Create a view from the canvas

The view designer turns a table on the project canvas into a saved view object. A dataflow diagram guides you left to right, from the source tables you pick columns from to the view node you are building. You stay on the canvas the whole time.

Open the designer

1. On the project canvas, right-click the table you want to build the view from.
2. Choose Create View.

The designer opens as a full-surface overlay, with the anchor table’s schema preselected. The layout: a canvas on the left and a two-tab panel on the right with Details and Preview tabs. With nothing selected on the canvas, Details opens on the View section, which holds the schema, the view name, and the as-of setting.

Start from a chat prompt

You can also start a view without selecting tables first: open Ask AI on the canvas and describe the view you need. The assistant proposes it as a card in the chat, and Open in view designer lands you in this same designer with the tables, filter, as-of setting, and any key columns already in place. See Create a view with Ask AI.

Name the view and set the basics

1. In the View section of Details, confirm or change the Schema the view will be saved into.
2. Type a Name for the view. The view node’s header on the canvas updates as you type, so you can watch the target name take shape; leave it blank and the node shows the placeholder vw_....
3. Set As of to the point in time the view should be evaluated at:
   • Current timestamp (the default) evaluates the view at query time.
   • Specific date and time opens a date and time picker and pins the view to that instant.
   • Custom expression accepts any SQL timestamp expression, such as a column reference like t.effective_date.

The setting only affects joins with a temporal mode; a view over plain tables ignores it.

Anchoring the view on a Data Vault hub prefills the Name from the organization’s dimension-view naming pattern, for example dim_customer, and sets the Schema layer to Data Mart. Anchoring on a link prefills the fact-view pattern instead, for example fact_orders. Both prefills are a starting point: change the name or the layer at any time before you save.

Add related tables

The Add related tables panel sits in the top-right corner of the canvas whenever DeltaVault has a suggestion. Choose a table to add it to the diagram with one click. When the designer is building a dimension or a fact, the panel is titled Related objects and groups its suggestions by kind.

The diagram flows left to right, and a newly added table pans into view so you can see where it landed.

When the view is anchored on a Data Vault object, the list shows vault suggestions first: the anchor’s satellites, its links, and the hubs a link connects. Add one and it joins in automatically on the shared hash key, with point-in-time handling already applied, so you never define that join by hand. Generic foreign-key suggestions still follow below the vault suggestions.

Read the diagram

Source tables sit on the left; the view you are building is the accented node on the right.

• A solid connector is an inner join and a dashed connector is a left join. The arrowhead points at the referenced key.
• Blue connectors also run from every table into the view node, showing the direction of data flow. These are decoration, not joins: they cannot be clicked.

Expand a table and pick columns

1. Click the chevron in a table’s header to expand it to every column, each showing its PK or FK marker.
2. Click + on a column row to add it to the view. The plus becomes a check once the column is added; removing the column from the view re-enables the plus.
3. Use the table’s menu for Add all columns to add every column at once.

Added columns inherit the source column’s data type. If a column name collides with one already in the view, DeltaVault auto-suffixes it, for example name_2.

Work with the view node

The view node lists one row per column, in projection order.

• The function icon marks a transformation or a derived expression.
• The warning icon marks a validation problem.
• Click a row to edit it in Details.
• Choose the Remove column control (the minus sign) on a row to remove it.
• Drag a row by its grip handle to reorder it, or double-click a row to rename it in place.
• Use Add column (the plus sign in the view node’s header) to add a derived expression column.

Edit a column in Details

Click a row in the view node to open it in the Details tab. Details shows:

• Column name: the target column name.
• Data type: the select for the column’s output type.
• The Transformation editor (Expression for a derived column): a SQL editor with autocomplete for the expression that produces the column.
• Insert macro: drops a macro reference into the transformation.
• Sources: chips showing exactly which source columns the view will record for this column.
• Inline validation messages when something needs fixing.

See the lineage on the canvas

Selecting a view column highlights the source columns it reads:

• On an expanded table, the matching rows light up.
• On a collapsed table, the table’s header highlights instead.

The Sources chips in Details and the canvas highlight always agree: whatever a chip lists is exactly what lights up when you select the row.

Use macros

1. In the Transformation editor, type @macro: to autocomplete macro names, or click Insert macro to pick one from a list.
2. Fill in the macro’s arguments. The canonical form is @macro:name(arguments).

Unknown macro names and wrong argument counts are flagged inline; until they are fixed the save button reads Save draft and its tooltip lists them. Macros stay readable in what you save and only expand into SQL when the view is built. See Macro library for the full list of macros and what each one does.

Adjust the joins

The joins DeltaVault proposes are derived from the table relationships recorded on the tables, whether those came from the source system, from AI discovery, or were drawn by hand: see Table relationships to check or correct one at the source.

• With a connector selected on the canvas, the Details tab becomes the Join designer: it shows the join’s Confidence, its Join type, its Key pairs (with Add key pair), its Temporal columns, and a Form / SQL toggle for hand-writing the clause. Back to view returns to the view’s own details.
• With nothing selected, the Join section lists every join in the view; click one to select that connector and open the Join designer.

Filter the rows

Open the Where section to narrow the result. Use the Form / SQL toggle to switch between the condition builder and a hand-written clause.

In Form mode:

1. Choose Add condition to add a row.
2. Pick a column from the dropdown. Columns are grouped by table, so you can find the one you want even when several tables share a name.
3. Pick an operator. The list shown depends on the column’s data type and uses friendly labels: equals, not equals, less than, at most, greater than, at least, matches, in list, is null, is not null. Matches only appears for text columns.
4. Enter a value. The field is typed from the column: a number input for numeric columns, a date picker for date columns, a true or false select for boolean columns, and a comma-separated list for in list. Is null and is not null need no value at all.
5. On the second row and every row after it, choose and or or to combine that row with the one above it.
6. Choose the x on a row to remove it.

Switch to SQL mode to hand-write advanced predicates the form cannot express, such as parentheses, functions, or a BETWEEN clause. If a filter was written in SQL and is too complex for the form to represent, Form mode shows a notice with a Clear filter button instead of the condition rows.

Preview the SQL

Open the Preview tab to see the full CREATE VIEW statement, with any macros already expanded into their underlying SQL. If a macro has a problem, it appears as an issue above the SQL instead of failing the preview.

Save

Save is disabled only while the view has no schema or no name. Any other problem, a duplicate column name, a lineage error, a blank derived expression, an unknown macro or an argument-count mismatch, turns the button into Save draft: the view is saved as a draft, and the button’s tooltip lists what is still outstanding. Once everything is valid, choose Save. Beside it, Save and map saves and goes on to mapping, and Cancel leaves the designer without saving.

Troubleshooting

• A column row shows a warning icon. Open it in Details and read the inline message; it names exactly what needs to change.
• The button reads Save draft. Hover it to read what is outstanding, then check every row that shows a warning icon, including derived expression columns that were left blank.
• The preview shows an issue naming a macro. The macro was renamed or deleted. Fix the reference in the transformation, or recreate the macro.