In what order will select * return the data?

Question

hi guys,
1) if we just do a select * from emp, will it return the data in the order of the data blocks in the extents. i.e. return all data blocks from extent 1, then from extent 2 and so forth?
2) if we do a select * from emp, and say half the data blocks are already in the database buffer cache, how does Oracle know that the data blocks from emp are already in the database buffer cache? Does it keep some lookup table?
thanks

Answer 1

OracleGuy777 wrote:
thanks guys,
I will mark the question as closed, even though I am still a bit unsure as to why if we just start up a database so that the db cache is empty, and do a select * on a small table that is all on one disk, why the Oracle kernel would from a logical point of view choose different order to return the rows.
but thanks anywayThe order in which rows are returned is a side effect of the execution plan chosen by the Query Optimizer (which is one reason why the ORDER BY clause exists), in plain relational theory both your underlying table and the results are relations (simply an unordered set of tuples) so consequently the optimizer is free to follow whatever method it likes to return the results. In the case of a Rule Based optimizer then there is simply a set of rules that determine the choice of execution plan, in the case of a cost based optimizer then the optimizer attempts to minimize the time taken to return the resultset by choosing an execution plan based on statistical information and algorithms. The existence of the buffer cache is in fact ignored for most practical purposes and certainly for the purposes you imagined.
Now if your table is just that, a straight forward heap table with no indexes then you will likely get what you expect, that is rows returned in the order that the blocks in which they are held are located on disk; this isn't insert order necessarily by the way especially if ASSM is involved. This is only however because a full tablescan is the only available option for satisfying the query. Most tables however aren't like that, they have indexes, the query involves joins and filters and so on. Once you start allowing the possibility of other access paths then you allow the possibilty that rows or rather blocks will be visited in a different order to the simplistic one described above.
Niall Litchfield