Help please on advanced MySQL...

Question

Dear all:
Sorry for posting some MySQL-related topics here, which neverthess is a portion of a Java intensive project =) I tried to ask for some help on MySQL forum but it seems here would be more gurus on MySQL!!
Basically I need to do some manipulation on an existing table, with the task detailed as follows:
For each pairwise rows/records in a table which have disparate values on the same attribute/column, say A, check and then record (to be passed on and computed later in Java environment) all column names with identical values in these two rows. For example, suppose the table is like: A  B  C  D        <-- Field/Column Names
a1 b1 c1 d1
a2 b1 c2 d1
a2 b2 c1 d2 Then
(1). consider row1 & row2, which have different values on A, they coincide on column B and D, so 'BD' needs to be recorded;
(2). consider row1 & row3, which have different values on A, they coincide on column C only, so 'C' needs to be recorded.
So as for column A, {BD, C} is the wanted answer.
What Im planning now is just join the table with itself, specifying different values on A. Something like: SELECT [column name/s??] FROM theTable t1, theTable t2 WHERE t1.A <> t2.A AND
[some value comparisons on B, C and D, respectively to find out which one/s(column name/s) should be recorded] Any help is welcome & many thanks in advance!!
Frank =)
P.S.: I feel that there should be a way to do some computation within MySQL, but just cannot find the way to it!! Althernatively, the computation may be done in Java but this would cause a big performance degradation since it incurrs enormous information passing between Java & MySQL via JDBC.

Answer 1

Thanks very much for your reply!! =)
Sorry for not making the question a bit clearer.
So, for every pair of rows (or a subset of all thepairs?), you need to extract which columns match on
that pair
First of all, I need to look at each attribute once
at a time; the example in my 1st post was for field A
(subsequently I need to examine B, C, D one after
another)
When looking at a field, again let's say A, I need to
elicit ALL pairs of rows which have DIFFERENT values
on A, i.e. t1.A <>t2.A, and then compare if both have
the same values on any other columns; if so, record
the column names. However I cannot specify 'AB' in
the SELECT clause, since there's no way to find out
their values coincide on 'AB' prior to computation.Unless I'm totally missing the point, yeah, there is. The where clause tells us that they will coincide on exactly AB.
You need separate queries for each combination--one query for 'A', one for 'AB', one for 'AC', etc.
If you don't want to do that, you'd have to grab each pair of rows, and then basically do a bunch of if tests (or generate a map or something) to see which fields match. I don't know if you can do that in SQL or not. I suppose you could in a stored proc, but I wouldn't know how.
In Java it's easy, but it requires either storing the entire table in memory or making many trips to the DB.