Most Efficient Spreadsheet File

Hello,

I am going to write a Matlab script that will pull data from multiple Excel files. I imagine this script will take a long time to run. I was wondering if I could save my Excel files as a different file type to make the script work more efficiently. I will only be using one spreadsheet from each excel file and I am only going to be reading text from the Excel files.

Thanks.

[Guillaume]

Considering it's pretty simple to write your script using xlsread:

1. write the code as simply as possible not worrying about optimisation
2. if code is too slow, run the profiler
3. only then, if the profiler indicates that xlswrite is a bottleneck, replace it with something else.

There's little point in complicating the code if it brings little benefit (premature optimisation).

[Image Analyst]

Tell us how many files you will read. Is it more than about 5 or 10 or so?

[Daniel]

I am going to be reading 20-30 files. Each file will have roughly 2000 rows and 5 columns of data. Only one column of the data will be numeric, the other columns will be names, serial numbers, and comments.

Answer 1

[Image Analyst]

It will take a long time, unless you use ActiveX. See my attached ActiveX demo. With ActiveX it only opens and shutdowns Excel once instead of every single time you call xlsread().

[Image Analyst]

dlmread(), readtable(), and importdata() are also possible options if your data is formatted properly.

[Image Analyst]

For that amount of data, you certainly don't want to use xlsread(). You could use ActiveX, but if you can, I think the fastest way would be to convert the data to text (like a csv file) and use csvread(), dlmread(), or importdata(). ActiveX would be more complicated and probably not faster than these text-reading functions since ActiveX still has to launch Excel.

[Daniel]

I can't use csvread or dlmread because those are for numeric data only. I made a CSV file and tested the import speed using both importdata and textscan with a simple MATLAB script. Each time I imported the data, textscan was about 10-30 times faster than importdata.

Answer 2

[Evan]

If Image Analyst's solution doesn't work for you, you might try csvread. Assuming you're able to save your data in .csv format, I've found that csvread is a couple orders of magnitude faster than xlsread.