vars

Table of Contents

Variables in CFEngine are defined as promises that an identifier of a certain type represents a particular value. Variables can be scalars or lists of types string, int, real or data.

The allowed characters in variable names are alphanumeric (both upper and lower case) and underscore. Associative arrays using the string type and square brackets [] to enclose an arbitrary key are being deprecated in favor of the data variable type.

Scalar Variables

string

Description: A scalar string

Type: string

Allowed input range: (arbitrary string)

Example:

    vars:

     "xxx"    string => "Some literal string...";
     "yyy"    string => readfile( "/home/mark/tmp/testfile" , "33" );

int

Description: A scalar integer

Type: int

Allowed input range: -99999999999,99999999999

Example:

    vars:

     "scalar" int    => "16k";
     "ran"    int    => randomint(4,88);
     "dim_array" int =>  readstringarray(
         "array_name",
         "/etc/passwd",
         "#[^\n]*",
         ":",
         10,
         4000);

Notes:

Int variables are strings that are expected to be used as integer numbers. The typing in CFEngine is dynamic, so the variable types are interchangeable. However, when you declare a variable to be type int, CFEngine verifies that the value you assign to it looks like an integer (e.g., 3, -17, 16K).

real

Description: A scalar real number

Type: real

Allowed input range: -9.99999E100,9.99999E100

Example:

    vars:

     "scalar" real   => "0.5";

Notes:

Real variables are strings that are expected to be used as real numbers. The typing in CFEngine is dynamic, so the variable types are interchangeable, but when you declare a variable to be type real, CFEngine verifies that the value you assign to it looks like a real number (e.g., 3, 3.1415, .17, 6.02e23, -9.21e-17).

Real numbers are not used in many places in CFEngine, but they are useful for representing probabilities and performance data.

List variables

Lists are specified using curly brackets {} that enclose a comma-separated list of values. The order of the list is preserved by CFEngine.

slist

Description: A list of scalar strings

Type: slist

Allowed input range: (arbitrary string)

Example:

    vars:

     "xxx"    slist  => {  "literal1",  "literal2" };
     "xxx1"   slist  => {  "1", @(xxx) }; # interpolated in order
     "yyy"    slist  => {
                        readstringlist(
                                      "/home/mark/tmp/testlist",
                                      "#[a-zA-Z0-9 ]*",
                                      "[^a-zA-Z0-9]",
                                      15,
                                      4000
                                      )
                        };

     "zzz"    slist  => { readstringlist(
        "/home/mark/tmp/testlist2",
        "#[^\n]*",
        ",",
        5,
        4000)
        };

Notes:

Some functions return slists, and an slist may contain the values copied from another slist, rlist, or ilist. See policy.

ilist

Description: A list of integers

Type: ilist

Allowed input range: -99999999999,9999999999

Example:

    vars:

      "variable_id"

           ilist => { "10", "11", "12" };

      "xxx1" ilist  => {  "1", @(variable_id) }; # interpolated in order

Notes:

Integer lists are lists of strings that are expected to be treated as integers. The typing in CFEngine is dynamic, so the variable types are interchangeable, but when you declare a variable to be type ilist, CFEngine verifies that each value you assign to it looks like an integer (e.g., 3, -17, 16K).

Some functions return ilists, and an ilist may contain the values copied from another slist, rlist, or ilist. See policy

rlist

Description: A list of real numbers

Type: rlist

Allowed input range: -9.99999E100,9.99999E100

Example:

    vars:

      "varid" rlist => { "0.1", "0.2", "0.3" };

      "xxx1" rlist  => {  "1.3", @(varid) }; # interpolated in order

Notes:

Real lists are lists of strings that are expected to be used as real numbers. The typing in CFEngine is dynamic, so the variable types are interchangeable, but when you declare a variable to be type rlist, CFEngine verifies that each value you assign to it looks like a real number (e.g., 3, 3.1415, .17, 6.02e23, -9.21e-17).

