Using contains filter in DynamoDB scan with Java

Question

Suppose I have a working query such as:

ScanRequest scanRequest = new ScanRequest()
            .withTableName("myTable")
            .withFilterExpression("attr1 = :val1 and attr2 = :val2")
            .withExpressionAttributeValues(vm)  //contains values for :val1 and :val2
            .withLimit(PAGE_SIZE)
            .withConsistentRead(READ_TYPE);

Now I would like to extend this scan. Suppose my table also has an attribute attr3 which has the form:

"attr3": {
    "S": "AAA BBB CCC DDD"
}

How can I filter elements who's attr3 contains AAA? Or AAA and BBB?

Answer 1

The DynamoDB Condition Expressions Reference documentation is your friend!

In your specific case you can use the contains function to search for a sub-string. Your filter expression might look something like this:

"attr1 = :val1 and attr2 = :val2 and (contains(attr3, :val3a) or contains(attr3, :val3b))"             
// where :val3a and :val3b are value placeholders for say AAA and BBB

But I suspect that what you want to achieve is a bit more sophisticated than can be handled server-side by DynamoDB filters, so you have two options:

1. do the filtering in the application logic (ie. bring down your results to the client and filter there), or;
2. change your attribute from a string to a list, or string set (if duplicates not allowed)

In either case, you should know that a scan with filters is only more efficient than a regular scan in terms of network band-with (in the case when most results are excluded by the filter). But in terms of capacity consumed a Scan is equally expensive with or without filters. So if you can avoid it, don't rely on scanning your table too much!