Formulas have multiple uses in Pentaho Metadata.

The first use of formulas within Pentaho Metadata is in the constraint definition of a Metadata Query, also known as MQL. A constraint function references business table columns and uses various comparison operators to determine which subset of data the business user is interested in.

The second use is in the definition of Physical Table Columns. In addition to Physical table columns mapping directly to a database table column, physical table columns defined in Pentaho Metadata may also be defined as a formula. This allows for combining of multiple columns into a single column, and also for doing more advanced aggregate calculations within aggregate table definitions.

The third use is in the definition of complex joins within business model relationships. This allows for multiple key joins as well as other logic when joining tables.

The fourth use is row level security.

Under the covers, Pentaho Metadata uses JFreeReport's libFormula package for interpreting formulas. The goal is to support OpenFormula syntax within the Metadata environment. Formulas are first interpreted by libFormula, and then within the Metadata system are converted to native SQL depending on the type of database used.

Here is an example of an MQL Constraint formula:

OR([BT_CUSTOMERS.BC_CUSTOMERS_CUSTOMERNAME] = "EuroCars"; (([BT_CUSTOMERS.BC_CUSTOMERS_CREDITLIMIT] * 2) / 3 > 1000)) |

We'll walk through this example to help explain the core components of MQL formulas. First note the OR function. This is a boolean function which has two parameters, separated by semi-colons. These parameters are boolean expressions.

The first boolean expression first references a business column from our Metadata model. All references appear with brackets around them []. This reference first refers to the business table, and then to the business column. This boolean expression first does some arithmetic and checks to see if the final value us larger than 1000.

In the second expression, we compare the business column BT_CUSTOMERS.BC_CUSTOMERS_CUSTOMERNAME to EuroCars. Note that we use double quotes when referring to text. Double quotes are required.

Here is an example of a Physical Table Column Formula:

SUM([QUANTITYORDERED]*[PRICEEACH]) |

Aggregate functions are supported within physical table column formula definitions. The references here specifically refer to the database column, not derived physical column definitions. All operators and functions may be used in the definition of the physical table column. One special note, in order for this formula to be recognized, the "isExact" property of the physical table column must be set to true. Also note, the referenced physical column must be explicitly defined in the metadata model.

Since the latest versions (after 2008/03/14) it is possible to define formulas that use business columns from anywhere in the business model.

For example suppose we have 2 business tables:

- Orders (fact table), ID=BT_ORDER_FACT
- Product (dimension), ID=BT_PRODUCT

Suppose we want to calculate the turnover based on:

- the number of products sold, from the Orders table, ID=BC_FACT_ORDER_NRPRODUCTS
- the price of the product, from the Product table, ID=BC_DIM_PRODUCT_PRICE

To arrive there, we define a new business column, say in the Orders business table (although you could take Product too):

- Table: Orders (BT_ORDER_FACT)
- ID = BC_FACT_ORDER_TURNOVER
- Name = Turnover
- Formula = SUM( [BT_ORDER_FACT.BC_FACT_ORDER_NRPRODUCTS] * [BT_PRODUCT.BC_DIM_PRODUCT_PRICE] )
- Exact = Yes

The SQL generator is now going to replace the 2 business columns by their respective SQL variants. As such, we have to make sure that the business columns on which we base ourselves are resolving correctly. In this specific case, this means we want the 2 columns to be non-aggregated. If we now select the single business column BT_FACT_ORDER_TURNOVER, this is the SQL that is generated:

SELECT SUM( BT_ORDER_FACT.NRPRODUCTS * BT_PRODUCT.PRICE ) AS COL0 FROM FACT_ORDER BT_ORDER_FACT ,DIM_PRODUCT BT_PRODUCT WHERE ( BT_ORDER_FACT.PRODUCT_TK = BT_PRODUCT.PRODUCT_TK ) |

