I have a half-baked idea based on using Flexcel for profiling a workbook. I'm hoping that I could use some Delphi profiling code and then extract the parts that reflect the performance of the spreadsheet itself. Does this sound feasible - or have you done anything similar?
I am not 100% sure I understand the idea. Are you trying to find what is making recalc slow in an xls file?
If that is it, I am not sure it would work. FlexCel is a completely independent implementation from Excel, so it will probably have different performance bottlenecks. Profiling FlexCel will give you a hint in the parts where FlexCel is slow, but not really where Excel is slow.Specially for recalc, there is a big difference because Excel calculates only what is changed (as Excel is designed for interactive use), while FlexCel calculates the full thing everytime (since it is designed for server use, and you normally calculate once at the end, so the code to conditionally calculate what changed would only slow down the recalc).
You might get an idea of slow functions anyway (like lookup) but probably this is easier found in Excel itself.
As said, I am not sure I understood the idea, if I got it completely wrong please let me know.
Yes it's for profiling workbooks that will be used server-side with Flexcel. I realise the profile/behaviour will be different from Excel itself, but that's fine. I'll also be doing full recalcs each time since all of the input data will be getting changed between runs. In fact, Flexcel is ideal for what I'll be doing, and it would be nice to have a way of identifying and removing bottlenecks caused by poor workbook design.