[Qgis-user] Relation (1-N) Unpredictable Result (with default options)

Tue Sep 15 08:50:19 PDT 2020

Andreas, 

Thank you for your interest in my reported problem. I have worked with this some today in hopes of opening an issue on GitHub. Unfortunately, the problem seems even more mysterious to me after further testing. 

I have 2 data data tables generated extracted from a fulcrumapp.com app. These 2 tables have a foreign key relationship by design, so they should work well in QGIS using the relation feature. After today's testing, I can tell you that they actually do work fine so long as I use a very small subset of the actual data. This is the mysterious part. For example, if I limit the parent table to only 10 records and then limit the child table to those records linked to the 10 parents, QGIS performs as expected. If I use all of the data (around 4000 parent records and 9000 child records), then I get an incorrect parent-to-child assignment for new child records that I try to create. 

With this being the case, I don't feel confident posting an issue on GitHub. I would have to anonymize the data to post on GitHub, and that would require further work on my end. If I have more time today or tomorrow, I will try to further investigate the problem to see if I can learn more about the behavior. 

-Jeffrey 

-- 
Jeffrey Durrence 
jeffrey.durrence at mcleanengineering.com 
McLean Engineering Company 
815 South Main Street 
Moultrie, GA 31768 
1.229.985.1148 (Voice) 
1.229.985.2248 (FAX) 
1.229.798.0480 (Mobile) 

From: "Andreas Neumann" <a.neumann at carto.net> 
To: "Jeffrey Durrence" <jeffrey.durrence at mcleanengineering.com> 
Cc: "qgis-user" <qgis-user at lists.osgeo.org> 
Sent: Tuesday, September 15, 2020 2:12:12 AM 
Subject: Re: [Qgis-user] Relation (1-N) Unpredictable Result (with default options) 

Hi Jeffrey, 

Can you please open an issue on Github ( [ https://github.com/qgis/QGIS/issues | https://github.com/qgis/QGIS/issues ] ) and include your DDL SQL statements you used for these tests - including all pkey/fkey and other constraints you used and a simple QGIS project file demonstrating the issue. It is easier to discuss such issues on Github. 

Maybe it is related to the fact that Comboboxes aren't properly initialized with NULL values, for cases that don't allow NULL values. 

This is important to investigate - as it is unacceptable that corrupt data is collected. 

Thanks, 

Andreas 

On 2020-09-15 04:32, Jeffrey Durrence wrote: 

Today I was trying to set up an edit form for a spatial table with a related, non-spatial table. I have done this many times in the past, but it's been a while. After using Project Properties to set up my relation, I made several attempts to add a new record in the "child" table. Each time I tried, the new record did not get the correct value for the field which relates the two tables. Existing records were displayed as expected and could be edited, but I could not add new features. 
I went to the docs to see if something had changed that might be affecting me. I downloaded sample shapefile data and used that to demonstrate that related records could be created with that data. I set up a clean database and QGIS project file and made several attempts with my PostGIS data. I did observe that some new options were available to me under the widget area of the child table. What had me confused was that with the default options selected (the way I used to do it), the "reference field" values generated for the "new" child records were real values -- just not at all the correct values. 
I tried several things with the PostGIS data, but what finally fixed the behavior for me was to make one change in the default Widget settings for the "reference field" in the child table properties. In my case, enabling the option to "Use a read-only line edit instead of a combobox" resolved the issue. When that is enabled, new child records get the correct value for the reference field. I don't completely understand what determined the inserted values when this option was not enabled, but I would think that enabling this by default would save a lot of confusion for users like me. I was quite surprised that I could not find folks talking about this in the usual places. Could be that there is something local to my environment, but I thought it worth mentioning should someone else stumble upon it. 
Using QGIS LTS (3.10.5) on Ubuntu 20.4 and Windows 10 64-Bit with PostGIS backend (observed same behavior with Postgres 10.14 / PostGIS 2.4.4 and Postgres 12.4 / PostGIS 3.0.0) 
-- 
Jeffrey Durrence 
jeffrey.durrence at mcleanengineering.com 
McLean Engineering Company 
815 South Main Street 
Moultrie, GA 31768 

_______________________________________________ 
Qgis-user mailing list 
[ mailto:Qgis-user at lists.osgeo.org | Qgis-user at lists.osgeo.org ] 
List info: [ https://lists.osgeo.org/mailman/listinfo/qgis-user | https://lists.osgeo.org/mailman/listinfo/qgis-user ] 
Unsubscribe: [ https://lists.osgeo.org/mailman/listinfo/qgis-user | https://lists.osgeo.org/mailman/listinfo/qgis-user ] 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-user/attachments/20200915/042a7c42/attachment.html>