Set<Mixed> should not consider string and binary data equivilent

[sync-by-unito]

The MongoDB behaviour is to consider a string and a binary different even if they contain the same data, but in a Realm Set they are considered the same. Realm should align to the MongoDB behaviour. In practice this should be a rare case because most strings have a null byte at the end while binaries do not.

Set<Mixed> set;
set.insert(StringData("abc", 3));
set.insert(BinaryData("abc", 3));
REALM_ASSERT(set.size() == 2); // currently fails with size of 1

This change will require a file format bump, but no data migration will be needed.

[sync-by-unito]

➤ Jonathan Reams commented:

The issue here isn't so much that you'd do

Set<Mixed> set;
set.insert(StringData("abc", 3));
set.insert(BinaryData("abc", 3));
REALM_ASSERT(set.size() == 2); // currently fails with size of 1

from a realm SDK, It's that you'd do

db.update({...}, { $addToSet: { set: [ BinData("abc"), "abc" ] } });

from atlas.

[sync-by-unito]

➤ Jørgen Edelbo commented:

We should probably do this when we have to bump the file format version anyway. We will need to re-sort the Set columns.

[sync-by-unito]

➤ Ian Ward commented:

Would this require a breaking API change or anything public facing?

FYI - this feature is still in Beta, so it would be good to make any changes now rather than when we go out of Beta.

[sync-by-unito]

➤ Kiro Morkos commented:

One of our fuzz tests ran into this issue recently, is this going to be picked up soon?