Hi @bbastian ,

Unfortunately, as of now there is strict limitation regarding size - a variant column cannot contain a value larger than 16 MiB.

Variant support in Delta Lake | Databricks on AWS

And tbh you cannot compare the size of this json string to how it is represented internally in memory because there's a lot that is happening there (i.e JSON is tokenize into objects, arrays, keys, value).

In your case you have 3 options:

1. Store this columns as a JSON string. You can store data in a single string column using standard JSON formatting and then query fields in the JSON using : notation. JSON string supports string of arbitrary length

2. Use structs - the provide great performance on read but you lose flexibility (you need to define schema upfront)

3. Preprocess your data - it depends on your data, but you can try divide this array of items into 2 or 3 array and store them in separate variant columns. This way you can overcome 16Mib size limitation