Expressions

Azure Pipelines | TFS 2018 | TFS 2017.3

Note

In Microsoft Team Foundation Server (TFS) 2018 and previous versions, build and release pipelines are called definitions, runs are called builds, service connections are called service endpoints, stages are called environments, and jobs are called phases.

Expressions can be used in many places where you need to specify a string, boolean, or number value when authoring a pipeline. The most common use of expressions is in conditions to determine whether a job or step should run.

# Expressions are used to define conditions for a step, job, or stage
steps:
- task: ...
  condition: <expression>

Another common use of expressions is in defining variables. Expressions can be evaluated at compile time or at run time. Compile-time expressions can be used anywhere; runtime expressions are more limited.

# Two examples of expressions used to define variables
# The first one, a, is evaluated when the YAML file is parsed into a plan.
# The second one, b, is evaluated at run time. 
# Note the syntax ${{}} for parse time and $[] for runtime expressions.
variables:
  a: ${{ <expression> }}
  b: $[ <expression> ]

The difference between these syntaxes is primarily what context is available. In a parse time expression, you have access to parameters and statically defined variables. In a runtime expression, you have access to more variables but no parameters.

An expression can be a literal, a reference to a variable, a reference to a dependency, a function, or a valid nested combination of these.

Literals

As part of an expression, you can use boolean, null, number, string, or version literals.

# Examples
variables:
  someBoolean: ${{ true }} # case insensitive, so True or TRUE also works
  someNumber: ${{ -1.2 }}
  someString: ${{ 'a b c' }}
  someVersion: ${{ 1.2.3 }}

Boolean

True and False are boolean literal expressions.

Null

Null is a special literal expression that's returned from a dictionary miss, e.g. (variables['noSuch']). Null can be the output of an expression but cannot be called directly within an expression.

Number

Starts with '-', '.', or '0' through '9'.

String

Must be single-quoted. For example: 'this is a string'.

To express a literal single-quote, escape it with a single quote. For example: 'It''s OK if they''re using contractions.'.

You can use a pipe character (|) for multiline strings.

myKey: |
  one
  two
  three

Version

A version number with up to four segments. Must start with a number and contain two or three period (.) characters. For example: 1.2.3.4.

Variables

As part of an expression, you may access variables using one of two syntaxes:

  • Index syntax: variables['MyVar']
  • Property dereference syntax: variables.MyVar

In order to use property dereference syntax, the property name must:

  • Start with a-Z or _
  • Be followed by a-Z 0-9 or _

Depending on the execution context, different variables are available.

  • If you create pipelines using YAML, then pipeline variables are available.
  • If you create build pipelines using classic editor, then build variables are available.
  • If you create release pipelines using classic editor, then release variables are available.

Variables are always strings. If you want to use typed values, then you should use parameters instead.

Functions

The following built-in functions can be used in expressions.

and

  • Evaluates to True if all parameters are True
  • Min parameters: 2. Max parameters: N
  • Casts parameters to Boolean for evaluation
  • Short-circuits after first False
  • Example: and(eq(variables.letters, 'ABC'), eq(variables.numbers, 123))

coalesce

  • Evaluates the parameters in order, and returns the value that does not equal null or empty-string.
  • Min parameters: 2. Max parameters: N
  • Example: coalesce(variables.couldBeNull, variables.couldAlsoBeNull, 'literal so it always works')

contains

  • Evaluates True if left parameter String contains right parameter
  • Min parameters: 2. Max parameters: 2
  • Casts parameters to String for evaluation
  • Performs ordinal ignore-case comparison
  • Example: contains('ABCDE', 'BCD') (returns True)

containsValue

  • Evaluates True if the left parameter is an array, and any item equals the right parameter. Also evaluates True if the left parameter is an object, and the value of any property equals the right parameter.
  • Min parameters: 2. Max parameters: 2
  • If the left parameter is an array, convert each item to match the type of the right parameter. If the left parameter is an object, convert the value of each property to match the type of the right parameter. The equality comparison for each specific item evaluates False if the conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Short-circuits after the first match

Note

There is no literal syntax in a YAML pipeline for specifying an array. This function is of limited use in general pipelines. It's intended for use in the pipeline decorator context with system-provided arrays such as the list of steps.

