Commented by: Baldur Fürchau (bfuerchau)

writing error;-).

Commented by: @cincuranet

But the `IscException` is not handled, isn't it?

Commented by: Baldur Fürchau (bfuerchau)

Thats right, but you can do try/catch around to change the exception.

By th way:
In SQL-Reader i get an exception if the destination array is shorter than the field count.
I think this is better also to implement.

If you are interested i create my own FbBulkCopy like SQL-Server has.
At the moment i can make 1500 Insert/Second, with 4 parallel connections up to 10.000, depending on fieldcount and sizes.
So when i use all available cpus i can fast load my datawarehouse. I can give you than the solution;-).

Commented by: Baldur Fürchau (bfuerchau)

Finally i made an additional enhencement:

	public override int GetValues\(object\[\] values\)
	\{
		CheckState\(\);
		CheckPosition\(\);

		var count = \_fields\.Count;
		if \(values\.Length < count\)
			throw new ArgumentException\("Argument too small to hold values"\);

		try
		\{
			for \(var i = 0; i < count; i\+\+\)
			\{
				values\[i\] = \_row\[i\]\.Value;
			\}
			return count;
		\}
		catch \(IscException ex\)
		\{
			throw new FbException\(ex\.Message, ex\);
		\}
	\}

	private void CheckIndex\(int i\)
	\{
		//if \(i < 0 \|\| i \>= FieldCount\)
		//	throw new IndexOutOfRangeException\("Could not find specified column in results\."\);
	\}

1) if the argument too small i send an exception like other readers does. It makes no sence, to get less values as available.
2) Catch for iscExcepption is done only once and not for each column.
3) I commented out the checkindex because the clr send the same exception.

With this changes i get now 15.500 Rows/Second on my machine.

Also if DataTable or DataAdapter fills DataTable-Objects, for every column is checked the index which is always correct. Additional the loadingmechanism gets the GetValues() with check index for each column and does call also IsDBNull for each column where the index checks again.

So this improve also the loading behavier.
It seams for each its "microoptimization", but the functions are called, in my test 100,000 Rows with 92 columns 18,400,000 times.

Try it;-).

Your code is missing the code with ExpectedColumnTypes.

Also I don't think that skipping CheckIndex is a good idea. Yes, same exception will be thrown, but from different place and with less descriptive message.

Quickly looking into npgsql I see that it's doing the same in case the array is smaller.

I'm closing this now, because I believe the proposed changes are not equivalent. But, that doesn't mean I'm not open to suggestions for optimizations, etc., quite the opposite.