We run into the following problem: When we connect a table with data about places with latitude and longitude from BigQuery to a Dataset (Query mode SPICE), latitude and longitude get sometimes wrong values in the analysis but correct values are shown in the dataset view. In BigQuery they are stored as float. We tried to cast them in the dataset as float, numeric and rounded numeric but this doesn’t help.
When we use a table with only 39,999 entries, we get it to work. If there are 40,000 entries, latitude and longitude become ‘null’ in the analysis, but are still correct in the dataset preview.
Then cutoff of 40,000 entries remains the same even if we change the the ordering, so it doesn’t seem to hinge on specific entries.
This problem only seems to occur if we use SPICE.
This seems to be a Quick Sight bug or a certain limit. Or is there any way to fix this?
Hi @DanielH,
I can’t seem to find any quota restrictions listed anywhere in relation to lat/long fields but it certainly sounds like there is one if you’ve tried changing the ordering and the various field types.
As this seems to be a set limitation, I will leave this marked as a feature request to promote visibility to the support team.
While this may not be an ideal work around, if you don’t need all points on the map to coincide with each other, maybe try and split up the rows into separate datasets and then incorporate each of them to your analysis. That way you can still utilize more than 40,000 rows, just not in one visual.
In Google BigQuery, the fields latitude and longitude contain floats.
In the QuickSight Dataset editor, they also contain decimal values with data type latitude and longitude.
In the analysis, the longitude and latitude data from SPICE are Null but also with the correct data type latitude and longitude.
A way to solve this problem is to rename the field num_la und num_lo in BigQuery with FLOAT as data type and then add calculated fields latitude and longitude in the QuickSight Dataset with the correct data types latitude and longitude. But this is very cumbersome when there are many changes to the Dataset.
Hi @DanielH,
Thank you for your follow up on this initial post and the additional information. I’ll replace the tag with ‘error’ instead to denote a bug. I would suggest creating a support ticket to report this behavior, in addition, I’ll try and raise the information to support as well