Skip to main content

Celonis Product Documentation

LAG
Description

LAG returns the row that precedes the current row by offset number of rows. It is possible to specify a column based ordering and partitioning. Null values are skipped.

Syntax
  LAG ( column [, ORDER BY ( sort_column [sorting], ... )] [, PARTITION BY ( partition_column, ... )] [, offset ] )
 
  • column: The source column where preceding rows are taken from.

  • sort_column: Optional sorting column to specify an order.

  • sorting: Each of these columns can have an optional tag specifying the ordering of the column. Default is ascending:

    • ASC: Ascending order

    • DESC: Descending order

  • partition_column: Optional partition column to specify groups in which LAG should operate.

  • offset: The number of non-NULL rows preceding the current row. The default value is 1.

Examples

[1]

Simple example of LAG returning the preceding row. If a row does not have a preceding row NULL is returned.

Query

Column1

         LAG ( "Table1"."column" )
        

Input

Output

Table1

column : string

'C'

'D'

'A'

'E'

'B'

Result

Column1 : string

null

'C'

'D'

'A'

'E'

[2]

LAG with an offset.

Query

Column1

         LAG ( "Table1"."column" , 3 )
        

Input

Output

Table1

column : string

'C'

'D'

'A'

'E'

'B'

Result

Column1 : string

null

null

null

'C'

'D'

Ordering

One or more columns can be given to specify an ordering. This tells the LAG function what the preceding element actually is. Optionally every column can be tagged as ascending or descending.

[3]

LAG with a single order column.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."order" ) )
        

Input

Output

Table1

column : string

order : int

'A'

3

'B'

2

'C'

5

'D'

1

'E'

4

Result

Column1 : string

'B'

'D'

'E'

null

'A'

[4]

LAG with a single order column tagged as descending.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."order" DESC ) )
        

Input

Output

Table1

column : string

order : int

'A'

3

'B'

2

'C'

5

'D'

1

'E'

4

Result

Column1 : string

'E'

'A'

null

'B'

'C'

[5]

LAG with multiple order columns.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."order1" , "Table1"."order2" ) )
        

Input

Output

Table1

column : string

order1 : int

order2 : string

'A'

3

'z'

'B'

1

'z'

'C'

1

'y'

'D'

2

'y'

'E'

2

'x'

Result

Column1 : string

'D'

'C'

null

'E'

'B'

Partitioning

The partition columns specify groups. The LAG function operates independently within every group. This means when an ordering is given it is applied within every group and the last offset elements in every group have a lagging value of null.

[6]

LAG with a single partition column.

Query

Column1

         LAG ( "Table1"."column" , PARTITION BY ( "Table1"."Country" ) )
        

Input

Output

Table1

column : int

Country : string

1

'Germany'

2

'Germany'

3

'Germany'

1

'USA'

2

'USA'

Result

Column1 : int

null

1

2

null

1

[7]

LAG with multiple partition columns.

Query

Column1

         LAG ( "Table1"."column" , PARTITION BY ( "Table1"."Country" , "Table1"."State" ) )
        

Input

Output

Table1

column : int

Country : string

State : string

1

'Germany'

'Bavaria'

2

'Germany'

'Berlin'

2

'Germany'

'Bavaria'

3

'Germany'

'Bavaria'

1

'USA'

'California'

Result

Column1 : int

null

null

1

2

null

[8]

LAG with multiple partition columns.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."year" ) , PARTITION BY ( "Table1"."Country" , "Table1"."State" ) )
        

Input

Output

Table1

column : int

year : date

Country : string

State : string

1

Tue Dec 31 2002 00:00:00.000

'Germany'

'Bavaria'

2

Fri Dec 31 1999 00:00:00.000

'Germany'

'Berlin'

2

Mon Dec 31 2001 00:00:00.000

'Germany'

'Bavaria'

3

Fri Dec 31 1999 00:00:00.000

'Germany'

'Bavaria'

1

Fri Dec 31 1999 00:00:00.000

'USA'

'California'

Result

Column1 : int

2

null

3

null

null

NULL handling

The lagging value for a NULL value is the same value as the lagging value for the next non-NULL value. The offset parameter counts only non-NULL values.

[9]

Example for LAG applied to a column including NULL values.

