Remove selectors from old interest format
Delete the code with selectors and related test, simplify the Interest-Data matching process.
The new version NDN interest simplify the selector function, so we need to update the corresponding part in repo-ng.
Updated by Junxiao Shi over 3 years ago
Is there a specific reason that names are stored as blob type?
Yes. Names are binary, always. They should be processed and stored as such.
It's not easy to do name prefix matching using blob. If we store name as string in the database, then it will be easy to implement prefix matching.
In SQL there’s no “prefix matching”. Instead, you convert the query to “WHERE name >= query AND name < successor-of(query)”.
See example in repo-sql: https://github.com/3rd-ndn-hackathon/repo-sql/blob/47671612ef8b7c40931e5129c4d82626033f90a1/src/query-processor.cpp#L98-L99
Updated by Alex Afanasyev over 3 years ago
Junxiao, I think there is a problem with this logic. As far as I understand, it would work if we exclude T and L for the overall name in the stored value, but should have undefined results when comparing name as a whole. Am I misremembering our previous discussion of http://named-data.net/doc/NDN-packet-spec/current/name.html#canonical-order ?
Updated by Junxiao Shi about 3 years ago
it would work if we exclude T and L for the overall name in the stored value
name column should only store NAME-VALUE.