load

Generate a runnable DuckDB CREATE TABLE AS statement from CSV profiling — typed columns, safe casts.

Profile a CSV file and generate a ready-to-run DuckDB CREATE TABLE AS SELECT statement. Every column is cast to its detected type using the expression from FineType's taxonomy — no manual type mapping needed.

Usage

finetype load [OPTIONS] --file <FILE>

Options

Flag	Type	Default	Description
`-f`, `--file`	path	—	Input CSV file (required)
`--table-name`	string	derived from filename	Override table name
`-m`, `--model`	path	`models/default`	Model directory
`--sample-size`	integer	`100`	Maximum values to sample per column
`--delimiter`	character	auto-detect	CSV delimiter character
`--no-header-hint`	flag	—	Disable column name header hints
`--model-type`	string	`char-cnn`	Model type: `transformer`, `char-cnn`, `tiered`
`--sharp-only`	flag	—	Disable Sense classifier (Sharpen-only pipeline)
`--limit`	integer	`10`	Number of preview rows in trailing SELECT
`--no-normalize-names`	flag	—	Disable DuckDB `normalize_names` (preserve original column names)
`--enum-threshold`	integer	`50`	Cardinality threshold for ENUM columns

Examples

Generate a typed table from CSV

$ finetype load -f contacts.csv

-- Generated by FineType (250 types, Sense→Sharpen pipeline)
-- Source: contacts.csv (12 rows, 6 columns)
CREATE TABLE contacts AS
SELECT
    CAST(id AS BIGINT) AS id,
    name,
    email,
    strptime(created_at, '%Y-%m-%dT%H:%M:%SZ') AS created_at,
    ip_address,
    CAST(REGEXP_REPLACE(...) AS DECIMAL(18,2)) AS amount
FROM read_csv('contacts.csv', all_varchar=true);

SELECT * FROM contacts LIMIT 10;

Pipe directly into DuckDB

$ finetype load -f contacts.csv | duckdb my.db

The generated SQL reads the original CSV with all_varchar=true (so nothing is silently coerced), then applies explicit casts. The trailing SELECT * FROM ... LIMIT 10 gives you an immediate preview.

Custom table name

$ finetype load -f data/2024-q1-sales.csv --table-name quarterly_sales

By default the table name is derived from the filename (2024_q1_sales). Use --table-name to override.

What gets generated

The output is a single SQL script containing:

A CREATE TABLE AS SELECT that reads the CSV with all_varchar=true
A CAST or transform expression for each column, based on its detected type
A trailing SELECT * FROM <table> LIMIT <n> for preview

Columns that FineType classifies as plain VARCHAR are passed through without a cast. Every other column uses the x-finetype-transform expression from its type schema.