On 10/28/10, Emmanuel Lecharny <[email protected]> wrote: > On 10/28/10 9:55 AM, Kiran Ayyagari wrote: >> >>>> One small last thing : currently, indexes files are named using the >>>> Attribute OID. I find it not very friendly. Can't we use the Alias >>>> instead >>>> ? >> -1, there is a reason we changed back to OID, it is to avoid creating >> duplicate indexes >> on the same AT. >> >> If we use alias then there is no way to check them, >> e.x if user creates two indexes on "cn" and "commonName" then jdbm >> creates "cn.db" and "commonName.db" which we cannot distinguish during >> creation > We can check that using the SchemaManager, which is available when we > create the files. AT.getName() always return the first alias. we can't depend on this as the schema *might* change(e.x a custom schema) the order of aliases. Checking for the existing indexes requires to get all the *.db files and perform various checks on their names (after removing ".db") but this check needs to be performed after *every* single index instantiation
>> Another case is to find which indexes are no longer needed when user >> removes them from config. > Same, using the SM will do the trick. >> We store a OID.txt file with AT's definition alongside the index files >> to let the user know what that indexed attribute is. > Do we? Never saw this file. yes, we create them > > -- > Regards, > Cordialement, > Emmanuel Lécharny > www.iktek.com > > -- Kiran Ayyagari
