Sample ETL transformations
This article demonstrates the following example transformations that an extract, transform, load (ETL) developer can encounter.
Flat CSV to hierarchy
Sample files
Sample CSV and JSON files are available from the public ETL Reference GitHub repo maintained by Adobe:
Example CSV
The following CRM data has been exported as CRM_profiles.csv
:
TITLE F_NAME L_NAME GENDER DOB EMAIL CRMID ECID LOYALTYID ECID2 PHONE STREET CITY STATE COUNTRY ZIP LAT LONG
Mr Ewart Bennedsen M 2004-09-25 ebennedsenex@jiathis.com 71a16013-d805-7ece-9ac4-8f2cd66e8eaa 87098882279810196101440938110216748923 2e33192000007456-0365c00000000000 55019962992006103186215643814973128178 256-284-7231 72 Buhler Crossing Anniston Alabama US 36205 33.708276 -85.7922905
Dr Novelia Ansteys F 1987-10-31 nansteysdk@spotify.com 2eeb6532-82e1-0d58-8955-bf97de66a6f5 50829196174854544323574004005273946998 2e3319208000765b-3811c00000000001 65233136134594262632703695260919939885 704-181-6371 79 Northfield Hill Charlotte North Carolina US 28299 35.2188655 -80.8108885
Mr Ulises Mochan M 1996-03-20 umochanco@gnu.org 6f393075-addb-bdd6-73f8-31c393b700f5 70086119428645095847094710218289660855 2e33192080003023-26b2000000000002 82011353387947708954389153068944017636 720-837-4159 00671 Mifflin Trail Lacolle Québec CA E5A 45.08338 -73.36585
Mrs Friederike Durrell F 1979-01-3 fdurrellbj@utexas.edu 33d018ec-5fed-f1a3-56aa-079370a9511b 50164729868919217963697788808932473456 2e33192080006dfc-0cdf400000000003 64452712468609735658703639722261004071 798-528-3458 47 Fremont Hill Independencia Veracruz Llave MX 91891 19.3803931 -99.1476905
Rev Evita Bingall F 1974-02-28 ebingallod@mac.com 8c93db88-f328-8efb-dc73-d5654d371cbe 74973364195185450328832136951985519627 2e331920800038db-0559e00000000004 58945501950285346322834356669253860483 397-178-5897 56 Crescent Oaks Court Buenavista Oaxaca MX 71730 19.4458447 -99.1497665
Mr Eugenie Bechley F 1969-05-19 ebechley9r@telegraph.co.uk b0c76a3f-6526-0ad0-e050-48143b687d18 67119779213799783658184754966135750376 2e331920800001a4-24b2800000000005 59715249079109455676103900762283358508 718-374-7456 5760 Southridge Junction Staten Island New York US 10310 40.6307451 -74.1181235
Dr Cammi Haslen F 1973-12-17 chaslenqv@ehow.com 56059cd5-5006-ce5f-2f5f-15b4d856a204 61747117963243728095047674165570746095 2e33192080007c25-2ec0600000000006 86268258269066295956223980330791223320 865-538-8291 83 Veith Street Knoxville Tennessee US 37995 35.95 -84.05
Mapping
The mapping requirements for the CRM data are outlined in the following table and include the following transformations:
- Identity columns to
identityMap
properties - Date of Birth (DOB) to Year and Month-Day
- Strings to Doubles or Short Integers.
