How Space modules work

The following steps occur when running a Space module:

  1. Bash will run the space shell script and then space will perform the init stage:
    1.1. Initialize the script context.
    1.2. Load given YAML file(s) into namespace(s).
    1.3. Fill up to three namespaces, if applicable.
    1.4. Expand given node dimensions.
    1.5. Fill up to three dimensions, if applicable.
    1.6. Setup node combo iteration, save state and hand over to the build stage.

  2. After the init stage Space enters the build stage.
    2.1. Load node(s) in the node combo.
    2.2. Assemble the script for export.
    2.3. Run script or print it to stdout.

1. Space Module init stage

1.1 Initialize the script context

Space will first unset all the Space variables to not inherit them from the shell environment.
After that, all variables defined with -e on the command line will be set to the Space context. These could include Space variables.

1.2 Load YAML files into namespaces

All YAML files and Space modules YAML files that were stated on the command line (using -f/-m switches) will be preprocessed and loaded into their respective namespace.

1.3 Fill up namespaces

If less than three namespaces were loaded, Space will look into the YAML of the first namespace after the special node /_namespaces/ and load files/modules defined there.
Normally, only one namespace is used. Further explanation can be found in Namespaces subsection.

1.4 Expand node dimensions

During this step, command line node targets are expanded for wild cards and regular expressions against their respective namespace.

1.5 Fill up dimensions

If less than three dimensions were given on the command line, Space will look into the YAML of the first namespace after the special node /_dimensions/ and, if matched, fill up dimensions 2 and 3 with node targets in their respective namespace.
Usually, only one dimension is used. Further explanation can be found in Dimensions subsection.

1.6 Setup node combo iteration

Depending on how many dimensions are specified, the node combination will result in a matrix of size X * Y * Z. The resulting size will be the number of node combinations to run.
After that, the script contexts state is saved and ready to be handed over to the build stage. Normally, there is only one dimension with single target node, so there is no iteration involved. Whenever multiple iterations occur, state is reset before proceeding to the next iteration.

2. Space Module build stage

2.1 Load node combo

The build process will be handed one node combo, which could consist of one, two or three node targets e.g. /mysite/ /myserver/ /deploy/.
Each node target is loaded from the YAML structure. If more than one node is present then they are loaded left to right and the resulting environment variables will override from left to right.

Analyzing the following node combo:
/node1/ /node2/ /node3/

Before passing over execution to the build stage, the init stage has setup a script context. This context contains the environment variables that are inherited from the shell and those which were defined on the command line using the -e switch. Also some other internal variables have been defined, such as _VERSION.

The build process will load each node in series, starting with /node1/. After that, the YAML document is read and all environment variables defined are extracted and applied onto the script context. A variable defined in the YAML document will overwrite a variable with the same name already existing in the script context (i.e. set using the -e switch or set by prior node target), if not the YAML respects the original value as default value, as:

        var1: ${var1-The value}  

Then /node2/ and /node3/ are loaded in the same way, each with the ability to override variables already defined before it.

2.2 Assemble export

Now that the script context is setup particularly for the node combo, Space will assemble the parts that will build up the final output - the exported script.
Space looks for the RUN variable. If it refers to a shell function that has not been loaded from a known module, that is, it is anonymous, then RUN will be put inside a function named _anonymous.
A module function could have a Space header. These are the first lines that set the special SPACE Header variables. The SPACE Header variables are always evaluated during the build step and affect the export.

2.3 Spacefile YAML description

The YAML structure sets up the environment and then execution is passed on to a shell script function that executes the actual task.

A YAML file is parsed and may refer to shell script functions loaded from a shell script file. The YAML structure is referred to as “nodes”, and a “node” can have shell variables associated to it.

A node can be executed if it has the environment variable RUN set. The RUN variable typically refers to a shell script function, or is a shell script snippet in itself. A node could have many environment variables attached to it (using the _env sub node). When executing a node that has many levels such as /a/b/c/, all shell script variables in each level will be loaded, where the deeper levels override their parent levels.


A namespace is where a YAML document is loaded. Each document has its own namespace to prevent conflicts.

When referring to YAML files directly (using -f) or indirectly (using -m) on the command line each document is loaded into its own namespace. There can be a maximum of three namespaces, and therefore a maximum of three -f/-m switches on the same command line.

There can also be a maximum of three node dimensions defined.

