LAE

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: dhrobertson

hi,

I thought that I had my head around how the fuzzy node works but like a lot of things, the more you stare at something the more it doesn't make sense or looks wrong.

I don't know whether I have the concept clear in my mind or not in the first place and reading the help doesn't help me....

my understanding is that if you have 2 datasets, for instance, that both have a key consisting of 2 fields and then another that may have a DateTime field

if you do a fuzzy Xref on these (or a fuzzy join) then the node will find all instances between the 2 datasets where the keys EXACTLY match first to narrow down the set to do the fuzzy logic on and then apply the fuzzy logic on the defined field with the specified weight or distance, in this case a DateTime distance of 10 seconds, which would then narrow the EXACT matches down to those that have a datetime stamp of <= 10 seconds. these results show in the multimatch output.

I'm confused as to what I should expect in the uniquematch output?

also when I try to test this by running a fuzzy Xref without any fuzzy logic specified, I would expect it to behave like a join, only spitting out the data where the keys match exactly and they would then be in the uniquematch output.
however, when I test this it seem that even if 1 field out of the key pair, matches 1 field in the other dataset key pair, then that is deemed to be a match??

Can someone explain precisely in as basic language as possible, how the fuzzy node works and what I should expect to get out with possibly an example to show this?

using the basic data below as a test. if I connect to an Xref I get the desired join of 2 rows (exact match of 1,2 and 2,9).... if I stick in a Fuzzy Xref without fuzzy logic I get 5 multimatches..... I don't know why. I know that it is obviously correct, I just don't understand what it is trying to do and tell me..... please can someone explain ?? much appreciated !

table1:

A:int, B:int,C:DateTime, D:String
1,2,2014-08-10 10:00:00,Red
2,9,2014-08-10 10:10:00,Yellow
3,4,2014-08-10 10:07:00,Blue
2,4,2014-08-10 10:08:00,Yellow
1,6,2014-08-10 10:15:00,Red
2,10,2014-08-10 10:09:55,Blue

table2:

XA:int, XB:int,Xy:DateTime,Xz:String
1,2,2014-08-10 10:10:00,Corvette
2,9,2014-08-10 10:02:00,Porsche
6,4,2014-08-10 10:07:00,fiat
9,4,2014-08-10 10:10:55,Ferrari
6,6,2014-08-10 10:34:00,honda
9,9,2014-08-10 10:23:55,ford

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: Labayawardana

Hi,

I came across this myself, I am relatively new hire at Lavastorm, so I might be wrong in my assumption. But I do have a solution for you regardless. The way I see it. The 'exact match' section of the fuzzy nodes work as 'OR' rather than a 'AND' when multiple data fields are included.

Simple solution is to concatenate the multiple fields in the exact match section. i.e match on strcat('A'," ",'B') and strcat('XA'," ",'XB'). This will give you 2 exact matches as expected. When you include both data fields as two parameters, it results in 5 multi matches because, it looks for exact matches on each field one at a time. I've included a sample graph using your data.

fuzzy_sample.brg

If someone from Product Dev can confirm this that would be great!.. but till then, I use the fuzzy match nodes as above depending on my requirements and works as expected.

Thanks

Note: This was originally posted by an inactive account. Content was preserved by moving under an admin account.

Originally posted by: ryeh

This is the second time this has come up in the past few days. Also see:
http://community.lavastorm.com/threa...n-in-Exact-Tab

I've checked with development. Right now it's only matching on the first key. A request has been made to extend the match to multiple keys.