Query

Column1

         LAG ( "Table1"."column" , PARTITION BY ( "Table1"."Country" ) )
        

Input

Output

Table1

column : int

Country : string

1

'Germany'

null

'Germany'

3

'Germany'

4

'Germany'

null

'USA'

1

'USA'

2

'USA'

Result

Column1 : int

null

1

1

3

null

null

1

[10]

Offset ignores NULL values.

Query

Column1

         LAG ( "Table1"."column" , PARTITION BY ( "Table1"."Country" ) , 2 )
        

Input

Output

Table1

column : int

Country : string

1

'Germany'

null

'Germany'

3

'Germany'

4

'Germany'

1

'USA'

null

'USA'

null

'USA'

4

'USA'

5

'USA'

Result

Column1 : int

null

null

null

1

null

null

null

null

1

Advanced Examples

[11]

LAG can be used to simulate ACTIVITY_LAG.

Query

Column1

         LAG ( "Table1"."activity" , ORDER BY ( "Table1"."timestamp" ) , PARTITION BY ( "Table1"."case" ) )
        

Input

Output

Table1

case : int

activity : string

timestamp : date

1

'A'

Mon Feb 01 2016 01:00:00.000

1

'B'

Mon Feb 01 2016 02:00:00.000

1

'C'

Mon Feb 01 2016 03:00:00.000

2

'A'

Mon Feb 01 2016 01:00:00.000

2

'B'

Mon Feb 01 2016 02:00:00.000

2

'C'

Mon Feb 01 2016 03:00:00.000

2

'D'

Mon Feb 01 2016 04:00:00.000

Result

Column1 : string

null

'A'

'B'

null

'A'

'B'

'C'

[12]

It is possible to combine an arbitrary amount of order columns with an arbitrary amount of partition columns.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."year" DESC , "Table1"."value" ) , PARTITION BY ( "Table1"."Country" , "Table1"."State" ) )
        

Input

Output

Table1

column : int

year : date

value : float

Country : string

State : string

1

Tue Dec 31 2002 00:00:00.000

1.0

'Germany'

'Bavaria'

2

Fri Dec 31 1999 00:00:00.000

2.0

'Germany'

'Berlin'

2

Mon Dec 31 2001 00:00:00.000

3.0

'Germany'

'Bavaria'

3

Fri Dec 31 1999 00:00:00.000

4.0

'Germany'

'Bavaria'

1

Fri Dec 31 1999 00:00:00.000

5.0

'USA'

'California'

1

Sun Dec 31 2000 00:00:00.000

6.0

'Germany'

'Berlin'

5

Tue Dec 31 2002 00:00:00.000

7.0

'USA'

'California'

6

Mon Dec 31 2001 00:00:00.000

8.0

'Germany'

'Berlin'

4

Tue Dec 31 2002 00:00:00.000

9.0

'Germany'

'Bavaria'

7

Wed Dec 31 2003 00:00:00.000

10.0

'USA'

'California'

Result

Column1 : int

null

1

4

2

5

6

7

null

1

null

[13]

All optional parameters can be combined.

Query

Column1

         LAG ( "Table1"."column" , ORDER BY ( "Table1"."year" DESC , "Table1"."value" ) , PARTITION BY ( "Table1"."Country" , "Table1"."State" ) , 2 )
        

Input

Output

Table1

column : int

year : date

value : float

Country : string

State : string

1

Tue Dec 31 2002 00:00:00.000

1.0

'Germany'

'Bavaria'

2

Fri Dec 31 1999 00:00:00.000

2.0

'Germany'

'Berlin'

2

Mon Dec 31 2001 00:00:00.000

3.0

'Germany'

'Bavaria'

3

Fri Dec 31 1999 00:00:00.000

4.0

'Germany'

'Bavaria'

1

Fri Dec 31 1999 00:00:00.000

5.0

'USA'

'California'

1

Sun Dec 31 2000 00:00:00.000

6.0

'Germany'

'Berlin'

5

Tue Dec 31 2002 00:00:00.000

7.0

'USA'

'California'

6

Mon Dec 31 2001 00:00:00.000

8.0

'Germany'

'Berlin'

4

Tue Dec 31 2002 00:00:00.000

9.0

'Germany'

'Bavaria'

7

Wed Dec 31 2003 00:00:00.000

10.0

'USA'

'California'

Result

Column1 : int

null

6

1

4

7

null

null

null

null

null

Use Cases
See also: