"Luca Ferrari" <[EMAIL PROTECTED]> writes: > In short each transaction with xmin<=xid<=xmax can see such tuple, > otherwise it cannot (of course beeing xmin and xmax different transtaction > from xid and beeing committed), isn't it?
No, it's more complicated than that. We can't simply compare xids numerically as you point out later. Even for regular transactions that doesn't work because there's no guarantee that transactions will commit in the same order they start and what matters is when they commit. We actually check what transactions have committed at the time your transaction starts (or command starts in read-committed mode). That's called your "snapshot". We check the xmin and xmax from a tuple against that list. If you really want to understand how snapshots work at this level you could read (slowly -- it's pretty dense stuff) through src/backend/utils/time/tqual.c:HeapTupleSatisfiesMVCC() -- Gregory Stark EnterpriseDB http://www.enterprisedb.com ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