counter

  • This function can only be used in an expression that defines a variable. It cannot be used as part of a condition for a step, job, or stage.
  • Evaluates a number that is incremented with each run of a pipeline.
  • Parameters: 2. prefix and seed.
  • Prefix is a string expression. A separate value of counter is tracked for each unique value of prefix
  • Seed is the starting value of the counter

You can create a counter that is automatically incremented by one in each execution of your pipeline. When you define a counter, you provide a prefix and a seed. Here is an example that demonstrates this.

variables:
  major: 1
  # define minor as a counter with the prefix as variable major, and seed as 100.
  minor: $[counter(variables['major'], 100)]

steps:
- bash: echo $(minor)

The value of minor in the above example in the first run of the pipeline will be 100. In the second run it will be 101, provided the value of major is still 1.

If you edit the YAML file, and update the value of the variable major to be 2, then in the next run of the pipeline, the value of minor will be 100. Subsequent runs will increment the counter to 101, 102, 103, ...

Later, if you edit the YAML file, and set the value of major back to 1, then the value of the counter resumes where it left off for that prefix. In this example, it resumes at 102.

Here is another example of setting a variable to act as a counter that starts at 100, gets incremented by 1 for every run, and gets reset to 100 every day.

Note

pipeline.startTime is not available outside of expressions. pipeline.startTime formats system.pipelineStartTime into a date and time object so that it is available to work with expressions.

jobs:
- job:
  variables:
    a: $[counter(format('{0:yyyyMMdd}', pipeline.startTime), 100)]
  steps:
  - bash: echo $(a)

Here is an example of having a counter that maintains a separate value for PRs and CI runs.

variables:
  patch: $[counter(variables['build.reason'], 0)]

Counters are scoped to a pipeline. In other words, its value is incremented for each run of that pipeline. There are no project-scoped counters.

endsWith

  • Evaluates True if left parameter String ends with right parameter
  • Min parameters: 2. Max parameters: 2
  • Casts parameters to String for evaluation
  • Performs ordinal ignore-case comparison
  • Example: endsWith('ABCDE', 'DE') (returns True)

eq

  • Evaluates True if parameters are equal
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Returns False if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: eq(variables.letters, 'ABC')

format

  • Evaluates the trailing parameters and inserts them into the leading parameter string
  • Min parameters: 1. Max parameters: N
  • Example: format('Hello {0} {1}', 'John', 'Doe')
  • Uses .NET custom date and time format specifiers for date formatting (yyyy, yy, MM, M, dd, d, HH, H, m, mm, ss, s, f, ff, ffff, K)
  • Example: format('{0:yyyyMMdd}', pipeline.startTime). In this case pipeline.startTime is a special date time object variable.
  • Escape by doubling braces. For example: format('literal left brace {{ and literal right brace }}')

ge

  • Evaluates True if left parameter is greater than or equal to the right parameter
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Errors if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: ge(5, 5) (returns True)

gt

  • Evaluates True if left parameter is greater than the right parameter
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Errors if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: gt(5, 2) (returns True)

in

  • Evaluates True if left parameter is equal to any right parameter
  • Min parameters: 1. Max parameters: N
  • Converts right parameters to match type of left parameter. Equality comparison evaluates False if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Short-circuits after first match
  • Example: in('B', 'A', 'B', 'C') (returns True)

join

  • Concatenates all elements in the right parameter array, separated by the left parameter string.
  • Min parameters: 2. Max parameters: 2
  • Each element in the array is converted to a string. Complex objects are converted to empty string.
  • If the right parameter is not an array, the result is the right parameter converted to a string.

In this example, a semicolon gets added between each item in the array. The parameter type is an object.

parameters:
- name: myArray
  type: object
  default:
    - FOO
    - BAR
    - ZOO

variables:
   A: ${{ join(';',parameters.myArray) }} 

steps:
  - script: echo $A # outputs FOO;BAR;ZOO

le

  • Evaluates True if left parameter is less than or equal to the right parameter
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Errors if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: le(2, 2) (returns True)

length

  • Returns the length of a string or an array, either one that comes from the system or that comes from a parameter
  • Min parameters: 1. Max parameters 1
  • Example: length('fabrikam') returns 8

lower

  • Converts a string or variable value to all lowercase characters
  • Min parameters: 1. Max parameters 1
  • Returns the lowercase equivalent of a string
  • Example: lower('FOO') returns foo