Some functions return rlists, and an rlist may contain the values copied from another slist, rlist, or ilist. See policy

Data container variables

The data variables are obtained from functions that return data containers, such as readjson(), readyaml(), parsejson(), or parseyaml(), the various data_* functions, or from merging existing data containers with mergedata(). They can NOT be modified, once created.

Inline YAML and JSON data

Data containers can be specified inline, without calling functions.

Inline YAML data has to begin with the --- preamble followed by a newline. This preamble is normally optional but here (for inline YAML) it's required by CFEngine. To generate that in CFEngine, use $(const.n) or a literal newline as shown in the example.

Inline JSON or YAML data may contain CFEngine variable references. They will be expanded at runtime as if they were simply calls to readjson() or readyaml(), which also means that syntax error in the JSON or YAML data will only be caught when your policy is actually being evaluated.

If the inline JSON or YAML data does not contain CFEngine variable references, it will be parsed at compile time, which means that cf-promises will be able to find syntax errors in your JSON or YAML data early. Thus it is highly recommended that you try to avoid variable references in your inline JSON or YAML data.

For example:

Inline Yaml example

bundle agent example_inline_yaml
{
  vars:
      # YAML requires "---" header (followed by newline)
      # NOTE \n is not interpreted as a newline, instead use $(const.n)

      "yaml_multi_line" data => '---
- "CFEngine Champions":
   - Name: "Aleksey Tsalolikhin"
     Year: 2011
   - Name: "Ted Zlatanov"
     Year : 2013';


      "yaml_single_line" data => '---$(const.n)- key1: value1$(const.n)- key2: value2';

  reports:
      "Data container defined from yaml_multi_line: $(with)"
        with => storejson( @(yaml_multi_line) );

      "Data container defined from yaml_single_line: $(with)"
        with => storejson( @(yaml_single_line) );
}
bundle agent __main__
{
  methods:
      "example_inline_yaml";
}
R: Data container defined from yaml_multi_line: [
  {
    "CFEngine Champions": [
      {
        "Name": "Aleksey Tsalolikhin",
        "Year": 2011
      },
      {
        "Name": "Ted Zlatanov",
        "Year": 2013
      }
    ]
  }
]
R: Data container defined from yaml_single_line: [
  {
    "key1": "value1"
  },
  {
    "key2": "value2"
  }
]

This policy can be found in /var/cfengine/share/doc/examples/inline-yaml.cf and downloaded directly from github.

Inline Json example

bundle agent example_inline_json
{
  vars:
      "json_multi_line" data => '{
  "CFEngine Champions": [
    {
      "Name": "Aleksey Tsalolikhin",
      "Year": "2011"
    },
    {
      "Name": "Ted Zlatanov",
      "Year": "2013"
    }
  ]
}';


      "json_single_line" data => '[{"key1":"value1"},{"key2":"value2"}]';

  reports:
      "Data container defined from json_multi_line: $(with)"
        with => storejson( @(json_multi_line) );

      "Data container defined from json_single_line: $(with)"
        with => storejson( @(json_single_line) );
}
bundle agent __main__
{
  methods:
      "example_inline_json";
}
R: Data container defined from json_multi_line: {
    "CFEngine Champions": [
      {
        "Name": "Aleksey Tsalolikhin",
        "Year": "2011"
      },
      {
        "Name": "Ted Zlatanov",
        "Year": "2013"
      }
    ]
}
R: Data container defined from json_single_line: [
  {
    "key1": "value1"
  },
  {
    "key2": "value2"
  }
]

This policy can be found in /var/cfengine/share/doc/examples/inline-json.cf and downloaded directly from github.

Passing data containers to bundles

Data containers can be passed to another bundle with the @(varname) notation, similarly to the list passing notation.