Now, suppose we want to generate the multiplication of the 2 sums (different use-case), we define the formula as "[BT_ORDER_FACT.BC_FACT_ORDER_NRPRODUCTS] * [BT_PRODUCT.BC_DIM_PRODUCT_PRICE]" (without the SUM) and specify an aggregation for the 2 used business columns. The generated SQL will then be:

SELECT SUM( BT_ORDER_FACT.NRPRODUCTS ) * SUM( BT_PRODUCT.PRICE ) AS COL0 FROM FACT_ORDER BT_ORDER_FACT ,DIM_PRODUCT BT_PRODUCT WHERE ( BT_ORDER_FACT.PRODUCT_TK = BT_PRODUCT.PRODUCT_TK ) |

It is obviously possible to create 2 versions of the used business columns, one aggregated (exposed to the users) and one non-aggregated (hidden from the users) for example.

The SQL generator works recursively. That means that it is possible to create a formula that calculates 7% (taxes for example) of the turnover:

- ID = BC_FACT_ORDER_TURNOVER_TAXES
- Name = Turnover Taxes
- Formula = [BT_ORDER_FACT.BC_FACT_ORDER_TURNOVER] * 7 / 100
- Exact = Yes

If we add that column to the selection, we get one extra column like this:

( SUM( BT_ORDER_FACT.NRPRODUCTS * BT_PRODUCT.PRICE ) * 7 / 100) AS COL1 |

Function syntax

FUNCTION_NAME ( PARAM ; PARAM ) |

Text (requires double quotes)

"TEXT" |

Parenthesis are used for formula precedence:

( 1 + 2) * 3 |

Business Column References:

[<BUSINESS_TABLE_ID>.<BUSINESS_COLUMN_ID>] |

Physical Column References (only used in physical column formula definitons):

[<PHYSICAL_COLUMN_NAME>] |

Function Name |
Parameters |
Description |
Example |
---|---|---|---|

OR |
2 or more boolean expression parameters |
Returns true if one or more parameters are true |
OR( |

AND |
2 or more boolean expression parameters |
Returns true if all parameters are true |
AND( |

LIKE |
2 parameters |
Compares a column to a regular expression, using "%" as wild cards |
LIKE([BT_CUSTOMERS.BC_CUSTOMERS_CUSTOMERNAME]; "%SMITH%") |

IN |
2 or more parameters |
Checks to see if the first parameter is in the following list of parameters |
IN([BT_CUSTOMERS.BC_CUSTOMERS_CUSTOMERNAME]; "Adam Smith"; "Brian Jones") |

NOW |
none |
The current date |
NOW() |

DATE |
3 numeric parameters, year, month, and day |
A specified date |
DATE(2008;4;15) |

DATEVALUE |
1 text parameter "year-month-day" |
A specified date |
DATEVALUE("2008-04-15") |

CASE |
2 or more parameters |
Evaluates the first, third, etc parameter, and returns the second, fourth, etc parameter value |
CASE( |

COALESCE |
1 or more parameters |
returns the first non null parameter |
COALESCE( |

ISNA |
1 parameter |
returns true if 1st parameter is null. Otherwise, false. |
ISNA([BT_CUSTOMERS.BC_CUSTOMERS_LAST_LOGIN]) |

- see below for aggregate functions

Operator |
Description |
---|---|

= |
returns true if two expressions are equal |

> |
returns true if first expression is larger than the second |

< |
returns true if first expression is smaller than the second |

>= |
returns true if first expression is larger than or equal to the second |

<= |
returns true if first expression is smaller than or equal to the second |

<> |
returns true if two expressions are not equal |

+ |
adds two values |

- |
subtracts two values |

* |
multiplies two values |

/ |
divides two values |

Aggregate functions may only be used in physical column definitions.

Function Name |
Description |
---|---|

SUM |
sums a specific columns values determined by grouping |

COUNT |
counts a specific columns values determined by grouping |

AVG |
averages a specific columns values determined by grouping |

MIN |
selects the minimum column value determined by grouping |

MAX |
selects the maximum column value determined by grouping |