Issues:MatchUp Object: Difference between revisions

From Melissa Data Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
Line 19: Line 19:


==Large KeyFile Size effect on Memory resources==
==Large KeyFile Size effect on Memory resources==
By default, MatchUp object allocates a large SetUserInfo, the unique identifier attached to built match key - 1024 bytes.  
By default, MatchUp object allocates a large SetUserInfo, the unique identifier attached to built match key - 1024 bytes.
 
See [[MatchUp Object:Best Practices|MatchUp Object Best Practices]] for override instructions.
See [[MatchUp Object:Best Practices|MatchUp Object Best Practices]] for override instructions.


This is for the ReadWrite Interface only.





Revision as of 17:02, 28 October 2015

← MatchUp Object



.NET wrapper namespace

build 2628
The namespace in the new mdMatchUp.cs classes is 'namespace MDClasses'. The previous wrapper namespace is 'namespace MelissaData' thus breaking backwards compatibility. Updating to the new wrapper will require you do change the namespace and recompile. We will rollback the namespace name in the next build.


Fuzzy Algorithms: Accurate Near , Frequency, and other Levenshtein based algorithms

Using the Accurate Near fuzzy algorithm on one or more components, running the same data set repeatedly could sporadically return different dupe counts.

This may not be apparent for a single run because the fuzzy algorithms do not return a percent difference for used algorithms, only returning a status whether the algorithm found a match between records.

We identified a new compiler issue and have also reviewed our other fuzzy algorithms which use similar computations and compiler variable initialization.

If you use any of the Fuzzy algorithms, please contact us and we will provide you with the available patch.


Large KeyFile Size effect on Memory resources

By default, MatchUp object allocates a large SetUserInfo, the unique identifier attached to built match key - 1024 bytes.

See MatchUp Object Best Practices for override instructions.

This is for the ReadWrite Interface only.


Fuzzy: Legacy Matchcodes

Legacy Matchcodes, imported from previous versions, allowed a Fuzzy: Near setting of '0'. This is incompatible with the current version and can cause an error. Using the interface to edit the matchcode by changing the Distance to 1 will resolve the problem


Fuzzy: First component with set distance missing dupes

Setting a distance for a first component forces the component to use the Intersecting deduper. This may result in records within a set distance to be put in different clusters, and therefore may never get compared.
Workaround: Use an exact algorithm in the first component and keep a distance component, if required further down the component list. This will prevent missed dupes (and give you better speed benchmarks.
Resolution: This may require an advanced change to the deduper. Development is aware of the issue and is exploring options.