Glossary

Base: for a changeset, the rows the changeset is measured from: what the workspace file started from, normally main's rows as the branch left them, or a branch or database the file was started again from. Not stored; worked out by taking the component document and undoing the changeset.

Basket: see Deployment basket.

Category: a top-level grouping of components in a workspace, e.g. Tables, Views, Procedures. Derived from the templates the workspace uses.

Changeset: the rows one work item changed in a data component, each as a before and after pair, kept at changesets/<work item>/<category>/<name>.json. Deploying it writes only those rows, after checking the target still holds each before. See Deploy only your rows. Not to be confused with a TFVC changeset, which is a check-in; the Add Components dialog calls the former DataChangesets for that reason.

Component: a single database artefact represented as a SQL file, or for a data template with OutputFormat="json" a data document, in the workspace.

Data component: a component that captures data in reference or configuration tables as a re-runnable script.

Data document: a data component stored as rows in a .json file rather than as a MERGE script, for a template with OutputFormat="json". It is inflated to the script when it deploys. See Store data as documents.

Deployment basket: the ordered set of components staged for the next release. Ephemeral until saved as a deployment file.

Deployment file: the saved, durable form of a basket. XML manifest that lists components, order, filters, and metadata.

Draft: a temporary extraction of a component from the database, held separately from the workspace copy. Used for review before overwriting.

Extract: reading a component from the database into the workspace (or into a draft).

Inflate: turning a data document or changeset into the SQL script it stands for. Happens when a deployment runs, when datastar --command-name build packages a release, or on request through View Inflated SQL.

MCP: Model Context Protocol. An open protocol for AI assistants to invoke tools over HTTP. DataStar embeds an MCP server.

Object component: a component that captures a schema object (table, view, stored procedure, function, trigger) as a CREATE script.

Reversal: the operation that undoes a prior deployment by applying its paired rollback script.

Release Workflow: the four-step wizard (Work Item → Commit → Deployment File → Push) that prepares a release.

Shared definition: the description of a data template's tables and columns, written once to definitions/<category>/<template>.<hash>.json and pointed at by every document the template generates, rather than repeated in each. The default for data documents. See Sharing one definition.

Template: an XML file that describes how DataStar should extract a particular kind of component from the database.

Work item: the unit of tracked work a release is tied to. DataStar uses this as a neutral term across trackers. It maps to a Jira story, task, or issue; an Azure Boards user story, task, or work item; or an equivalent ticket in whatever system you use.

Workspace: a project root on disk that DataStar knows about. Contains components, templates, and a .ds/ configuration folder.