WINDOW order by changes row order of table for no reason #20015

What happens?

I have two different WINDOWs in my select statement. If I order window2 then it orders the full table for some reason placing rows with a NULL value at the end before calculating window1: row_number() OVER () despite OVER () guaranteeing preservation of row order according to:
https://duckdb.org/docs/stable/sql/dialect/order_preservation#clauses

Sure I could do this in two steps, row_number() first, put it in a nested select statement and then do my calculations for window2 but that seems like a workaround for a bug.

To Reproduce

WITH test_data AS (
    SELECT ['', 'P diff. (psi)', '0', '1', '2', '3', '4'] AS str_list
)
SELECT
    row_number() OVER () AS row,
    TRY_CAST(str_col AS DOUBLE) AS elapsed_time,
    elapsed_time - LAG(elapsed_time) OVER (ORDER BY elapsed_time ASC) AS interval_s
FROM (
    SELECT
        UNNEST(str_list) AS str_col,
    FROM test_data
)

VS

WITH test_data AS (
    SELECT ['', 'P diff. (psi)', '0', '1', '2', '3', '4'] AS str_list
)
SELECT
    row_number() OVER () AS row,
    TRY_CAST(str_col AS DOUBLE) AS elapsed_time,
    elapsed_time - LAG(elapsed_time) OVER () AS interval_s
FROM (
    SELECT
        UNNEST(str_list) AS str_col,
    FROM test_data
)

They have a different row column/row order for some reason.

OS:

Windows 10 Pro 64bit

DuckDB Version:

1.4.2

DuckDB Client:

Python

Hardware:

Intel Core i3-6006U, 8GB RAM, integrated GPU

Full Name:

Felix Baumann

Affiliation:

student

Did you include all relevant configuration (e.g., CPU architecture, Linux distribution) to reproduce the issue?

• Yes, I have

Did you include all code required to reproduce the issue?

• Yes, I have

Did you include all relevant data sets for reproducing the issue?

Yes