I try to count words and have problems in string like this: 1. 2. 3. 4. 5.
If I count words using myTextStorage words] count] it returns 1 (one) word, but if I count words using nextWordFromIndex:, it returns 11 (eleven). And this problem exists only with numbers and period (or comma after it). If I have each number like x.x (where x is any digit) or string like this a. b. c. d. e. then it counts correctly and both methods returns 5 words.
(I know I can workaround it by counting words by counting whitespaces between them, but then it doesn’t work correctly with japanese language - that language doesn’t have whitespaces between words, but my method ( nextWordFromIndex:) works fine with any language except the case with numbers). So is there any workaround or explanation of why nextWordFromIndex: count a number with period like two words? Can anyone check is it fixed in Tiger? (Yes, I’m on Panther still)
Personally, I would avoid using [[NSTextStorage’s -words method anyway. It is slow, and so no good if you want to implement a live word count, and also it yields strange results - in one test I did, a document with over 40,000 words came up as having only just ove 2,000 word objects in the -words array.
This topic has been helpful to me - I was parsing whitespace to get my word count, but hadn’t considered how this would only be suitable for some languages. The trouble is that -nextWordFromIndex:forward: is not, according to the docs, intended for use with statistical analysis such as a word count:
“This method is intended for moving the insertion point during editing, not for linguistic analysis or parsing of text.”
That said, Douglas Davidson, one of Apple’s text gurus, actually recommended the use of this method for precisely this purpose:
So I guess it must be safe to use. I have therefore taken this route in in my own app now.
To (hopefully) answer your question:
The way I am doing this is as follows:
This way, punctuation and whitespace doesn’t get counted.
For a live word count, you can use NSTextView’s -shouldChangeTextInRange:replacementString: and -didChangeText: to calculate the difference in word count for the changed text - that way you can have a live word count that doesn’t slow down typing on even very large documents.
Hope that is of some help. Cheers, KB
Thanks for input! I did use shouldChangeTextInRange:replacementString: and -didChangeText: but had some problems, so I would not recommend this way. It’s better to calculate everything just in shouldChangeTextInRange:replacementString:
After posting on Cocoa-dev, Aki Inoue at Apple recommended using -doubleClickAtIndex: instead of -nextWordFromIndex:, so I am now using that for the actual word count (though I use -nextWordFromIndex: still for my -wordCountForCharRange: method, as it is easier to prevent that method going out of bounds). Personally I never had any problems using both -shouldChangeTextInRange:replacementString: and -didChangeText: for my live word count - this way I can get the range of affected words before and after the change, which requires replacementString being in the context of the rest of the text. Incidentally, you need to support -shouldChangeTextInRanges:replacementStrings: under Tiger. Cheers, KB
What about using this:
NSArray *words= [myString componentsSeparatedByString:@” “]; int count = [words count];
This won’t work with japanese language.
I use NSSpellChecker’s -countWordsInString:language: method … it works quite well for me.