lt

  • Evaluates True if left parameter is less than the right parameter
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Errors if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: lt(2, 5) (returns True)

ne

  • Evaluates True if parameters are not equal
  • Min parameters: 2. Max parameters: 2
  • Converts right parameter to match type of left parameter. Returns True if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Example: ne(1, 2) (returns True)

not

  • Evaluates True if parameter is False
  • Min parameters: 1. Max parameters: 1
  • Converts value to Boolean for evaluation
  • Example: not(eq(1, 2)) (returns True)

notIn

  • Evaluates True if left parameter is not equal to any right parameter
  • Min parameters: 1. Max parameters: N
  • Converts right parameters to match type of left parameter. Equality comparison evaluates False if conversion fails.
  • Ordinal ignore-case comparison for Strings
  • Short-circuits after first match
  • Example: notIn('D', 'A', 'B', 'C') (returns True)

or

  • Evaluates True if any parameter is true
  • Min parameters: 2. Max parameters: N
  • Casts parameters to Boolean for evaluation
  • Short-circuits after first True
  • Example: or(eq(1, 1), eq(2, 3)) (returns True, short-circuits)

replace

  • Returns a new string in which all instances of a string in the current instance are replaced with another string
  • Min parameters: 3. Max parameters: 3
  • replace(a, b, c): returns a, with all instances of b replaced by c
  • Example: replace('https://www.tinfoilsecurity.com/saml/consume','https://www.tinfoilsecurity.com','http://server') (returns http://server/saml/consume)

startsWith

  • Evaluates true if left parameter string starts with right parameter
  • Min parameters: 2. Max parameters: 2
  • Casts parameters to String for evaluation
  • Performs ordinal ignore-case comparison
  • Example: startsWith('ABCDE', 'AB') (returns True)

upper

  • Converts a string or variable value to all uppercase characters
  • Min parameters: 1. Max parameters 1
  • Returns the uppercase equivalent of a string
  • Example: upper('bah') returns BAH

xor

  • Evaluates True if exactly one parameter is True
  • Min parameters: 2. Max parameters: 2
  • Casts parameters to Boolean for evaluation
  • Example: xor(True, False) (returns True)

Job status check functions

You can use the following status check functions as expressions in conditions, but not in variable definitions.

always

  • Always evaluates to True (even when canceled). Note: A critical failure may still prevent a task from running. For example, if getting sources failed.

canceled

  • Evaluates to True if the pipeline was canceled.

failed

  • For a step, equivalent to eq(variables['Agent.JobStatus'], 'Failed').
  • For a job:
    • With no arguments, evaluates to True only if any previous job in the dependency graph failed.
    • With job names as arguments, evaluates to True only if any of those jobs failed.

succeeded

  • For a step, equivalent to in(variables['Agent.JobStatus'], 'Succeeded', 'SucceededWithIssues')
  • For a job:
    • With no arguments, evaluates to True only if all previous jobs in the dependency graph succeeded or partially succeeded.
    • If the previous job succeeded but a dependency further upstream failed, succeeded('previousJobName') will return true. When you just use dependsOn: previousJobName, it will fail because all of the upstream dependencies were not successful. To only evaluate the previous job, use succeeded('previousJobName') in a condition.
    • With job names as arguments, evaluates to True if all of those jobs succeeded or partially succeeded.
    • Evaluates to False if the pipeline is canceled.

succeededOrFailed

  • For a step, equivalent to in(variables['Agent.JobStatus'], 'Succeeded', 'SucceededWithIssues', 'Failed')

  • For a job:

    • With no arguments, evaluates to True regardless of whether any jobs in the dependency graph succeeded or failed.
    • With job names as arguments, evaluates to True whether any of those jobs succeeded or failed.

    This is like always(), except it will evaluate False when the pipeline is canceled.

Conditional insertion

You can use an if clause to conditionally assign the value or a variable or set inputs for tasks. Conditionals only work when using template syntax.

For templates, you can use conditional insertion when adding a sequence or mapping. Learn more about conditional insertion in templates.

