[Roger Binns] > I'd recommend you write code in your application first that knows how to > calculate the values you want. That way you can ensure the calculations > are correct, you have something for test harnesses that produces "good" > values and you have something to port to your final solution. > > I'd probably stop there. Remember that your application code and the > SQLite library are running in the same process. It is almost certain that > it is (better/more convenient/easier to develop and use) for this code to > be app code than within SQLite.
OK. But then I wont have an SQL interface for accessing interpolated data. It's acceptable. An SQL interface is mostly a "nice to have", so I could do some quick queries in SQL instead of having to write C code for it. > If someone will be doing queries expecting to match a row with second > granularity then your SQLite side solutions are virtual tables and > functions. The former is well documented. For the latter you can make a > function like "reading" which behind the scenes calls your app code which > prepares a statement, finds neighbouring readings and returns the > interpolated result - eg `select reading("2012-01-01T012345")` So, if I go for that approach, you'd recommend that I add functions, such as "reading", and if I want a mean temperature, I should add a function "mean" and not try to change what AVG will do? Thanks, -- Steinar _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users