Currently, the iter_by_ methods for non-unique indices work by iterating over all index values and then iterating over the values associated to each index. So basically, the difference to a simple iter is that the values will be ordered differently: with `iter_by_´, they will be grouped by index.
However, this is not that useful to my use case. When I saw that there are iter_by_ functions (without checking the full signature), I assumed that they worked like this: accept a value for the corresponding index and iterate to all map values associated to it. Basically, first look up the indices of the map values for the desired index value (through a simple hash map retrieval), then return an iterator used to iterate over the corresponding map values.
This is more efficient than the current iter_by_ methods as we do not have to filter for the desired index first. Effectively, the described functionality above would be the same as get_by_*().into_iter(), except that it avoids allocating.
Currently, the
iter_by_methods for non-unique indices work by iterating over all index values and then iterating over the values associated to each index. So basically, the difference to a simpleiteris that the values will be ordered differently: with `iter_by_´, they will be grouped by index.However, this is not that useful to my use case. When I saw that there are
iter_by_functions (without checking the full signature), I assumed that they worked like this: accept a value for the corresponding index and iterate to all map values associated to it. Basically, first look up the indices of the map values for the desired index value (through a simple hash map retrieval), then return an iterator used to iterate over the corresponding map values.This is more efficient than the current
iter_by_methods as we do not have to filter for the desired index first. Effectively, the described functionality above would be the same asget_by_*().into_iter(), except that it avoids allocating.