I've been trying to work around this issue for a couple of days by now without success so far.

Imagine you have these two dummy datasets

dataset_1

id,latitud,longitude
1,-0.023437,-0.070068
2,-0.069099,-0.069099

dataset_2

id,name
1,"site one" 
2,"site two"

and you want to JOIN them by id. This is very straightforward with the QuickSight dataset editor. The issue happens when you change the data type of latitud and longitude to their geospatial type, since the error shown in the title pops-up and won't let you save the dataset.

The weird thing is that the error suggests the fields latitude and/or longitude being used to make the JOIN instead of id.

Before contacting AWS for a possible bug have anyone had and solved this issue before?

CodePudding user response：

This seems to be a limitation with quicksight. Can you do the join in custom sql and then add the fields as geographical coordinates in data preparation?

CodePudding user response：

At the end we contacted AWS support. It seems they have this feature in consideration but it's still not addressed. They suggested us a work-around though :

1. Change the datatype of the Geo-spatial field to string and perform the join

2. Once the join is successful, go back to the dataset page, click on the dataset and select "Use in a new Dataset" option

3. This will create a new child dataset for the main dataset

4. Here you can change the datatype back to Geo-spatial and save it

Have in mind that the option Use in a new Dataset is disabled if your dataset have Row-level security or if it exceeds 3 levels of JOIN (in which case you'd have to follow @darcoli's answer first)