Free cookie consent management tool by TermsFeed Casting
Casting

Overview

All variables in Appian are strongly typed. In some cases, when using the Expression Editor, a variable's value must be cast from one data type to another. For instance, the value 123.45 has a decimal floating point number data type. This value can be cast to an integer number as the number 123. Similarly, 123 can be cast to a decimal type as 123.0.

See also: Data Types

Note:  Some casts lose information (such as 123.45 to 123), while others preserve all information (123 to 123.0). In general, the casts that are supported are intended to provide ease of use while acting in an expected manner. Casts that never make sense for a type are disallowed and will issue an error message during evaluation of the expression.

Casting is done consistently throughout the application (except for nested arrays that only flatten when saved to a process variable).

For example, if a process variable is defined as an integer, and a decimal result of an expression is stored in the process variable, then the result will be the same as if running the data through the tointeger() function.

Also, if you pass a value of type decimal to a rule, function, or type constructor that expects a value of type integer, the value that passes would be the same as if running it through the tointeger() function first.

Operator casting

Comparison operators

Before performing a comparison with a comparison operator, the left and right sides of an expression are normalized to the same type automatically (where possible) according to the following rules:

Left Right Cast to before operation
Integer Decimal Decimal
Integer Boolean Integer
Decimal Integer Decimal
Boolean Integer Integer
Text Any Text
Date and time Date Date and time
Date Date and time Date and time
Date and time Time Date and time
Time Date and time Date and time
Date Time Date and time
Time Date Date and time
User or group User User or group
User User or group User or group
User or group Group User or group
Group User or group User or group
Document Document or folder Document or folder
Document or folder Document Document or folder
Document or folder Folder Document or folder
Folder Document or folder Document or folder
Duration Decimal Duration
Decimal Duration Duration
Email address Email recipient Email recipient
Email recipient Email Address Email recipient
Text Email recipient Email recipient
Email recipient Text Email recipient
User Email recipient Email recipient
Email recipient User Email recipient
Group Email recipient Email recipient
Email recipient Group Email recipient

Arithmetic

Before performing arithmetic operations, the left and right sides are normalized to the same type where possible. Arithmetic operations on combinations not listed here are disallowed. For the sake of consistency, expressions that have the same data type on each side are also listed.

Left Right Result of operation
Boolean Boolean Integer
Boolean Integer Integer
Boolean Decimal Decimal
Boolean Text Decimal
Boolean Duration Duration
Integer Boolean Integer
Integer Integer Integer
Integer Decimal Decimal
Integer Text Decimal
Integer Duration Duration
Decimal Boolean Decimal
Decimal Integer Decimal
Decimal Decimal Decimal
Decimal Text Decimal
Decimal Duration Duration
Text Boolean Decimal
Text Integer Decimal
Text Decimal Decimal
Text Text Decimal
Text Duration Duration
Duration Boolean Duration
Duration Integer Duration
Duration Decimal Duration
Duration Text Duration

Addition, subtraction, and division operations also support the following combinations (combinations specific to just subtraction or division appear in more tables below):

Left Right Result of operation
Integer Date Date
Integer Date and time Date and time
Integer Time Time
Decimal Date Date
Decimal Date and time Date and time
Decimal Time Time
Text Date Date
Text Date and time Date and time
Text Time Time
Date Integer Date
Date Decimal Date
Date Duration Date
Date and time Integer Date and time
Date and time Decimal Date and time
Date and time Text Date and time
Date and time Duration Date and time
Time Integer Time
Time Decimal Time
Time Text Time
Time Duration Time
Duration Duration Duration
Duration Date Date
Duration Date and time Date and time

Subtraction and division operations also support the following combinations (combinations specific to just division appear in table after this):

Left Right Result of operation
Date and time Date and time Duration
Date and time Date Duration
Date Date Duration
Date Date and time Duration

Only division operations support the following combinations in addition to the other arithmetic combinations mentioned above:

Left Right Result of operation
Boolean Boolean Decimal
Boolean Integer Decimal
Integer Boolean Decimal
Integer Integer Decimal

General casting

The casting rules for each data type are listed in the following format:

To From Comment
New Type    
  Existing Type1 Comment
  Existing Type2 Comment

A null is a special value, an absence of value, or term indicating that a value is not applicable. Any null of any type may be cast to any other type. All nulls are considered the same null. Null is excluded from consideration below.

Types are often based on other types or categorized as entirely within a given type. For instance, Knowledge Center and Document data types are both Integer Keys. They are listed only as Integer Keys below. (All Appian object types other than users are treated as Integer Keys for this page.)

If a casting rule is not listed, it is not supported and will issue an error. If no comment is given, then the casting rule is that the entire value is retained. For instance, a text string of abc cast to a password is still abc.

Types are always castable to themselves without change. This is known as the "identity" cast and is not listed below.