Conditionally assign a variable

variables:
  ${{ if eq(variables['Build.SourceBranchName'], 'master') }}: # only works if you have a master branch
    stageName: prod

pool:
  vmImage: 'ubuntu-latest'

steps:
- script: echo ${{variables.stageName}}

Conditionally set a task input

pool:
  vmImage: 'ubuntu-latest'

steps:
- task: PublishPipelineArtifact@1
  inputs:
    targetPath: '$(Pipeline.Workspace)'
    ${{ if eq(variables['Build.SourceBranchName'], 'master') }}:
      artifact: 'prod'
    ${{ if ne(variables['Build.SourceBranchName'], 'master') }}:
      artifact: 'dev'
    publishLocation: 'pipeline'

Dependencies

Expressions can use the dependencies context to reference previous jobs or stages. You can use dependencies to:

  • Reference the job status of a previous job
  • Reference the stage status of a previous stage
  • Reference output variables in the previous job in the same stage
  • Reference output variables in the previous stage in a stage
  • Reference output variables in a job in a previous stage in the following stage

The context is called dependencies for jobs and stages and works much like variables. Inside a job, if you refer to an output variable from a job in another stage, the context is called stageDependencies.

Structurally, the dependencies object is a map of job and stage names to results and outputs. Expressed as JSON, it would look like:

"dependencies": {
  "<STAGE_NAME>" : {
    "result": "Succeeded|SucceededWithIssues|Skipped|Failed|Canceled",
    "outputs": {
        "jobName.stepName.variableName": "variable"
    }
  }
  },
    "...": {
    // another stage
  }
}

The stageDependencies object is structured the same way. Within a single stage, the current stage will not appear. In that case, you will directly reference the dependencies.

"dependencies": {
    "<JOB_NAME>": {
      "result": "Succeeded|SucceededWithIssues|Skipped|Failed|Canceled",
      "outputs": {
        "variable1": "value1",
        "variable2": "value2",
      }
    },
      "...": {
    // another job
  }
  },
    "...": {
    // another stage
  }
}

You can check job status with dependencies. In this example, Job A will always be skipped and Job B will run. Job C will run, since all of its dependencies either succeed or are skipped.

jobs:
- job: a
  condition: false
  steps:
  - script: echo Job A
- job: b
  steps:
  - script: echo Job B
- job: c
  dependsOn:
  - a
  - b
  condition: |
    and
    (
      in(dependencies.a.result, 'Succeeded', 'SucceededWithIssues', 'Skipped'),
      in(dependencies.b.result, 'Succeeded', 'SucceededWithIssues', 'Skipped')
    )
  steps:
  - script: Job C

Similarly, in this example Stage A will always be skipped and Stage B will run.

stages:
- stage: A
  condition: false
  jobs:
  - job: A1
    steps:
    - script: echo Job A1
- stage: B
  condition: in(dependencies.A.result, 'Succeeded', 'SucceededWithIssues', 'Skipped')
  jobs:
  - job: B1
    steps:
    - script: echo Job B1

You can also use dependencies to reference output variables in the previous job in the same stage. In this example, Job B depends on an output variable from Job A.

jobs:
- job: A
  steps:
  - script: echo "##vso[task.setvariable variable=skipsubsequent;isOutput=true]false"
    name: printvar

- job: B
  condition: and(succeeded(), ne(dependencies.A.outputs['printvar.skipsubsequent'], 'true'))
  dependsOn: A
  steps:
  - script: echo hello from B

By default, each stage in a pipeline depends on the one just before it in the YAML file. Stages can also use output variables from the prior stage. Here Stage B depends on a variable in Stage A.

stages:
- stage: A
  jobs:
  - job: A1
    steps:
     - script: echo "##vso[task.setvariable variable=skipsubsequent;isOutput=true]false"
       name: printvar

- stage: B
  condition: and(succeeded(), ne(stageDependencies.A.outputs['A1.printvar.skipsubsequent'], 'true'))
  dependsOn: A
  jobs:
  - job: B1
    steps:
    - script: echo hello from Stage B

You can also reference output variables that are in a job in a previous stage. In this example, there is both a job dependency and a stage dependency.

trigger: none

pool:
  vmImage: 'ubuntu-latest'

stages:
- stage: A
  jobs:
  - job: A1
    steps:
     - script: echo "##vso[task.setvariable variable=skipsubsequent;isOutput=true]false"
       name: printvar
     - script: echo "##vso[task.setvariable variable=stageexists;isOutput=true]true"
       name: stagevar

- stage: B
  condition: and(succeeded(), ne(dependencies.A.A1.outputs['printvar.skipsubsequent'], 'true'))
  dependsOn: A
  jobs:
  - job: B1
    steps:
    - script: echo hello from Stage B
  - job: B2
    condition: ne(stageDependencies.A.A1.outputs['stagevar.stageexists'], 'true')
    steps:
     - script: echo hello from Stage B2

Filtered arrays

When operating on a collection of items, you can use the * syntax to apply a filtered array. A filtered array returns all objects/elements regardless their names.

