ArcGIS Pro Create Locator fails

[JoeBorgione]

Pro 2.4

The Create Locator tool with a Point Address role, bails out on a feature class of roughly 500K address points:

Messages
Start Time: Thursday, July 11, 2019 9:58:35 AM
Start building locator...
Start transforming data...
Start processing Point Address role
Error happened while transforming data
Failed to execute (CreateLocator).
Failed at Thursday, July 11, 2019 9:58:55 AM (Elapsed Time: 19.91 seconds)

Not much to work with with respect to error messages.  This data resides in in and enterprise gdb, and it bailed out  using the egdb as a source, so I saved it to a fgdb and bailed out there as well.  Anyone else using this new approach to creating a locator?

edited to add: the create address locator tool in 2.4 works just fine...

Kory Kramer‌

[JoeBorgione]

Michael-  I haven't gone back to it as of yet.  I've been re-working a python script that uses 10.6.1 locators.  Currently I'm trouble shooting why my Alt Names table with centerlines fails  miserably. (50% + missed....  argh...)

The source data that was giving me the issues is in a Enterprise geodatabase (aka SDE).

I ran a ton of sample batches where I narrowed the number of points, trying to find the offending value; when it was all said and done, I never found it.  But I do appreciate you posting this as I'm totally into personal vindication...   

Good luck!

Shana Britt‌ : kind of weird that there is another one so soon, eh?

[MichaelVolz]

Why would the data need to be transformed?  I don't recall that being necessary with earlier address locator versions?

[ShanaBritt]

Joe and Michael:

It would be best that you submit your data through Support so that a bug can be logged. Without the data and the documented bug we cannot star investigating this behavior during the next ArcGIS Pro development cycle and identify why it's failing. I've tested the Create Locator tool in Pro 2.4with different datasets and not encountered this error.

[JoeBorgione]

I'll submit a case later today....

eta: Just created a case....

[EricAnderson17 · Accepted solution]

Hi Joe,

Just to follow up and close the loop in this thread, I took a look at your data with the analyst assigned to your Support case, and the issue is related to null geometry in your data. Running Check Geometry on the data shows two records with null geometry:

Running Repair Geometry on the feature class and then attempting to run the Create Locator tool again was successful.

-Eric

[MichaelVolz]

Eric:

I am having the same error occur when using the Create Locator tool with a Point Address role.  I ran the Check Geometry tool on my data in ArcMap and found that most, but not all, unmatched records had Null geometry.  As such I do have a few questions about the Create Locator tool in Pro compared to the Create Address Locator tool in both Pro and ArcMap:

1.) Is the validation for the Create Address Locator at a lower threshold as Null geometry does not crash the creation of a point addess locator with this tool in either Pro or ArcMap?

2.) Why would the Check Geometry tool only catch some of the unmatched records where none of the unmatched records would have a point associated with them, so they would all have Null geometry?

Your feedback is greatly appreciated.

[EricAnderson17]

Hi Michael,

I think I'm a little confused by your use of the word "unmatched". Are we talking about records that are coming back as unmatched when going to batch geocode with the locator? The error Joe was getting was during the locator creation (Create Locator tool), and thus he was never actually able to use the locator to perform batch geocoding. 

I agree that there might be some room for improvement here, if the Create Address Locator tool allows for the locator to be created successfully even with null geometry, while the Create Locator tool does not. This is something I can test on my end, but do you know what happens when you attempt to search for one of those records that had null geometry with the locator outputted from the Create Address Locator tool? 

Best regards,

Eric

[MichaelVolz]

Eric:

The unmatched records are from a prior process that is used to generate the points that are used as a source for the point address locator.  If my org continues with this process using Pro (unlikely as we are looking to change the way we create and maintain address points in the near future), it looks like we would need to add an additional step to our python pre-processing to remove these unmatched records so the Create Locator tool new with Pro can finish without crashing.  I was just thrown off by the fact that the same data can successfully generate an address locator with the Create Address Locator tool in either ArcMap or Pro.

Also, the Details pane of a geoprocessing tool such as Check Geometry only shows a subset of the bad records that were detected (Not sure if this is by design by ESRI to minimize the amount of data getting sent to this pane), but the output table has all of the bad records.  Confusing, but not a big deal.

Thanks again for your responses on this thread as it showed the root cause of the issue.

[EricAnderson17]

Got it, thank you for the clarification Michael! 

To skip ahead, the issue you are reporting about the Details pane/pop-up not showing all the records that were picked up by the Check Geometry tool (but all of them are present in the output table), definitely seems like a usability issue to me. Would you be able to work with our Support team to look into this further?

I'll talk a bit offline with some members of our Geocoding team to see how we can improve the null geometry issue. I would argue that I would rather have the Create Locator tool error out than produce a locator that could potentially have issues down the road if some records had null geometry, but maybe the error messaging could be improved since this was a pain point for both you and Joe.

Cheers,

Eric

[JoeBorgione]

Thanks Eric- we have a state holiday today (July 24th) and I'll be back to the office on the 25th.

Two questions still linger:

1. Seems like the original check and repair geometry tools did not work on an EGDB. Is that still the case?

2. I assume I can script those two tools, at least for a FGDB?