Hi Aneta,

Thank you for the code sample. It seems that combining the ApplyCellFormula method with calling Rows.Insert is producing the slow results. I've attached an updated code sample with using the Rows.Insert method. Rows 1- 1000 takes 10 seconds, Rows 1001 - 2000 takes 32 seconds, Rows 2001 - 3000 takes 55s.

The results were pretty similar when using SuspendCalculations. Rows 1- 1000 takes 10 seconds, Rows 1001 - 2000 takes 32 seconds, Rows 2001 - 3000 takes 53s.

If I remove the ApplyCellFormula calls, but still use Rows.Insert, then 3000 rows takes about a second.

So I believe this is caused when combining calls to ApplyCellFormula and Rows.Insert. I do need to use Rows.Insert because we are using the Infragistics Excel classes to insert data into an existing Excel workbook.

Thanks,

Paul

Hello Paul,

Let me make sure I understand you clearly - you would like to ‘insert’ new rows between already existing rows or at the beginning of the worksheet. When you are calling Insert method in the loop, the preexisting rows are moved downward on every iteration. That can definitely cause excessive overhead, especially if you are moving thousands of rows on every iteration and that is true even without formula columns. My suggestion is to insert all the needed rows in advance.

ws.Rows.Insert(0, 3000 + 1);

for (int i = 0; i < 3000; i++)

{

ws.Rows[i].Cells[0].Value = 11 +i;

       ws.Rows[i].Cells[1].Value = 21 + i;

       ws.Rows[i].Cells[2].Value = 31 + i;

       ws.Rows[i].Cells[3].Value = 41 + i;

       …

}

This will move the already existing rows and after that fill the data in the empty rows. I modified the sample. When I test it and the time to add rows is less than a second. Please test it on your side and let me know how this works for you.

Regards,

Aneta Gicheva,

Infragistics

Hi Aneta,

Let me explain my business use case; I think this will help. We have an Excel "Templating" feature which allows our clients to take an existing Excel workbook and template out headers and footers and then our system fills in the data in the middle. The client also has the ability to take a single datasource and parse it out into as many sheets as they want. So we don't necessarily know how many rows to insert into each sheet until we actually add in the data. This works perfectly fine with the Infragistics excel classes when there are not many rows or Excel formulas are not used in each row, but seriously slows down when many rows are used with formulas. We don't typically consider 10k rows to be a lot of data (I could see a client writing up to 100k rows of data) and Excel can inherently handle inserting a row at a time without being too slow (more on this below).

I don't think pushing down the data in the cells is the issue. Try creating two separate sheets and writing the data in one, while inserting the rows in the other. This takes about the same amount of time as doing this in a single sheet:

Hello Paul,

Just wanted to follow up – have you been able to update your project to use the more bulk approach to inserting rows?

Generally speaking, some penalty is to be expected for doing separate inserts as each requires some validation tasks to be performed (like making sure tables are not split, end of worksheet) as well as check and potentially adjust formula references – not even the actual calculation yet mind you, the latter also applies across all worksheets in case they have formulas that reference the affected area. The last part is what you saw even doing inserts in a separate worksheet, which should not be affected as much.
We’ve discussed this internally and there are potential improvements we’ve set to investigate further, but in any case we can all agree it’s always advisable to insert large blocks of rows at once where possible.

Regards,
Damyan Petev
Software Developer
Infragistics, Inc.

Hi Damyan,

I haven't had a chance to implement this yet; however, there is no reason that it shouldn't work. Thanks for discussing this internally. I've verified your answer.

Thanks,

Paul