Performance of scripted vs. regular field mapping for reference fields in transform maps

[Max Nowak]

Hi,

I'm curious about the overhead of using a scripted field mapping for a reference field in a transfor map. Just thinking about it, I feel like a scripted field mapping might have better performance if the table I'm referencing is relatively large, and I'm able to reduce the number of records to search through a gliderecord query.

For example, just for fun, imagine I'd be importing into a table that references the CMDB, and for whatever reason, the field is not referencing a specific CMDB table, but just cmdb_ci. Now, that table could be very large, possibly millions of records.

Now, let's say I know that the records I want to import only ever reference a specific CI Class, AD Controllers ("cmdb_ci_ad_controller"). While I have a few AD Controllers, it's certainly much, MUCH less than the millions of records in the whole CMDB.

In that case, if I'd use a scripted field mapping like this:

answer = (function transformEntry(source) {
  var adcGr = new GlideRecord('cmdb_ci_ad_controller'); // Narrowing the search down to only AD Controllers
  var found = adcGr.get('name', source.u_name); // Assume that I can safely identify records by name

  if(found) {
    return adcGr.getUniqueValue();
  }
})(source);

Would that increase the performance, since I narrowed the search space down from millions of records to less than 100, or is the overhead from using a scripted field mapping so great that it doesn't matter?

Thanks for your help,

Max

[Mark Manders]

In this case always use a script. You are getting a string and need to find the reference. The chance that there is more than one CI with the same name, is theoretically not small. So you want the correct table. I have never tested the difference in performance, but the chances of getting the wrong CI would be too large to not use a script.