To From Comment
Password    
  Binary  
  Text  
Text    
  Binary  
  Integer The decimal base format of the number, for example: 123 as "123"
  Decimal The decimal base format of the number, with a decimal (.) point (for example, 123.45 as "123.45")
  Duration Days::Time
  Integer Key [typename:integer value]

All Appian object types other than users are handled as Integer Keys. These include Knowledge Centers and Documents.
  Email Address  
  Password * appears in place of each password character, without necessarily matching the number of characters in the password.
  Boolean Yes or No
  Date Local date format
  Time Local time format
  Date and time Local Date and time format
  User "username"
  List value,value
  User or group As user or group Integer Key
  Document or folder As Document or Folder Integer Key
  Custom Data Type Each record can be stored as text
  Safe URI The underlying text string does not change.
Integer    
  Decimal Rounding
  Text If the minus character "-" is included in the text string, it is negative. Any data after the decimal point is dropped. The remaining digits (0-9) are used in forming the number.
  Duration Truncated
  Boolean 0 for False, 1 for True
  Date Serial date value (number of days since January 1st, 2035)
  Date and time Serial date value (number of days since January 1st, 2035)
  Integer Key The Integer ID number
  List The head of the list is cast to an Integer
Decimal    
  Integer Number as Decimal
  Text Similar to casting a text string to an integer, except it retains data that appears after the decimal point.
  Duration Number of days, time in fractional days.
  Boolean 0.0 for False, 1.0 for True
  Time Fraction of day, 12:00 pm yields 0.5.
  Integer Key The Integer ID number as a Decimal number
  List The head of the list is cast to a Decimal number
Duration    
  Integer Number as Decimal, number of days
  Decimal Number as Decimal, number of days, any fractional component is converted to a fraction of a day
  Text Similar to casting a text string to a Decimal number
  Boolean 0.0 days for False, 1.0 days for True
  Time Fraction of a day as Duration
  List The head of the list is cast to a Duration
Integer Key    
  Integer Integer ID of the correct form
  Decimal Rounded to Integer ID
  List The head of the list is cast to an Integer Key
NOTE: One form of Integer Key is not allowed to be cast to another form of Integer Key (for example, you cannot convert a Page to a Group). To do this, first cast to Integer and then back to the desired form of Integer Key (such as casting a Page to an Integer to a Group). However, this is very rarely useful.
User    
  Text The User with a given Username
  User or group The User if it represents a user; otherwise null
  List The head of the list is cast to user
Boolean    
  Integer False if 0, otherwise True
  Decimal False if 0.0, otherwise True
  Text True if the first character is 1, t, T, y, or Y; otherwise False
  List The head of the list is cast to Boolean
Date    
  Text Date in local format (discouraged, only for user input)
  Integer Serial date value (number of days since January 1st, 2035)
  Decimal Truncated serial date value (number of days since January 1st, 2035)
  Date and time Truncated serials date value (number of days since January 1st, 2035)
  List The head of the list is cast to Date
(Time is explicitly disallowed. For example, "is 2pm on a Tuesday" doesn't provide useful data)
Time    
  Text Time in local format (discouraged, only for user input)
  Date and time Time component of Date and time
  Decimal The fractional component of Date and time
  Duration The fractional component is fraction of day
  Date Midnight
  Integer From miliseconds after midnight, up to one day
  List The head of the list is cast to time
Date and time    
  Text Date and time in local format (discouraged, only for user input)
  Date Midnight on the given date
  Integer Serial date value (number of days since January 1st, 2035)
  Decimal Truncated serial date value (number of days since January 1st, 2035)
  List The head of the list is cast to Date and time
  User or Group  
  User  
  Group  
  Document or Folder  
  Document  
  Folder  
  List of type x  
  List of type y Each element is cast; if successfully cast, it is included in the final list. If not, it is ignored. This implies that a miscast list can yield an empty list.
Safe URI    
  Text The underlying text string does not change. If the URI is not considered safe, the system produces an error.

Casting custom data

One custom data type (CDT) can be cast to another CDT, when the field names match.

For example, given the following CDTs:

1
2
3
4
5
HRName 
	|
	|-lastname (text)
	|
	|-firstname (text)

— AND —

1
2
3
4
5
PayrollName 
	|
	|-lastname (text)
	|
	|-firstname (text) 

Casting HRName to PayrollName maps firstname to firstname and lastname to lastname as records with the PayrollName data type.

  • General casting rules apply to types used by fields in the CDT. (Casting may fail if the fields are of different, uncastable types.)
  • Fields in a CDT do not have to have a common origin to allow casting, only a common structure (field names).

See above: General Casting Rules

Casting record data

A record data type can be cast to a dictionary, map, CDT, or string. You can cast the record fields and use the relationships defined in the record type to cast related record fields. You cannot cast fields that have been excluded by a query, or were not mapped from the record type's data source.

Note:  A CDT may have more fields than the corresponding record type it supports.

Let's look at some examples of how to cast record data. See Casting related record data for examples of how to use your relationships to case related record fields.

Casting a record to a CDT

In the first example, we will cast the Employee record, which has four record fields: firstName, lastName, age, and phoneNumber to the Person CDT.

When casting a record to a CDT, general casting rules apply to data types used by fields in the record.