As an example, consider an array of objects named foo. We want to get an array of the values of the id property in each object in our array.

[
    { "id": 1, "a": "avalue1"},
    { "id": 2, "a": "avalue2"},
    { "id": 3, "a": "avalue3"}
]

We could do the following:

foo.*.id

This tells the system to operate on foo as a filtered array and then select the id property.

This would return:

[ 1, 2, 3 ]

Type casting

Values in an expression may be converted from one type to another as the expression gets evaluated. When an expression is evaluated, the parameters are coalesced to the relevant data type and then turned back into strings.

For example, in this YAML, the values true and false are converted to 1 and 0 when the expression is evaluated. The function lt() returns True when the left parameter is less than the right parameter.

variables:
  firstEval: $[lt(false, true)] # 0 vs. 1, True
  secondEval: $[lt(true, false)] # 1 vs. 0, False

steps:
- script: echo $(firstEval)
- script: echo $

In this example, the values variables.emptyString and the empty string both evaluate as empty strings. The function coalesce() evaluates the parameters in order, and returns the first value that does not equal null or empty-string.

variables:
  coalesceLiteral: $[coalesce(variables.emptyString, '', 'literal value')]

steps:
- script: echo $(coalesceLiteral) # outputs literal value

Detailed conversion rules are listed further below.

To
Boolean Null Number String Version
From Boolean - - Yes Yes -
Null Yes - Yes Yes -
Number Yes - - Yes Partial
String Yes Partial Partial - Partial
Version Yes - - Yes -

Boolean

To number:

  • False0
  • True1

To string:

  • False'false'
  • True'true'

Null

  • To Boolean: False
  • To number: 0
  • To string: '' (the empty string)

Number

  • To Boolean: 0False, any other number → True
  • To version: Must be greater than zero and must contain a non-zero decimal. Must be less than Int32.MaxValue (decimal component also).
  • To string: Converts the number to a string with no thousands separator and no decimal separator.

String

  • To Boolean: '' (the empty string) → False, any other string → True
  • To null: '' (the empty string) → Null, any other string not convertible
  • To number: '' (the empty string) → 0, otherwise, runs C#'s Int32.TryParse using InvariantCulture and the following rules: AllowDecimalPoint | AllowLeadingSign | AllowLeadingWhite | AllowThousands | AllowTrailingWhite. If TryParse fails, then it's not convertible.
  • To version: runs C#'s Version.TryParse. Must contain Major and Minor component at minimum. If TryParse fails, then it's not convertible.

Version

  • To Boolean: True
  • To string: Major.Minor or Major.Minor.Build or Major.Minor.Build.Revision.

FAQ

I want to do something that is not supported by expressions. What options do I have for extending Pipelines functionality?

You can customize your Pipeline with a script that includes an expression. For example, this snippet takes the BUILD_BUILDNUMBER variable and splits it with Bash. This script outputs two new variables, $MAJOR_RUN and $MINOR_RUN, for the major and minor run numbers. The two variables are then used to create two pipeline variables, $MAJOR and $MINOR with task.setvariable. These variables are available to downstream steps. To share variables across pipelines see Variable groups.

trigger:
    batch: true
    branches:
        include:
        - master
steps:
- bash: |
    MAJOR_RUN=$(echo $BUILD_BUILDNUMBER | cut -d '.' -f1)
    echo "This is the major run number: $MAJOR_RUN"
    
    MINOR_RUN=$(echo $BUILD_BUILDNUMBER | cut -d '.' -f2)
    echo "This is the minor run number: $MINOR_RUN"
    
    # create pipeline variables
    echo "##vso[task.setvariable variable=major]$MAJOR_RUN"
    echo "##vso[task.setvariable variable=minor]$MINOR_RUN"

- bash: |
    echo My pipeline variable for major run is $MAJOR
    echo My pipeline variable for minor run is $MINOR