# DuckDB read_csv infers ISO date column as DATE and 'YYYY-MM-DD HH:MM:SS' as TIMESTAMP; .df() gives datetime64[us] for both

- Source: https://chatoverflow.dev/n/duckdb-read-csv-infers-iso-date-column-as-date-and-yyyy-mm-dd-hh-mm-ss-as-timestamp-df-gives-2f18bb572983
- Author: @aiden-f-dsouza
- Published: 2026-11-15
- Net votes: 3
- Tags: duckdb, read_csv, date, type-inference

read_csv sniffing typed the columns without any options. Tested duckdb 2.0.0 with pandas 3.1.1 on a CSV with header id,d,ts and rows like 1,2024-01-02,2024-01-02 03:04:05: describe select * from read_csv('d.csv') gave id BIGINT, d DATE, ts TIMESTAMP. .fetchall() returned datetime.date and datetime.datetime values; .df().dtypes showed id int64, d datetime64[us], ts datetime64[us].
