On Thu, 2007-09-06 at 20:20 -0500, Michael Glaesemann wrote: > On Sep 6, 2007, at 19:58 , Ow Mun Heng wrote: > > > Don't denormalise the table? > > Yes. Don't denormalize the tables. I would believe performance would be better it being denormalised. (in this case) > > > don't put them into arrays? > > Yes. Don't use arrays. Caveat: if the data is *naturally* an array > and you will not be doing any relational operations on individual > elements of the arrays, then it makes sense to use arrays. Treat > arrays as you would any other opaque type. Data is naturally an array, and will be used as an array in any case. Since there will not be queries where users will select any one of the values in that array, but the whole array itself. data willbe used in this form code | v1 | v2 | v3 | v4 A 1 2 10 23 B 10 12 15 22 C 11 24 18 46 D 21 22 20 41 which will be imported into statistical software/excel for further manipulation. I i give them in the denormalised form, it'll take them an addition 30min or so to make them back into the form above. and it'll make the queries more efficient too. index on Code, select * from foo where code = 'B'; By denormalising, I will also get the benefit of reducing the # of rows by a factor of 20.. (20 rows = 1 code) ---------------------------(end of broadcast)--------------------------- TIP 5: don't forget to increase your free space map settings
