On Mon, Jul 2, 2012 at 1:14 PM, Simon Slavin <slav...@bigfraud.org> wrote: > Your argument is for SQLite to allow users to implement their own affinities > (datatypes ?) must like SQLite3 allows users to implement collation > algorithms and functions. But I don't think you're going to get something as > rarely used in SQLite4 when it's specially designed to be tiny and light. > SQLite stays that way by refusing to add rarely-used facilities.
No simple user-defined type is likely to allow for user-defined literals for user-defined types. What does matter, I think, is how commonly used some type might be. IP addresses are very commonly used in relational DBs. > But I still question whether you're engaged in premature optimization. Would > things really be that much slower if you stored IPv4/6 as Higits ? IPv4 > stored as hex is readable, sortable, can be transferred from place to place > as an 8 character string and works perfectly with the standard distribution > of SQLite3 with no changes. Implement conversion in- and out- functions > either as SQLite functions or in your programming language, as suits you. > You're concerned about the performance you lose by storing 10 octet strings > instead of having a custom type ? Possibly if you're running a major ISP or > doing packet analysis and logging for one. But you wouldn't be using SQLite > to do that anyway, you'd be running something with built-in caching and > redundancy. Hex is not good enough for CIDR blocks (i.e., bit strings of length <= sizeof(address-type)), unless you encode the bit string length mod 8 into the last byte, with appropriate zero-bit padding. Then it's good enough, but hardly user-friendly. IP addresses are useful in databases that store configuration information. NICs, for example, must have such databases, and any sufficiently large network is going to need one too. I've seen a number of home-grown Moira-like databases, and I maintained one called UName*It (which used UniSQL/X underneath). Any organization with home-grown DNS management is going to need this, and any open source and/or commercial DNS management tools will need this under the hood. IP addresses are really quite common in databases. This is why Postgres has native IP address support. Nico -- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
