Quoth da...@cantrell.org.uk (David Cantrell): > On Sat, Jan 10, 2009 at 07:41:22PM -0500, Jonathan Yu wrote: > > On Fri, Jan 9, 2009 at 12:57 PM, nadim khemir <na...@khemir.net> wrote: > > > 3/ don not put all your data in the module. Although this module is very > > > specific and unlikely to be loaded with a lot of other modules at the same > > > time, it is good to conserve memory. Split your data and load it > dynamically, > > > compress the data if possible. > > How would you suggest doing this? Do you think using SQLite as a > > backing store would be a good idea, or perhaps the Berkeley DB? > > If you put the data in a __DATA__ segment, it only gets loaded into > memory when accessed. My own experience with Number::Phone::UK::Data - > which has a very big database in __DATA__ is that DBM::Deep is very > frugal with memory - as the data is indexed, it is quite happy to seek > back and forth on the DATA filehandle and only load from disk those bits > that are needed.
If you're going to do that (and it's a good idea), surely it would be better to put the data in a real file and use File::ShareDir or some such to find it? Ben
