KeyBy Stream(source is Kafka) based on 2 fields where the 2 fields can come in any order but belong to same group

message1: { otherfields: {...}, field1: "abc", field2: "def"}
message2: { otherfields: {...}, field1: "def", field2: "abc"}

In Flink, if we keyBy the stream coming from Kafka Topic and if we keyBy based on field1 and field2, then the above 2 messages will go to different keyed-state but I want them to go to same keyed-state.

My current code is something like:

streams
.keyBy(GroupByKeyGenerator())
.process(CustomProcessFunction())

One solution I have is to combine the 2 fields as one, order them ascending order and then use it as one key to keyBy the stream. This way both messages will go to same keyed-state. (key in both the cases will be abcdef)

But the issue is that the 2 keys are uuids and they are generated randomly also from the source.
What are the chances if I combine them and order them in ascending order, then they will be mapped to different category/key-group in same keyed-state?

For example:

message1:
field1: aef,
field2: bcd =>
combinedOrderedKey: abcdef

message2:
field1: abc,
field2: def =>
combinedOrderedKey: abcdef

The combinedKey will be same but they are not part of same group/keyed-state.

Please help me out with the solution that I can use. (some Hash function can be used?)

The simplest solution to above problem is:

1. Put both the keys in a list
2. sort the list
3. Put any delimiter between the keys(in sorted order)
4. This will act as a common key for the states.

Example:
message1: {key1: "bcf" , key2: "adf"}
message1: {key1: "adf" , key2: "bcf"}

In both cases, the key will be: adf_bcf
Steps:

1. [bcf, adf]
2. [adf, bcf]
3. adf_bcf
4. common key: adf_bcf