Some useful tips for using data containers

  • to extract just container[x], use mergedata("container[x]")
  • to wrap a container in an array, use mergedata("[ container ]")
  • to wrap a container in a map, use mergedata('{ "mykey": container }')
  • they act like "classic" CFEngine arrays in many ways
  • getindices() and getvalues() work on any level, e.g. getvalues("container[x][y]")
  • in reports, you have to reference a part of the container that can be expressed as a string. So for instance if you have the container c with data { "x": { "y": 50 }, "z": [ 1,2,3] } we have two top-level keys, x and z. If you report on $(c[x]) you will not get data, since there is no string there. But if you ask for $(c[x][y]) you'll get 50, and if you ask for $(c[z]) you'll get implicit iteration on 1,2,3 (just like a slist in a "classic" CFEngine array).
  • read the examples below carefully to see some useful ways to access data container contents

Iterating through a data container is only guaranteed to respect list order (e.g. [1,3,20] will be iterated in that order). Key order for maps, as per the JSON standard, is not guaranteed. Similarly, calling getindices() on a data container will give the list order of indices 0, 1, 2, ... but will not give the keys of a map in any particular order. Here's an example of iterating in list order:

body common control
{
      bundlesequence => { run };
}

bundle agent run
{
  vars:
      "x" data => parsejson('[
 { "one": "a" },
 { "two": "b" },
 { "three": "c" }
]');

      # get the numeric indices of x: 0, 1, 2
      "xi" slist => getindices(x);

      # for each xi, make a variable xpiece_$(xi) so we'll have
      # xpiece_0, xpiece_1, xpiece_2. Each xpiece will have that
      # particular element of the list x.
      "xpiece_$(xi)" string => format("%S", "x[$(xi)]");

  reports:
      "$(xi): $(xpiece_$(xi))";
}

Output:

R: 0: {"one":"a"}
R: 1: {"two":"b"}
R: 2: {"three":"c"}

Often you need to iterate through the keys of a container, and the value is a key-value property map for that key. The example here shows how you can pass the "animals" container and an "animal" key inside it to a bundle, which can then report and use the data from the key-value property map.

body common control
{
      bundlesequence => { run };
}

bundle agent run
{
  vars:
      "animals" data => parsejson('
      {
         "dog": { "legs": 4, "tail": true, "names": [ "Fido", "Cooper", "Sandy" ] },
         "cat": { "legs": 4, "tail": true, "names": [ "Fluffy", "Snowball", "Tabby" ] },
         "dolphin": { "legs": 0, "tail": true, "names": [ "Flipper", "Duffy" ] },
         "hamster": { "legs": 4, "tail": true, "names": [ "Skullcrusher", "Kimmy", "Fluffadoo" ] },
      }');

      "keys_unsorted" slist => getindices("animals");
      "keys" slist => sort(keys_unsorted, "lex");

      "animals_$(keys)" data => mergedata("animals[$(keys)]");

  methods:
      # pass the container and a key inside it
      "any" usebundle => analyze(@(animals), $(keys));
}

bundle agent analyze(animals, a)
{
  vars:
      "names" slist => getvalues("animals[$(a)][names]");
      "names_str" string => format("%S", names);

  reports:
      "$(this.bundle): possible names for animal '$(a)': $(names_str)";
      "$(this.bundle): describe animal '$(a)' => name = $(a), legs = $(animals[$(a)][legs]), tail = $(animals[$(a)][tail])";
}

Output:

R: analyze: possible names for animal 'cat': { "Fluffy", "Snowball", "Tabby" }
R: analyze: describe animal 'cat' => name = cat, legs = 4, tail = true
R: analyze: possible names for animal 'dog': { "Fido", "Cooper", "Sandy" }
R: analyze: describe animal 'dog' => name = dog, legs = 4, tail = true
R: analyze: possible names for animal 'dolphin': { "Flipper", "Duffy" }
R: analyze: describe animal 'dolphin' => name = dolphin, legs = 0, tail = true
R: analyze: possible names for animal 'hamster': { "Skullcrusher", "Kimmy", "Fluffadoo" }
R: analyze: describe animal 'hamster' => name = hamster, legs = 4, tail = true

data

Description: A data container structure

Type: data

Allowed input range: (arbitrary string)

