"Specified cast is not valid" when populating DataTable from OracleDataAdapter.Fill()

Question

I can't seem to find this question anywhere on Google (or StackOverflow), which really surprised me, so I'm putting it on here to help others in the same situation.

I have a SQL query which runs fine on Oracle Sql Developer, but when I run it through C# usingadapter.Fill(table) to get the results, I get Specified cast is not valid errors (System.InvalidCastException).

Here is cut-down version of the C# code:

var resultsTable = new DataTable();

using (var adapter = new OracleDataAdapter(cmd))
{
    var rows = adapter.Fill(resultsTable);  // exception thrown here, but sql runs fine on Sql Dev

    return resultsTable;
}

And here is a simplified version of the SQL:

SELECT acct_no, market_value/mv_total
FROM myTable
WHERE NVL(market_value, 0) != 0
AND NVL(mv_total, 0) != 0

If I remove the division clause, it doesn't error - so it's specific to that. However, both market_value and mv_total are of type Number(19,4) and I can see that the Oracle adapter is expecting a decimal, so what cast is taking place? Why does it work on SqlDev but not in C#?

Solution 1

Answering my own question:

So it seems that the Oracle number type can hold many more decimal places than the C# decimal type and if Oracle is trying to return more than C# can hold, it throws the InvalidCastException.

Solution?

In your sql, round any results that might have too many decimal places to something sensible. So I did this:

SELECT acct_no, ROUND(market_value/mv_total, 8)  -- rounding this division solves the problem
FROM myTable
WHERE NVL(market_value, 0) != 0
AND NVL(mv_total, 0) != 0

And it worked.

The take away is: Incompatibility between Oracle number type and C# decimal. Restrict your Oracle decimal places to avoid the invalid cast exceptions.

Hope this helps someone else!

Solution 2

I know this thread is really old.. However I had a similar issue.

The best solution I have to use the Oracle method TO_BINARY_DOUBLE on Oracle Decimal column

Solution 3

I know this has been answered already, but I also found another alternative that I use as well. I used a CAST on the field that was giving me troubles.

Based on OP's SELECT command:

SELECT acct_no, CAST((market_value/mv_total) AS DECIMAL(14,4))  -- CAST to decimal here
FROM myTable
WHERE NVL(market_value, 0) != 0
AND NVL(mv_total, 0) != 0

Comments

[Chrisi]

Thanks for posting a solution to your own problem. probably saved me a couple of hours (and gray hair^^). But is there a way that avoids rounding numbers within the query?

[TheRoadrunner]

I know this post is kinda old, but this just happened to me, and rounding didn't help. In my case, I had some very high values. For completeness (hopefully not needed for market values :-) ) write: ROUND(GREATEST(LEAST(market_value/mv_total,7.9E28),-7.9E28),‌​8)

[Ali Umair]

You brother saved my day

[Lucas925]

You saved my life with this. Thanks!

[Crismogram]

while I did not follow this fixed. I salute you for posting it. I was running in circles for hours.

[Allen]

Ouch, Bitten again. So I did this: SELECT ...., ,to_char(Round(b.Length,2),'9,999.00') AS Feet,... from <some_table> which not only addressed the overly precise return value but also formatted the data in a tidy way, since I only needed it for reporting purposes.

[Allen]

For consistency I use a function taking the input value and returning a string to defend against this problem... plus it will round and format as desired. If I knew how I could post it but it is pretty trivial.