INSERT or UPDATE with multiple rows

Question

Hi there!
I want to ask what I should do in the following case: I have to handle mutliple rows of data to insert OR to update into the database.
The first question is about how to decide whether I should take INSERT or UPDATE. I read here in the forum that I could take a SELECT-statement before, and, if it isn't null, I could update the resultset..if it is null I can make an INSERT-statement.
But now I a have multiple rows to update or to insert which I want to handle as a transaction (with a batch), so I don't want to check each row the way I described above. Does anyone has a hint ?
Thanks a lot in advance.

Answer 1

This is not a problem with java but rather a problem
with databases in general. The solution generally
depends on the data that is being operated on.
If there is a primary key involved, and most records
were expected to NOT be in the database, then you
could just insert them in blocks (transaction/batch).
Each block would fail when a primary key duplicate.
Then you can process each block as individual
l statements. The ones that fail are done as
inserts.
The reverse of the above is used if you expect most
records to be in the database. Do updates and the
break out the blocks with failures to locate the
inserts.
Keep in mind that queries for keys probably will be
faster, but that requires that your keys are ordered.
If you keys are ordered then you can get a range from
the initial data. Use that to query the database for
keys between that range (only return the keys.)
Using the returned keys you can decide whether each
h record needs to be an update or insert (presort the
data into each group and batch each group for more
speed.)
If the data is really mixed and the database supports
it then you can write a stored proc (MySQL does not)
which decides whether an insert/update is needed.
Finally if you have large amounts of data, bulk
operations especially inserts are better done with
tools supplied by the database vendor. Instead of
using JDBC to do the insert/updates you write the
output to a file and pass the file to the tool. You
might use JDBC (again with the ordered key query) to
decide which operation to do. Although faster for
large sets this is problematic due to the error
handling that you have to add.
Thanks for this, jschell. I look for your answers, because they're on the money and written well. I learn a lot from them. - MOD