Example:

    vars:

     "loaded1" data => readjson("/tmp/myfile.json", 40000);
     "loaded2" data => parsejson('{"key":"value"}');
     "loaded3" data => readyaml("/tmp/myfile.yaml", 40000);
     "loaded4" data => parseyaml('- key2: value2');
     "merged1" data => mergedata(loaded1, loaded2, loaded3, loaded4);

     # JSON or YAML can be inlined since CFEngine 3.7
     "inline1" data => '{"key":"value"}'; # JSON
     "inline2" data => '---$(const.n)- key2: value2'; # YAML requires "---$(const.n)" header

Attributes

policy

Description: The policy for (dis)allowing (re)definition of variables

Variables can either be allowed to change their value dynamically (be redefined) or they can be constant.

Type: (menu option)

Allowed input range:

    free
    overridable
    constant
    ifdefined

Default value:

policy = free

Example:

    vars:

      "varid" string => "value...",
              policy => "free";

Notes:

The policy free and overridable are synonyms. The policy constant is deprecated, and has no effect. All variables are free or overridable by default which means the variables values may be changed.

The policy ifdefined applies only to lists and implies that unexpanded or undefined lists are dropped. The default behavior is otherwise to retain this value as an indicator of the failure to quench the variable reference, for example:

    "one" slist => { "1", "2", "3" };

    "list" slist => { "@(one)", @(two) },
          policy => "ifdefined";

This results in @(list) being the same as @(one), and the reference to @(two) disappears. This is useful for combining lists.

For example:

example_com::
  "domain"
     string => "example.com",
    comment => "Define a global domain for hosts in the example.com domain";

# The promise above will be overridden by one of the ones below on hosts
# within the matching subdomain

one_example_com::
  "domain"
     string => "one.example.com",
    comment => "Define a global domain for hosts in the one.example.com domain";

two_example_com::
  "domain"
     string => "two.example.com",
    comment => "Define a global domain for hosts in the two.example.com domain";

(Promises within the same bundle are evaluated top to bottom, so vars promises further down in a bundle can overwrite previous values of a variable. See normal ordering for more information).

Defining variables in foreign bundles

As a general rule, variables can only be defined or re-defined from within the bundle where they were defined. There are a few notable exceptions to this general rule which are described below.

Meta type promises

Variables defined by the meta promise type are defined in a bundle scope with the same name as the executing bundle suffixed with meta.

Example policy:

    bundle agent example_meta_vars
    {
      meta:
        "tags" slist => { "autorun" };

      vars:
        "myvar" string => "my value";

      reports:
        "$(with)" with => string_mustache( "", variablesmatching_as_data( ".*example_meta_vars.*" ) );

    }
    bundle agent __main__
    {
       methods: "example_meta_vars";
    }

Example output:

    R: {
      "default:example_meta_vars.myvar": "my value",
      "default:example_meta_vars_meta.tags": [
        "autorun"
      ]
    }

Injecting variables into undefined bundles

Variables can be directly set in foreign bundles if the bundle is not defined.

Example policy:

    bundle agent example_variable_injection
    {
      vars:
        "undefined.myvar" string => "my value";
        "cant_push_this.myvar" string => "my value";

      reports:
        "$(with)" with => string_mustache( "", variablesmatching_as_data( ".*myvar.*" ) );

    }
    bundle agent cant_push_this
    {
          # If a bundle is defined, you can't simply define a variable in it from
          # another bundle, unless the variable is defined via the module protocol.
    }
    bundle agent __main__
    {
       methods: "example_variable_injection";
    }

Example output:

       error: Ignoring remotely-injected variable 'myvar'
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Ignoring remotely-injected variable 'myvar'
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
    R: {
      "default:undefined.myvar": "my value"
    }
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6
       error: Remote bundle variable injection detected!
       error: Variable identifier 'cant_push_this.myvar' is not legal
       error: Promise belongs to bundle 'example_variable_injection' in file '/tmp/example_variable_injection.cf' near line 6

Module protocol

The module protocol allows specification of context (the bundle scope within which a variable gets defined).

