2013/06/03 21:43 +, Rick James
Soundex is the 'right' approach, but it needs improvement. So, find an
improvement, then do something like this...
Hashing involves somekind normalizing, and in my case I see no means to it;
otherwise I would not have considered something so
2013/06/03 18:38 +0200, Hartmut Holzgraefe
equality checks have a linear cost of O(min(len1,len2)) and can make
use of indexes, too, while Levenshtein cost is is almost quadratic
O(len1*len2) and can't make any good use of indexes ... even using
a C UDF would help only so far with this ki
I will second Rick's approach and have implemented something very
similar for a client when soundex feel short of expectation. It
worked very well.
On Mon, Jun 3, 2013 at 5:43 PM, Rick James wrote:
> Soundex is the 'right' approach, but it needs improvement. So, find an
> improvement, then do
Soundex is the 'right' approach, but it needs improvement. So, find an
improvement, then do something like this...
Store the Soundex value in a column of its own, INDEX that column, and JOIN on
that column using "=". Thus, ...
* You have spent the effort to convert to Soundex once, not on every
On 03.06.2013 17:29, h...@tbbs.net wrote:
> I wish to join two tables on likeness, not equality, of character strings.
> Soundex does not work. I am using the Levenstein edit distance, written in
> SQL, a very costly test, and I am in no position to write it in C and link it
> to MySQL--and join
- Original Message -
> From: h...@tbbs.net
>
> I wish to join two tables on likeness, not equality, of character
> strings. Soundex does not work. I am using the Levenstein edit
> distance, written in SQL, a very costly test, and I am in no
> position to write it in C and link it to MySQL-
