SQL Loader and record order

Question

I am using sqlloader utility to load data from a CSV file in to a table.
My .ctl file looks as below
------- 8< -------
options (errors=5,SILENT=(HEADER, FEEDBACK),direct=true)
load data
infile "mytest.csv"
discardmax 0
into table owneruser.MY_TABLE
fields terminated by "," optionally enclosed by "##"
(ID, ID1, VAL, VAL2)
------- 8< -------
sqlldr tool is run with this ctl file by another database user who has sufficient privileges to insert this data in to mytest.csv has about 400000 entries each entry maps to one row in MY_TABLE. Before loading data using the sqlldr MY_TABLE is truncated.
In mytest.csv, the value for ID field is a number starting at 1 which keeps incrementing by 1 for the next entry. The records are ordered by ID in the csv file.
After loading the data using sqlldr, when we query MY_TABLE (select * from MY_TABLE), so far the records are returned in same order in which they were inserted (i.e. ordered by ID). But off late they are not being returned in random order. This happens only on one database instance. On other test instances the the resultset is ordered. I agree that the only way the order can be guaranteed is by using the ORDER BY clause.
But, I was wondering why this has worked even when ORDER BY is not used.
This is the only way in which MY_TABLE is manipulated. Rest all use it only for querying.
ID is the primary key column in MY_TABLE and there is an index on (ID, ID1).
Thanks in advance.
S

Answer 1

There are any number of reasons that the data would be coming back in a different order since you're not using an ORDER BY. My guess is that the most likely reason is that you have one or more extents in your table that is physically before another extent that it is logically after, in which case a full scan would read that extent first. You may also be seeing differences in how ASSM happens to choose which block to insert into, in the use of parallelism, etc.
Justin