rank

DefinitionCopy Anchor Link

A window function that returns the rank of rows within a partition of a table. Tie values are assigned the same rank. The next rank in the sequence is not consecutive. E.g. 1, 1, 3, 4, 4, 6, ...

To get consecutive rank, use dense_rank.

Syntax

rank(order: order_expr)
rank(order: order_expr, ..., partition: partition_expr, ...)

rank(order: count(orders.id) | desc())
rank(order: count(orders.id), order: average(users.age))

// with partition
rank(order: count(orders.id), partition: orders.status)

// Axis-aware examples
rank(order: 'rows')                              // Rank by row/x-axis order
rank(order: 'columns' | desc())                  // Rank by column/legend order (descending)
rank(order: revenue | desc(), partition: 'rows') // Rank within each row

Input

• order (required, repeatable): A field that is used for ordering. The order defaults to ascending. The order can be set explicitly with asc() or desc(). You can also use axis references:
  • 'rows' or 'x_axis': Order by dimensions mapped to rows/x-axis
  • 'columns' or 'legend': Order by dimensions mapped to columns/legend
  • E.g. rank(order: count(orders.id) | desc()) or rank(order: 'rows')
• partition (repeatable, optional): A field that is used for partitioning the table. You can use axis references like 'rows', 'columns', 'x_axis', or 'legend'. E.g. rank(order: count(orders.id), partition: orders.status) or rank(order: revenue, partition: 'rows')

Output

Rank of the current row within its partition (if no partition is specified, the whole table is considered as a single partition). If two rows are tied for a rank, each tied rows receives the same rank. The next rank in the sequence is not consecutive. E.g. 1, 1, 3, 4, 4, 6, ...

Sample UsagesCopy Anchor Link

explore {
  dimensions {
    users.gender, users.age_group
  }
  measures {
    count(orders.id),
    _rank: rank(order: count(orders.id) | desc())
  }
}

explore {
  dimensions {
    users.gender, users.age_group
  }
  measures {
    count(orders.id),
    _rank: rank(order: count(orders.id) | desc(), partition: users.gender)
  }
}

Using rank in a dimension definition is straight-forward. Remember that the rank is calculated against all rows in the model.

dimension ranking_by_age {
  label: 'Rank By Age'
  type: 'number'
  hidden: false
  description: ''
  definition: @aql rank(order: users.age | desc()) ;;
}

You can see that users with the same age are assigned the same rank

explore {
  dimensions {
    users.age,
    users.ranking_by_age
  }
}

You can also use it in filter() to filter out the top 10 users by age:

users 
| select(users.full_name, _rank: rank(order: users.age))
| filter(_rank == 1)

See alsoCopy Anchor Link

• dense_rank(): same value gets same rank, no gaps
• percent_rank()
• ntile()
• Window Functions