Upload Parquet Files

Steps

Open the dataset containing the target table.
Click Add table next to the dataset name in the sidebar.
Choose or drag in your Parquet file.
Choose append, upsert, or replace, then select the existing target table. Upsert also requires one or more key columns.
Start the upload.

Loam reads the column types stored in the Parquet file footer and validates them against the existing table. Creating a new table from the upload currently requires CSV because create-if-missing needs a declared initial schema.

Why Parquet

Parquet stores a type for every column, so Loam does not have to infer types from sample rows. Use Parquet when you need to preserve supported column types from your file.

Parquet uses compression and is often smaller than CSV containing the same data.

Supported types

Loam accepts Parquet columns that map to these supported types:

Signed and unsigned integers
32-bit and 64-bit floating-point numbers
Text
True/false values
Dates and timestamps
16-byte UUIDs

Nested and other unsupported types cause the upload to fail.

File requirements

Maximum file size: 100 MB.
Must be a valid Apache Parquet file (not a CSV renamed to .parquet).

When to use Parquet vs. CSV

Use Parquet when	Use CSV when
You need to preserve supported column types	You're exporting from a spreadsheet
You want compression for a larger file	You want to edit the file in a text editor first
You're exporting from a data tool (dbt, Spark, pandas)	You don't have a Parquet file

See Uploading CSV files for CSV-specific details and File size limits for more on upload limits.

What to expect

You'll see upload progress while the file transfers, followed by a processing status while Loam validates and loads it. Keep the tab open until the run finishes.

After the upload completes, you can immediately query the data in the SQL Console (sign in to open this page) or open Explore from the table page.