Datasets and tables

How Loam organizes your data

A workspace contains datasets, tables, and views. Datasets keep related data together and can contain smaller datasets when you need another level of organization. Tables and views hold or present the data.

Datasets

A dataset groups related data. For example, you might create commerce, then stripe inside it, then production inside that. You see the same hierarchy in the sidebar and in SQL.

Create each parent before creating its child. Loam does not create missing datasets or an automatic default dataset.

Tables and views

A table holds rows and columns, like a spreadsheet. A view presents the result of a saved query. Both are leaves: they cannot contain other resources.

SQL paths match the hierarchy

Every SQL path is absolute from the workspace root. The workspace name is not part of the path. A table named charges inside commerce → stripe → production is queried as:

SELECT * FROM commerce.stripe.production.charges;

A table directly beneath the workspace root has a one-part name, such as charges. Loam does not search another dataset or insert a default part when a path is missing.

Creating a dataset

Choose New dataset at the workspace root or inside an existing dataset. The selected location becomes the parent.

You need Admin access on the parent to create or reorganize datasets. Editors can create and change tables and views in datasets where they have Editor access, but cannot change the dataset structure.

Adding a table

Select the dataset that should contain the table, then choose Add table. A CSV upload can create an ordinary table directly inside that dataset. Other supported formats load an existing table whose schema already matches the file.

See Your first upload for a step-by-step walkthrough.

Naming rules

• Names must be valid SQL identifiers. Quote names containing spaces, punctuation, uppercase letters, or reserved words.
• Unquoted names are normalized according to SQL identifier rules.
• A name must be unique among every dataset, table, and view with the same parent.
• A SQL path may contain up to 32 dataset components (called namespaces in SQL) plus its table or view name.