Fastest Way of Opening and Reading .csv Files (Currently using xlsread)

I am currently trying to convert 100,000+ csv files (all the same size, with the same data structuring on the inside) to mat files, and I am running into the issue that it takes an extremely long time, and sometimes Excel stops responding. Are there any other functions that could cut down on the read time of these .csv files?

I read something about trying the COM server that runs Excel, but I am not sure how to implement it. Any thoughts?

Answer 1

[Kirby Fears]

Thankfully, you don't need to interact with Excel to read csv files. You can use textscan to read csv files quickly.

I wrote a function called delimread that utilizes textscan with automatic parameters. It might be easier than learning how to parameterize textscan.

Answer 2

[Todd Leonhardt]

readtable

http://www.mathworks.com/help/matlab/ref/readtable.html

Answer 3

[Jeremy Hughes]

Since you have multiple files, you may want to consider using datastore. (Since R2014b)

In many cases, you can just use the following pattern to read a large collection of files,

ds = datastore('folder/containing/your/files')
while(hasdata(ds))
 t = read(ds)
 % do stuff to t.
end

Hope this helps,

Jeremy

Answer 4

[TastyPastry]

There's a function csvread() which only works on numeric data.

The other way you can do it is to use textscan(). Both of those methods should be faster than xlsread() since xlsread() uses Excel, which is pretty slow.

[Walter Roberson]

csvread calls dlmread calls textscan.

Answer 5

[Kristoffer Walker]

In my experience, the absolute fastest method is textscan. Here is a benchmark to support my claim using a 3.6 GB CSV file with 4 columns.

Using textscan: 81 seconds

Using readtable: 143 seconds

I tried dlmread and csmread, but they had problems with parsing input. They are not recommended.

Good luck.

Kris