person.birthDate: “YYYY-MM-DD”
person.birthYear: YYYY
Transform birthDate as string
Transform birthYear as short int
Output XDM
The following sample shows the first two rows of the CSV transformed to XDM, as shown in CRM_profiles.json
:
{
"person": {
"name": {
"courtesyTitle": "Mr",
"firstName": "Ewart",
"lastName": "Bennedsen"
},
"gender": "male",
"birthDayAndMonth": "09-25",
"birthDate": "2004-09-25",
"birthYear": 2004
},
"identityMap": {
"CRMID": [{
"id": "71a16013-d805-7ece-9ac4-8f2cd66e8eaa",
"primary": false
}],
"ECID": [{
"id": "87098882279810196101440938110216748923",
"primary": false
},
{
"id": "55019962992006103186215643814973128178",
"primary": false
}],
"LOYALTYID": [{
"id": "2e33192000007456-0365c00000000000",
"primary": true
}]
},
"homePhone": {
"number": "256-284-7231"
},
"personalEmail": {
"address": "ebennedsenex@jiathis.com"
},
"homeAddress": {
"street1": "72 Buhler Crossing",
"city": "Anniston",
"stateProvince": "Alabama",
"country": "US",
"postalCode": "36205",
"_schema": {
"latitude": 33.708276,
"longitude": -85.7922905
}
}
},{
"person": {
"name": {
"courtesyTitle": "Dr",
"firstName": "Novelia",
"lastName": "Ansteys"
},
"gender": "female",
"birthDayAndMonth": "10-31",
"birthDate": "1987-10-31",
"birthYear": 1987
},
"identityMap": {
"CRMID": [{
"id": "2eeb6532-82e1-0d58-8955-bf97de66a6f5",
"primary": false
}],
"ECID": [{
"id": "50829196174854544323574004005273946998",
"primary": false
},
{
"id": "65233136134594262632703695260919939885",
"primary": false
}],
"LOYALTYID": [{
"id": "2e3319208000765b-3811c00000000001",
"primary": true
}]
},
"homePhone": {
"number": "704-181-6371"
},
"personalEmail": {
"address": "nansteysdk@spotify.com"
},
"homeAddress": {
"street1": "79 Northfield Hill",
"city": "Charlotte",
"stateProvince": "North Carolina",
"country": "US",
"postalCode": "28299",
"_schema": {
"latitude": 35.2188655,
"longitude": -80.8108888
}
}
}
Dataframe to XDM schema
The hierarchy of a dataframe (such as a Parquet file) must match that of the XDM schema being uploaded to.
Example dataframe
The structure of the following example dataframe has been mapped to a schema that implements the XDM Individual Profile class, and contains the most common fields associated with schemas of that type.
[
StructField("person", StructType(
[
StructField("name", StructType(
[
StructField("courtesyTitle", StringType()),
StructField("firstName", StringType()),
StructField("lastName", StringType())
]
)),
StructField("gender", StringType()),
StructField("birthDayAndMonth", StringType()),
StructField("birthDate", StringType()),
StructField("birthYear", LongType())
]
)),
StructField("identityMap", MapType(
StructField("CRMID", ArrayType(
StructType(
[
StructField("id", StringType()),
StructField("primary", BooleanType())
]
)
)),
StructField("ECID", ArrayType(
StructType(
[
StructField("id", StringType()),
StructField("primary", BooleanType())
]
)
)),
StructField("LOYALTYID", ArrayType(
StructType(
[
StructField("id", StringType()),
StructField("primary", BooleanType())
]
)
))
)),
StructField("homePhone", StructType(
[
StructField("number", StringType())
]
)),
StructField("personalEmail", StructType(
[
StructField("address", StringType())
]
)),
StructField("homeAddress", StructType(
[
StructField("street1", StringType()),
StructField("city", StringType()),
StructField("stateProvince", StringType()),
StructField("country", StringType()),
StructField("postalCode", StringType()),
StructField("_schema", StructType(
[
StructField("latitude", DoubleType()),
StructField("latitude", DoubleType()),
]
))
]
))
]
When constructing a dataframe for use in Adobe Experience Platform, it is important to ensure that its hierarchical structure is an exact match to that of an existing XDM schema in order for the fields to map properly.
Identities to identity map
Array of identities
[
{
"xdm:id": "someone1@example.com",
"xdm:namespace": {
"xdm:code": "Email"
}
},
{
"xdm:id": "2eeb6532-82e1-0d58-8955-bf97de66a6f5",
"xdm:namespace": {
"xdm:code": "CRMID"
}
},
{
"xdm:id": "2e3319208000765b-3811c00000000001",
"xdm:namespace": {
"xdm:code": "LOYALTYID"
}
}
]
Mapping
The mapping requirements for the array of identities are outlined in the following table:
Output XDM
Below is the array of identities transformed to XDM:
"identityMap": {
"Email": [{
"id": "someone1@example.com"
}],
"CRMID": [{
"id": "2eeb6532-82e1-0d58-8955-bf97de66a6f5"
}],
"LOYALTYID": [{
"id": "2e3319208000765b-3811c00000000001"
}]
}