Incorrect implicit casting with duckdb 1.4.1

[AlexHelloWorld]

What happens?

When SET old_implicit_casting = true;, list type was casted to string unexpectedly, leading to incorrect results with list_extract. This seems to be a regression in 1.4.x as DuckDB 1.3.x works fine.

Reproduce:

import duckdb
import pyarrow as pa

data = {
    "str_col": ["1-2", "3-4", "a-z", None]
}
arrow_table = pa.table(data)

conn = duckdb.connect(":memory:")
conn.execute("SET old_implicit_casting = true;")
conn.register("input", arrow_table)

result1 = conn.execute("SELECT list_extract(string_split(str_col, '-'), 1) FROM input").fetch_arrow_table()
print(result1)

In DuckDB 1.4.0 and 1.4.1, we got

pyarrow.Table
list_extract(string_split(str_col, '-'), 1): string
----
list_extract(string_split(str_col, '-'), 1): [["[","[","[",null]] # instead of the first element of list, the first character "[" was returned

In DuckDB 1.3.2, we got

pyarrow.Table
list_extract(string_split(str_col, '-'), 1): string
----
list_extract(string_split(str_col, '-'), 1): [["1","3","a",null]]

To Reproduce

import duckdb
import pyarrow as pa

data = {
    "str_col": ["1-2", "3-4", "a-z", None]
}
arrow_table = pa.table(data)

conn = duckdb.connect(":memory:")
conn.execute("SET old_implicit_casting = true;")
conn.register("input", arrow_table)

result1 = conn.execute("SELECT list_extract(string_split(str_col, '-'), 1) FROM input").fetch_arrow_table()
print(result1)

OS:

linux

DuckDB Version:

1.4.1

DuckDB Client:

Python

Hardware:

No response

Full Name:

Hongyu Shi

Affiliation:

Benchling

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

[evertlammerts]

On v1.4.1:

❯ duckdb
DuckDB v1.4.1 (Andium) b390a7c376
Enter ".help" for usage hints.
Connected to a transient in-memory database.
Use ".open FILENAME" to reopen on a persistent database.
D SET old_implicit_casting = true;
D SELECT list_extract(string_split(str_col, '-'), 1) FROM (VALUES('1-2'), ('3-4'), ('a-z'), (NULL)) as t(str_col);
┌─────────────────────────────────────────────┐
│ list_extract(string_split(str_col, '-'), 1) │
│                   varchar                   │
├─────────────────────────────────────────────┤
│ [                                           │
│ [                                           │
│ [                                           │
│ NULL                                        │
└─────────────────────────────────────────────┘

On v1.3.2:

DuckDB v1.3.2 (Ossivalis) 0b83e5d2f6
Enter ".help" for usage hints.
Connected to a transient in-memory database.
Use ".open FILENAME" to reopen on a persistent database.
D SET old_implicit_casting = true;
D SELECT list_extract(string_split(str_col, '-'), 1) FROM (VALUES('1-2'), ('3-4'), ('a-z'), (NULL)) as t(str_col);
┌─────────────────────────────────────────────┐
│ list_extract(string_split(str_col, '-'), 1) │
│                   varchar                   │
├─────────────────────────────────────────────┤
│ 1                                           │
│ 3                                           │
│ a                                           │
│ NULL                                        │
└─────────────────────────────────────────────┘

[taniabogatsch]

I'll have a shot at this, since it's a duckdb/duckdb error (not Python). 👍