From: Kevin D. <ke...@tr...> - 2005-10-04 14:48:41
|
Markus- Good catch. =20 I have to admit I'm quite puzzled by this. The offset of the lookup slot should be specified in terms of offset from the front of the translation page. The translation page entry is 10 bytes. Unless the size of the translation page header (10 bytes) = is somehow variable (and I don't think it is), then how could this happen? There are 817 available entries in a translation page ((8192-10)/10) - assuming that the 1700 number is actually 1634, then it would appear that every two translation pages things get out of whack? All said, this is mostly an academic exercise. Now that we are going to be changing the size of the record header, I think we should use the full 8 bytes for the reverse lookup. It will be faster, and the extra disk usage will more than be made up by = the coallesc/split/allocate/compact/reclaim system... However, I am concerned that maybe we are missing something important about the way the logical row id translation works (I was pretty sure I understood it, but there's no way that the rowids would not be divisible by 10 in my understanding). Any ideas on what's going on? (Note: I haven't had time to test this myself - I'll try to do that this afternoon). Thanks again, - K > Kevin Day wrote: > By taking advantage of this, and the fact that all logical rowids are=20 > divisible by 10 (because of the slot size in the translation page)=20 I've a scenario where this is not the case. After a bunch of records=20 (about 1700) are stored, the logical ids are not anymore divisable by=20 10, e.g. there're the ids 6816222 or 203826. After such a "gap", they're increased by 10 (until the next gap occurs). Best regards -Markus ------------------------------------------------------- This SF.Net email is sponsored by: Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl _______________________________________________ Jdbm-developer mailing list Jdb...@li... https://lists.sourceforge.net/lists/listinfo/jdbm-developer < |