Recent changes to bugs

#5 Jaro returns score 0.0 for resembling strings

2015-12-09T12:33:16.483000Z

This is caused by a glitch in the source code of Jaro class. At function commonChars, the line:

for (int j=Math.max(0,i-halflen); !foundIt && j<Math.min(i+halflen,t.length()); j++) {

...is wrong. It should be:

for (int j=Math.max(0,i-halflen); !foundIt && j<Math.min(i+halflen**+1**,t.length()); j++) {

library not thread safe

2010-05-11T13:24:46Z

SimpleTokenizer has a static DEFAULT_TOKENIZER used by many methods in their constructors. However, SimpleTokenizer has a Map which is not thread safe. We have had infinite loops in the core java treemap get when used in a multi-threaded web environment. It would also seem incorrect to share a tokenizer between all methods.

One solution is to change the matching methods constructors to use a new SimpleTokenizer each.

Jaro returns score 0.0 for resembling strings

2010-02-05T12:29:43Z

When I give jaro the strings 'ihdienstdatuni' and 'indienstdatum' the returned score = 0.0, although there are only 3 differences.
When I append a space after 'indienstdatum' and leave the other string the same, the resulting score is 0.74. When I append a space to both strings, the score goes up to 0.78.

Jaccard.score() returns NaN if both strings empty

2009-04-08T01:35:20Z

Jaccard.score(s1, s2) returns NaN if both strings have zero tokens. this happens when both strings are empty, or the tokenizer doesn't find any tokens (NGram tokenizer with minNGramSize > size of the string)

It should *probably* return 0.

Probable typo in compute method in com/wcohen/ss/AffineGap.j

2008-04-09T13:35:09Z

Line 66 of AffineGap has a '=' where it should probably be a '-'.

The line in question is:
it.get(i=1,j-1) + matchScore

and i is one of the parameters to the method containing that line.

Level2 scoring issues when bags of different sizes

2008-02-28T20:51:49Z

I've encountered problems with Level2 not accurately scoring bags of different lengths. I'm including a patch which solved the problem for me.

The problem presented itself when Level2.score(s, t) was called with size(s) < size(t). For example:

s = {'Frances', 'Fyfe'}
t = {'Mary', 'Frances', 'Fyfe'}

Level2.score(s,t) -> 1.0
level2.score(t,s) -> 0.83

The problem is, I believe, that the algorithm always iterates over s. What should happen, in my opinion, is that the algorithm should iterate over the larger of the two sets.

I'm including a patch which does just that. I hope this is helpful.

Jaro metric uses halflength of the longer string

2004-04-14T08:29:34Z

Here's a patch:

--- Jaro.java.orig 2004-04-14 10:21:34.000000000 +0200
+++ Jaro.java 2004-04-14 10:21:26.000000000 +0200
@@ -51,7 +51,7 @@

private int halfLengthOfShorter(String
str1,String str2)
{
- return (str1.length() > str2.length())
? str1.length()/2 + 1 : str2.length()/2 + 1;
+ return (str1.length() < str2.length())
? str1.length()/2 + 1 : str2.length()/2 + 1;
}

private String commonChars(String s,String
t,int halflen)