Example policy:

    bundle agent example_variable_injection_via_module
    {
      commands:
        "/bin/echo '^context=undefined$(const.n)=myvar=my value" module => "true";
        "/bin/echo '^context=cant_push_this$(const.n)=myvar=my value" module => "true";

      reports:
        "$(with)" with => string_mustache( "", variablesmatching_as_data( ".*myvar.*" ) );

    }
    bundle agent cant_push_this
    {
        # If a bundle is defined, you can't simply define a variable in it from
        # another bundle, unless the variable is defined via the module protocol.
    }
    bundle agent __main__
    {
       methods: "example_variable_injection_via_module";
    }

Example output:

    R: {
      "default:cant_push_this.myvar": "my value",
      "default:undefined.myvar": "my value"
    }

Augments

Augments defines variables in the def bundle scope.

This augments file that defines my_var will be used for all examples shown here (/tmp/def.json).

{
  "vars": {
    "my_var": "My value defined from augments"
    }
}

This example policy illustrates how augments defines variables in the def bundle scope (/tmp/example.cf).

bundle common def
{
  vars:
    "some_var"
      string => "My value for $(this.promiser) defined in Policy";
}
bundle agent __main__
{
  reports:
    "$(with)"
      with => string_mustache( "", variablesmatching_as_data( "default:def.*") );
}

This command shows the execution and output from the policy above.

cf-agent -Kf /tmp/example.cf
R: {
  "default:def.jq": "jq --compact-output --monochrome-output --ascii-output --unbuffered --sort-keys",
  "default:def.my_var": "My value defined from augments",
  "default:def.some_var": "My value for some_var defined in Policy"
}

Note, augments defines the variables at the start of agent initialization. The variables can be re-defined by policy during evaluation.

This example policy illustrates how policy will, by default, re-define variables set via augments (/tmp/example2.cf).

bundle common def
{
  vars:
    "some_var"
      string => "My value for $(this.promiser) defined in Policy";
    "my_var"
      string => "My value for $(this.promiser) defined in Policy";
}
bundle agent __main__
{
  reports:
    "$(with)"
      with => string_mustache( "", variablesmatching_as_data( "default:def.*") );
}

This command shows the execution and output from the policy above.

cf-agent -Kf /tmp/example2.cf
R: {
  "default:def.jq": "jq --compact-output --monochrome-output --ascii-output --unbuffered --sort-keys",
  "default:def.my_var": "My value for my_var defined in Policy",
  "default:def.some_var": "My value for some_var defined in Policy"
}

Thus augments can be used to override defaults if the policy is instrumented to support it.

This example policy illustrates how policy can be instrumented to avoid re-definition of variables set via augments (/tmp/example3.cf).

bundle common def
{
  vars:
    "some_var"
      string => "My value for $(this.promiser) defined in Policy";

    # Here we set my_var if it has not yet been defined (as in the case where
    # augments would define it before the policy was evaluated).

    "my_var"
      string => "My value for $(this.promiser) defined in Policy",
      unless => isvariable( $(this.promiser) );

}
bundle agent __main__
{
  reports:
    "$(with)"
      with => string_mustache( "", variablesmatching_as_data( "default:def.*") );
}

This command shows the execution and output from the policy above.

cf-agent -Kf /tmp/example3.cf
R: {
  "default:def.jq": "jq --compact-output --monochrome-output --ascii-output --unbuffered --sort-keys",
  "default:def.my_var": "My value defined from augments",
  "default:def.some_var": "My value for some_var defined in Policy"
}