Projection framework

Current version: 9.3

The projection framework is a fluent API that transforms source data into tabular form, including the ability to group data by a key. The projection framework is available in the context of the projection worker and saves projected data to the Cortex Processing Storage database by default.

Keys, attributes, and measures

A projection includes at least one key, optional attributes, and optional measures.

  • A key  uniquely identifies a row and determines how source data should be grouped. Composite keys are supported.  Assuming the source data is a list of contacts, examples of keys include:

    • A contact ID.  This key is unique per contact and the resulting rows will not be grouped.

    • A combination of job title and year of birth. This key is unlikely to be unique per contact and the resulting rows will be grouped.

  • An attribute represent additional data about the key:

    • Given that the key is a contact ID, you might output that contact's total engagement value as an attribute.

    • Given that the key is a combination of job title and year of birth, you might call a method that returns the average starting salary as an attribute.

  • A measure  is a property of the key on which calculations can be made. Measures are only relevant if  projection reduces the source data set. For example, a source list of 40 contacts grouped by job title may results in six rows. In this scenario, you can:

    • Use .Measure("Count", x => 1)  to return the number of contacts that have each job title.

    • Use .Measure("YearsOfExperience", c => c.CareerFacet().YearsOfExperience) to return the total number of years of experience represented by contacts with each job title.

The following contact projection example groups contacts by birth year and includes a measure and an attribute:

RequestResponse
// Syntax example only
Projection = Engine.Projection.Projection.Of<XConnect.Contact>().CreateTabular(
    nameof(Model1_ContactProjection),
    Expansion,
    cfg => cfg
        .Key("BirthYear", c => c.Personal().Birthdate.Year.ToString())
        .Attribute("IsLeapYear", c => DateTime.IsLeapYear(c.Personal().Birthdate.Year))
        .Measure("NumberOfContactsInYear", c => 1)

Projection without grouping

The following sample projection outputs groups rows by contact ID and outputs each contact's job title and number of identifiers:

RequestResponse
// Syntax example only
Projection = Engine.Projection.Projection.Of<Contact>().CreateTabular(
    nameof(ProjectionOnlyModel),
    cfg => cfg
        .Key("ContactID", c => c.Id)
        .Attribute("IdentifierCount", c => c.Identifiers.Count)
        .Attribute("JobTitle", c=> c.Personal().JobTitle));

A sample set of 3 contacts would produce the following output:

ContactID

IdentifierCount

JobTitle

a3102a8c-ed77-492b-b447-4b6a0df3c412

3

Programmer Writer

c35556da-17dc-476a-98d5-56df981e4529

2

Developer

0600f5a2-4491-491e-b568-be09a8d801fe

5

Junior Developer

This example uses a unique contact ID as a key and the data set has not been reduced. There is a one-to-one relationship between the number of contacts in the source data set and the number of rows in the projected table. In scenarios like this:

  • Do not use  Measure(). Every contact ID in the source data set is unique, and the data set will not be reduced. Using Measure creates an unnecessary performance overhead.

Warning

Be wary of including contact IDs in a projection. Contact IDs uniquely identify a contact in the xDB and are not removed if a contact executes the right to be forgotten, and should not be persisted outside of the xDB. Projected data is deleted from the Cortex Processing Storage database at configurable intervals.

Projection with grouping

The following sample projection groups rows by contact birth years and job titles, and outputs a count for each row:

RequestResponse
// Syntax example only
Projection = Engine.Projection.Projection.Of<XConnect.Contact>().CreateTabular(
    nameof(Model1_ContactProjection),
    Expansion,
    cfg => cfg
        .Key("BirthYear", c => c.Personal().Birthdate.Year.ToString())
        .Key("JobTitle", c => c.Personal().JobTitle)
        .Measure("YearsExperience", c=> c.CareerInfo().YearsExperience) // Sample facet; int
        .Measure("Count", x => 1));

A sample set of 17 contacts that span 4 birth years and 3 job titles would produce the following output:

BirthYear

JobTitle

Count

YearsExperience

1987

Programmer Writer

1

10

1987

Developer

5

50

1972

Developer

10

92

1990

Junior Developer

1

3

This example uses a composite key consisting of contact birth years and job titles, and the data set has been reduced. In scenarios like this:

  • Do use Measure(). This example includes:

    • A sum of the number of contacts within each group.

    • The total number of years of work experience represented by the group.

Do you have some feedback for us?

If you have suggestions for improving this article,