Read a table from the command line

Use the CLI to see a Delta table's columns or save some of its rows to a file. This quickstart starts with a local table and reads at most ten rows.

Before you start

Install the CLI and have a Delta table available. Replace /path/to/table below with the table's directory, which contains _delta_log, rather than a single Parquet file. Run the shell examples in Bash.

The examples use the standalone command, dar. If you installed the Python package, use delta-arrow-reader in its place. Both commands accept the same arguments.

See the table's columns

dar inspect /path/to/table

The command prints JSON with the table's version in table_version and its columns in schema.fields. It reads the table's metadata without reading its rows. Use the column names from this output when selecting columns later.

Save a small result

dar scan --limit 10 /path/to/table > sample.arrow

This saves up to ten rows, with all columns, to sample.arrow. The limit keeps your first result small. Row order is not guaranteed.

The file contains an Arrow IPC stream, a binary format that Arrow libraries can read. It is not CSV or text, so scan requires output to a file or pipe instead of printing rows in the terminal.

Only use the file if the command succeeds. If it reports an error, discard the file: a failed scan can leave an incomplete result.

Read the saved result

Open the file with an Arrow IPC stream reader in your application. For example, if you have Python and PyArrow installed:

import pyarrow.ipc as ipc

with ipc.open_stream("sample.arrow") as reader:
    print(reader.read_all().to_pydict())

This prints the saved rows as a dictionary of columns. Collecting the result in memory is convenient for this small sample. For larger results, process one batch at a time.

Choose columns

Once the first scan works, repeat --column to select the columns you need:

dar scan --column id --column region --limit 10 /path/to/table > selected.arrow

Replace id and region with names from your table. The output contains those columns in the order you requested. Omit --limit when you are ready to read all rows.

Where to go next

• Filter rows when you only need rows matching a condition.
• Read a table from storage such as S3, with storage options when your environment needs them.
• Use a container or call the CLI from Go when adding the reader to a service.
• Look up flags, output formats, and exit statuses in the CLI reference.

For help in the terminal, run dar inspect --help or dar scan --help.