Skip to main content
Controls how a query can use a dimension that belongs to only a subset of member topics. This applies to any dimension that isn’t a shared dimension and isn’t on a shared view. By default, these dimensions can only act as filters on a query. To make the dimensions selectable for the query, you can decide whether measure values should be represented as NULL or repeated when the dimension is not shared across all topics.

​
Syntax

​
Properties

​
string
default:"filter_only"
  • filter_only - Default. A member topic’s own dimensions appear in the field picker as filter-only fields. A query that selects one anyway, for example through the API, returns an error.
  • null_fill - A member topic’s own dimensions can be selected in the field picker. The other member topics’ measures appear once per group, on a row where the dimension is NULL.
  • repeat - A member topic’s own dimensions can be selected in the field picker. The other member topics’ measures repeat on every row of the dimension.
When a topic-scoped dimension is selected under null_fill or repeat, its column header in the results table shows an info icon that explains which mode is active.

​
Examples

The examples below use a composite topic with two member topics:
  • orders: one row per order, with status (complete, shipped) and sale_price. Joins to users.
  • users: one row per user, with country.
users is also a shared view, so country is available from both topics. status exists only in orders.
Each example runs the same query: country, status, sale_price_sum, and user_count. Only the unrelated_dimension_handling value changes.

​
filter_only (default)

status is filter-only, so the query returns an error. Removing status from the selection returns:

​
null_fill

user_count lands on one extra row per country where status is NULL. sale_price_sum is NULL on that row because it’s already accounted for on the status rows. Each measure appears once per country, so every column still sums correctly.

​
repeat

user_count repeats on every status row for its country. Summing it across rows overcounts it, so use repeat when the value is meant to be read per row, for example as the denominator in a ratio.

​
Limitations

  • A NULL value coming from the table is indistinguishable from a filled-in row. With null_fill, a value that is genuinely NULL in the data gets the same key as fill row. In the example above, orders with a NULL status have the key US | NULL, so their sale_price_sum shows up in the US | NULL row next to user_count. The result gives no way to tell whether that NULL is a real value or a fill. Filter out NULL values of the dimension if you need the fill row to be unambiguous.
  • Pivots, fill fields, and totals on topic-scoped dimensions aren’t currently supported. Pivoting on a topic-scoped dimension, filling its missing values, or requesting totals in a query that selects one returns an error.
  • The mode applies to every member topic. It’s set once on the composite topic and covers all member topics, including imported ones.