Note:  Casting may fail if casting is not available on a field type or the fields in the record type do not match the fields on the CDT.

1
2
3
4
5
6
7
8
9
cast(
  type!Person,
  recordType!Employee(
    recordType!Employee.fields.firstName: "Jane",
    recordType!Employee.fields.lastName: "Doe",
    recordType!Employee.fields.age: "31",
    recordType!Employee.fields.phoneNumber: "540-234-8975"
  )
)

The expression output will result in the record field values cast to the Person CDT:

/casting record data to cdt result

Since the age field does not exist on the Person CDT, this field was dropped.

Casting a record to a map

In this example, we'll cast the Employee record to a map.

1
2
3
4
5
6
7
8
9
cast(
  typeof(a!map()),
   recordType!Employee(
    recordType!Employee.fields.firstName: "Jane",
    recordType!Employee.fields.lastName: "Doe",
    recordType!Employee.fields.age: "31",
    recordType!Employee.fields.phoneNumber: "540-234-8975"
  )
)

The expression output will result in the record field values cast to a map:

/casting a record type to map result

Casting a record type to a dictionary works the same way as casting a record type to a map. A map retains the type of each value while a dictionary wraps a variant AnyType around each value in the dictionary.

Casting a map to a record

Now let's use the cast() function to cast in the opposite direction. We'll cast a map of four field values to the Employee record. In this expression, we'll use the recordType! domain to specify the Employee record.

1
2
3
4
5
6
7
8
9
cast(
  recordType!Employee,
  a!map(
    firstName: "Jane",
    lastName: "Doe",
    age: "31",
    phoneNumber: "540-234-8975"
  )
)

The expression output will result in the following map cast to the Employee record:

/casting a map to record type result

Casting a record to text

In the last example, we will cast the Employee record to text.

1
2
3
4
5
6
7
8
9
cast(
  3 /* Type number (Text) */,
  recordType!Employee(
    recordType!Employee.fields.firstName: "Jane",
    recordType!Employee.fields.lastName: "Doe",
    recordType!Employee.fields.age: "31",
    recordType!Employee.fields.phoneNumber: "540-234-8975"
  )
)

The expression output will result in the following record field values cast to a text string:

/casting record data to text result

You can use the relationships defined in the record type to cast related record fields to a nested CDT, a map, or a dictionary.

Note:  A CDT may have more fields than the corresponding record type or related record type it supports.

Let's take a look at a few examples.

In this example, we'll cast a record from the Case record type to the Case CDT. The Case record type has two fields: Id and title, and has a many-to-one relationship with the Customer record type. Similarly, the Case CDT has the fields Id, title, and customer, where customer uses a nested CDT as the data type.

When casting a related record to a nested CDT, the relationship name must match the nested CDT field name. General casting rules also apply to data types used by fields in the related record type.

Note:  Casting may fail if casting is not available on a field type, the relationship name does not match the nested CDT field name, or the fields on the related record type do not match the fields on the CDT.

1
2
3
4
5
6
7
8
9
10
11
cast(
  type!CSA_Case,
  recordType!Case(
    recordType!Case.fields.id: 1,
    recordType!Case.fields.title: "Case 1",
    recordType!Case.relationships.customer: recordType!Customer(
      recordType!Customer.fields.id: 1,
      recordType!Customer.fields.name: "John Doe"
    )
   )
  )

The expression output will result in the record field values cast to the Case CDT.

/casting-related-record-to-cdt

In this example, we'll cast a record from the Case record type to a map.

1
2
3
4
5
6
7
8
9
10
11
cast(
  typeof(a!map()),
  recordType!Case(
    recordType!Case.fields.id: 1,
    recordType!Case.fields.title: "Case 1",
    recordType!Case.relationships.customer: recordType!Customer(
      recordType!Customer.fields.id: 1,
      recordType!Customer.fields.name: "John Doe"
    )
   )
  )

The expression output will result in the record field values cast to a map:

/casting-related-record-to-map

Casting a related record type to a dictionary works the same way as casting a related record type to a map. A map retains the type of each value while a dictionary wraps a variant AnyType around each value in the dictionary.

Now we'll cast in the opposite direction. We'll cast a map of two record field values and two related record field values to the Case record type. In this expression, we'll use the recordType! domain to specify the Case record type.

1
2
3
4
5
6
7
8
9
10
11
cast(
  typeof(recordType!Case()),
  a!map(
    id: 1,
    title: "Case 1",
    customer: a!map(
      id: 1,
      name: "John Doe"
    )
  )
)

The expression output will result in the following map cast to the Case record type:

/casting-map-to-related-record

In the last example, we will cast values from the Case CDT and the nested Customer CDT to the Case record type.

1
2
3
4
5
6
7
8
9
10
11
cast(
  typeof(recordType!Case()),
  type!Case(
    id: 1,
    title: "Case 1",
    customer: type!Customer(
      id: 1,
      name: "John Doe"
    )
  )
)

The expression output will result in the following map cast to the Case record type:

/casting-nested-cdt-to-record

Casting

FEEDBACK