Bug: The alphabetical order of the properties of an object-based Record ID is query-significant

[ioannist]

Describe the bug

REMOVE TABLE wallet;

UPDATE wallet:{
    a: 0,
    b: 0,
    c: 0,
    d: 5
};
UPDATE wallet:{
    a: 0,
    b: 1,
    c: 0,
    d: 6
};
UPDATE wallet:{
    a: 0,
    b: 1,
    c: 1,
    d: 7
};
UPDATE wallet:{
    a: 1,
    b: 0,
    c: 1,
    d: 8
};
UPDATE wallet:{
    a: 0,
    b: 0,
    c: 0,
    d: 9
};

RETURN array::len(SELECT * FROM wallet:{a:0, b:0, c:0, d: 9}..={a:1, b:1, c:1, d: 10});


REMOVE TABLE wallet;

UPDATE wallet:{
    d: 0,
    c: 0,
    b: 0,
    a: 5
};
UPDATE wallet:{
    d: 0,
    c: 1,
    b: 0,
    a: 6
};
UPDATE wallet:{
    d: 0,
    c: 1,
    b: 1,
    a: 7
};
UPDATE wallet:{
    d: 1,
    c: 0,
    b: 1,
    a: 8
};
UPDATE wallet:{
    d: 0,
    c: 0,
    b: 0,
    a: 9
};

RETURN array::len(SELECT * FROM wallet:{d:0, c:0, b:0, a: 9}..={d:1, c:1, b:1, a: 10});

Steps to reproduce

Execute above code.

The first array has length 4 and the second has length 1.

The datasets are exactly the same with the property names reversed, which implies that the value lookup happens in the alphabetical order of the properties.

Expected behaviour

The arrays should have equal lengths. Alternatively, the alphabetical ordering should be documented.

SurrealDB version

1.5

Contact Details

No response

Is there an existing issue for this?

• I have searched the existing issues

Code of Conduct

• I agree to follow this project's Code of Conduct

[Dhghomon]

You're right that objects are alphabetically sorted and in fact we were just discussing adding that documentation last week. One PR is on its way to clarify that:

https://github.com/surrealdb/docs.surrealdb.com/pull/534/files

Another one will be coming up on sorting and how comparisons/range scans can be done on all values regardless of the type. (Which is why you sometimes see record range scans starting from NONE, which is the absolute starting point)