entire map is
traversed.
Add tests that assert the map iterator functions as expected: traverses
the entire map, and returns each key exactly once.
Avoid panicing by letting Done proceed and retrospectively return
errors via Next as stated in MapIterator.Done interface
documentation.
masih
commented
2 years ago
The map iterator implementation was only handling buckets in the
Elementunions. This meant that if the length of HAMT is sufficiently large, such thatElements could contain a link to a node, the map iterator would panic.To avoid this, load the
Elementlink when it is encountered and populate the iterator steps appropriately so that the entire map is traversed.Add tests that assert the map iterator functions as expected: traverses the entire map, and returns each key exactly once.
Avoid
panicing by lettingDoneproceed and retrospectively return errors viaNextas stated inMapIterator.Doneinterface documentation.