`octa --export-schema`¶

Render a file's column schema as SQL DDL, a Pydantic model, a TypeScript interface, a JSON Schema document, or a Rust struct, and print it to stdout. Only the column list is read and no rows are serialised, so it's cheap even on large files.

Synopsis¶

octa --export-schema FILE [-t TARGET]
octa -e FILE [-t TARGET]

Flag	Required	Meaning
`-e FILE`, `--export-schema FILE`	yes	The file whose schema to export.
`-t TARGET`, `--target TARGET`	no	Output target (default `postgres`). See the table below.

-f / --format does not apply, because the output is source code, not a table. Like --convert, the schema-export action ignores it.

Targets¶

`--target` value	Output
`postgres` (default)	`CREATE TABLE` in the Postgres dialect, double-quoted identifiers.
`mysql`	`CREATE TABLE` in the MySQL dialect, backtick identifiers, `UNSIGNED` types.
`sqlite`	`CREATE TABLE` in the SQLite dialect, affinity types (INTEGER/REAL/TEXT/BLOB).
`databricks`	`CREATE TABLE` in the Databricks (Spark SQL / Delta) dialect, backticks, `STRING`.
`snowflake`	`CREATE TABLE` in the Snowflake dialect, double quotes, `VARCHAR` / `TIMESTAMP_NTZ`.
`pydantic`	Pydantic v2 `BaseModel` with annotated fields and the right imports.
`typescript`	`export interface` with `number` / `string` / `boolean` mappings.
`json-schema`	JSON Schema (draft 2020-12) with `properties` and `required`.
`rust`	`pub struct` with serde derives and chrono types.

The table / class / struct name is taken from the file stem (sample for sample.csv), sanitised and PascalCased where the target requires.

Examples¶

Postgres DDL (the default)¶

$ octa --export-schema sales.parquet
-- Generated by octa - Postgres dialect
CREATE TABLE sales (
    region TEXT,
    quarter TEXT,
    amount DOUBLE PRECISION,
    order_id BIGINT
);

Snowflake DDL¶

$ octa -e sample.csv -t snowflake
-- Generated by octa - Snowflake dialect
CREATE TABLE sample (
    id BIGINT,
    name VARCHAR,
    active BOOLEAN
);

Databricks DDL¶

$ octa --export-schema sample.csv --target databricks
-- Generated by octa - Databricks dialect
CREATE TABLE sample (
    id BIGINT,
    name STRING,
    active BOOLEAN
);

Identifiers are emitted bare. A column name that can't work bare, one with spaces, punctuation, or a leading digit, is quoted automatically ("…" for Postgres / SQLite / Snowflake, `…` for MySQL / Databricks).

A Pydantic model, redirected to a file¶

octa -e users.parquet -t pydantic > users_model.py

A Rust struct¶

octa -e events.parquet -t rust

Multi-table sources¶

For a SQLite / DuckDB / GeoPackage file the CLI exports the schema of the file's default table. To target a specific table inside a multi-table database, use the MCP export_schema tool, which accepts a table argument.

Type mapping¶

Octa stores column types as Arrow-style strings (Int64, Utf8, Float64, Timestamp(Microsecond, None), …). Each target maps them to its closest native type; Arrow types with no clean equivalent fall through to the target's TEXT-equivalent with a /* … */ comment so the output is never silently wrong. The full mapping table lives on the GUI Schema Export page, since the CLI, GUI, and MCP all share the same renderers.

octa --export-schema¶