Modeling with Uploaded Files

Hashboard can directly use CSVs and other common data files for small or adhoc analysis, without needing to set up a traditional database.

Uploading files

From the model builder page (opens in a new tab), select "Uploads" as the data connection. You can then either click the "Upload" button or drag a file from your computer into the tables list. This will upload the file to Hashboard and make it available as if it were a table in a databse.

You can update the contents of an existing file by uploading or dragging in a new version with the same name.

Supported files

Hashboard currently supports the following formats:

• Comma separated or tab separated values files (.csv, .tsv)
• JSON files containing a single array of objects. Object keys will be used as column names. (.json)
• Apache Parquet (opens in a new tab) files (.parquet, .pqt)

Hashboard enforces a max size of 50 MB for uploaded files.

Writing SQL for uploaded files

Hashboard uses DuckDB (opens in a new tab) as the SQL-engine for uploaded files.

Hashboard loads your files into the DuckDB instance as a table with a name matching the file. Since filenames contain special characters, you will need to quote the table name.

SELECT * FROM "my_file.csv";

Hashboard is compatible with the DuckDB SQL dialect (opens in a new tab). However, side-effects from statements such as INSERT, UPDATE, and CREATE will be discarded. The DuckDB instance is re-created and discarded between every query.