When the columns are built for the table based on the result data, it iterates over all the rows for the current page, and builds the set of the available columns. But due to the fact that some attribute(s) might not present in the first rows, but be alphabetically higher than other attributes, but because it is added in the further iterations it is put in the end of the columns list rather than in the proper alphabetical order.
The same also applies to the #195 where the sorting will be random at least, but not alphabetical.
Factor that complicates the sorting is that primary keys should come always first.
Although looking into the official DynamoDB web UI, they also do the same sorting as the app does right now.
But still good to investigate the different behavior.
When the columns are built for the table based on the result data, it iterates over all the rows for the current page, and builds the set of the available columns. But due to the fact that some attribute(s) might not present in the first rows, but be alphabetically higher than other attributes, but because it is added in the further iterations it is put in the end of the columns list rather than in the proper alphabetical order.
The same also applies to the #195 where the sorting will be random at least, but not alphabetical.
Factor that complicates the sorting is that primary keys should come always first.
Although looking into the official DynamoDB web UI, they also do the same sorting as the app does right now.
But still good to investigate the different behavior.