Try a StringSearcher with the following regex which is based on searching for all the substrings that does NOT contain a comma:

(o^,]+)

Also define a list name for the matches, e.g.

You can then use a ListElementCounter to find the number of comma-separated values in your input string.

If you need this functionality in several places, it's fairly easy to pack it into a custom transformer.

Could you use an attributeSplitter (separator: comma) to create a list and then a listElementCounter to get the number of items in the list?

Thanks David. Magic worked well.

So much valuable. I appreciate

Yes, that's the way to go. Search the strings and store the matched items. Then use the stored items as input to ListElementCounter.

This made my life easier.

I agree that it's a good approach to split the string into individual items then count the number of them. 

Alternatively, you can also replace each item (including trailing delimiter) with a single character (e.g. 'x') then get the length (i.e. number of characters) of the resulting string ('xxx...'), which you can consider as the number of items. For example, this expression returns the number of comma-separated items in the value of 'Building Classification' attribute.

@StringLength(@ReplaceRegEx(@Value(Building Classification),".+?(,|$)",x))

(o^,]+)

This is brilliant.  Exactly what I was looking for!

I have Postcode data with an attribute that has either YES or NO in it.  I'm aggregating data by Postcode Sector, then I want to report how many YESs and how many NOs there are in that Sector.  I spent hours and got nowhere putting the values into a List.  But concatenating the values in the Aggregator then using this answer did the job in 5 minutes!

If you want to count occurrences of values in a list, the ListHistogrammer is what you want

I tried that, having found the suggestion in a different question. But I couldn't get it to work. I just got blank values. I could have persevered but this worked instantly:

ListHistogrammer may do the counting better, but you've still got to know which element of the _histogram List NO and YES are in to be able to extract the count. And if you get other values in the data, these could upset your _histogram List, for example:

_histogram{0}.value = 'FISH'

_histogram{0}.count = 3

_histogram{1}.value = 'NO'

_histogram{1}.count = 2

_histogram{2}.value = 'YES'

_histogram{2}.count = 1

If it was only YES and NO in the data, I could probably sort the List, then assume element 0 was NO and element 1 was YES, but FISH puts a spanner in the works. If I have to test the value of each _histogram element to see if it's NO or YES, I will have to work out how many elements there are and loop through the List, testing each one.

The benefit of david_r's solution is that you only count the things you want to count, and don't have to worry about rogue values. That's also the downside though. The sum of the counts may not equal the total number of aggregated features. For example, if I aggregate 10 features but one of them has FISH instead of YES or NO, I will end up with say 4 YESs and 5 NOs, which is one short of the total. But the data could be validated upstream of this part of the process.