Hello All,
Mike Speirs at Sun in New Zealand pointed me toward you-all. I have
several sets of questions, so I plan to group them and send several emails.
This question is about the name/attribute mapping layer in ZFS. In the
last version of the source-code that I read, it provides hash-tables.
They are a good way of finding an exact match for a name.
I''m from a database background, so I am thinking of the hash-tables as
if they were an index to some data that could have been stored in a
different form.
Are there any plans to provide the following searches efficiently:
- matching prefixes
- eg I give it a telephone number, and it gives me a list of file
descriptors for the:
- suburb
- city
- province/state/etc
- country
- hemisphere
each of which matches fewer digits at the beginning of the number
[[ general tries work for this ]]
- upper and lower bounds
- eg imagine fixed-length dates and times that are all digits;
I give it a date and time, and it gives me the file descriptor
for the latest date and time <= the one I give it
[[ tries can be made to do this, but B-<variant>-trees are better
]]
Regards,
James.