Determine if column in ResultSet contains values in all rows

Question

In my application, I perform a costly query that takes minutes to produce a report. I am trying to make a generic class that transforms a ResultSet to and Excel spreadsheet, where a column is excluded from the spreadsheet if it only contains nulls. I can remove the columns from the Excel sheet after the fact easily, but it is difficult to "glue" worksheets back together after I have already split them when there are too many columns.

I could do a query to check if each column is null, but this would entail running the costly query all over again, perhaps multiple times, which would make the generation of the spreadsheet take too long.

Is there a way that I can query the ResultSet object that I already have (a little like ColdFusion) and remove columns from it?

EDIT I ended up adding a pre-processing step where I added the column numbers of the used columns to a List<Integer> and then iterating through that collection rather than the set of all columns in the ResultSet. A few off-by-one errors later, and it works great.

Answer 1

Can you extract the data from the ResultSet and store it in memory first, before creating the work sheet, or is it too large? If so, then while you're extracting it you could remember whether a non-null value has been seen in each column. Once you're done extracting, you know exactly which columns can be omitted. Of course this doesn't work so well if the amount of data is so large that you wouldn't want to store it in memory.

Another solution would be to store the results of the costly query in a "results" table in the database. Each row for a given query execution would get stamped with a "query id" taken from a database sequence. Once the data is loaded into this table, subsequent queries to check whether "all values in column X are null" should be pretty speedy.

Note: if you're going to take this second approach, don't pull all the query data up to your application before storing it back to the results table. Rewrite the original "costly" query to do the insert. "insert into query_result(columns...) select {costly query}".

Answer 2

I could do a query to check if each column is null

Better still you could incorporate that check into the original query, via a COUNT etc. This will be miles quicker than writing Java code to the same effect.