The following table presents how many dimensions are loaded in relation to how many namespaces are defined.

# Namespaces Dimension 1 Dimension 2 Dimension 3
1 namespace 1 1 1
2 namespaces 1 2 1
3 namespaces 1 2 3

Given the following example:

space -f a.yaml /node1/ /node2/ -f b.yaml /node3/

Both node1 and node3 are expected to be defined in a.yaml and they will be associated with that same YAML file. On the other hand, node2 will be associated with b.yaml.
The order of namespaces in which nodes are presented is irrelevant, thus:

space -f a.yaml -f b.yaml /node1/ /node2/ /node3/

will give the exact same result as the first example.

However, when defining nodes before namespaces as:

space /node1/ -f b.yaml /node2/

b.yaml will become the second namespace and the default namespace will be loaded by automatically injecting -f Spacefile.yaml in front of/node1/`.

space -f Spacefile.yaml /node1/ -f b.yaml /node2/
Automatic loading of namespaces 2 and 3

If less than three namespaces have been loaded using -f/-m switches, Space will look into the first namespace’s YAML structure after the special node /_namespaces/. If that node is present, then more namespaces could be loaded automatically.

In the following example, only one namespace is loaded on the command line, so Space will automatically load the module named somemodule into namespace 2, and then, since only 2 namespaces are loaded, it will load the file b.yaml into namespace 3.

        module: somemodule  
        file: b.yaml  

When defining two namespaces on the command line only the third namespace will be automatically loaded, in this case the file b.yaml (not somemodule).


For more complex combinations of modules, there is also the concept of dimensions. Space can handle up to three dimensions. This is a cuboid X*Y*Z, where X, Y and Z are one or more nodes that will be run in combination.

An example of that is X could be a list of things to operate on, Y could be one or more operations to apply to each thing and Z could be one or more “hosts” where these things exist.

The Z dimension nodes will typically provide an environment variable named SPACE_WRAP that would wrap the RUN inside another RUN, for example to run the given RUN on another host over SSH.

Each Dimension could have its own namespace or share namespaces. A namespace is a loaded YAML structure. Space could, from a given first and optional second dimension, fetch the second or third dimension.

A node dimension is a list of one or more node targets in a YAML file. Space supports up to three node dimensions together.
A node dimension can be described either from command line or from the _dimensions node in the first dimensions YAML file.

Node dimensions on the command line

Consider the following YAML file a.yaml:

# a.yaml
            - ADDRESS:
            - RUN: SERVER_UP
            - ADDRESS:
            - RUN: SERVER_UP

Running Space with the -l switch to list the nodes result in dimension 1 having a single item in its list of node targets:

space -f a.yaml /servers/alpha/ -l

However, running with a wildcard for node target result in dimension 1 having a list composed of two node targets:

space -f a.yaml "/servers/.*/" -l

Note the quotes around the node targets, that is because your shell should not start globbing on the *.

Consider having a second YAML file (dimension) named b.yaml:

# b.yaml
            - RUN: SERVER_PING
            - RUN: SERVER_STATUS

Running Space referring to nodes both in a.yaml and b.yaml results in two node dimensions, that is, two columns, each with one node target in it:

space -f a.yaml /servers/alpha/ -f b.yaml /tasks/ping/ -l
/servers/alpha/ /tasks/ping/

The wildcard can be also be used:

space -f a.yaml /servers/alpha/ -f b.yaml "/tasks/.*/" -l
/servers/alpha/ /tasks/ping/
/servers/alpha/ /tasks/status/

Wildcards also work for multiple dimensions at the same time:

space -f a.yaml "/servers/.*/" -f b.yaml "/tasks/.*/" -l
/servers/alpha/ /tasks/ping/
/servers/alpha/ /tasks/status/
/servers/beta/ /tasks/ping/
/servers/beta/ /tasks/status/

In this hypothetical example we run the targets of two YAML files against each other to first ping a server and then to query its status, for the two servers alpha and beta.

You are welcome to try it with the -d switch to see the exported script:

space -f a.yaml "/servers/.*/" -f b.yaml "/tasks/.*/" -d

Until now, examples only took advantage of two out of three possible dimensions defined by the two node targets. One could refer to a module instead of a file by using the -m switch:

space -f a.yaml "/servers/.*/" -f b.yaml "/tasks/.*/" -m ssh /wrap/ -eSSHHOST="" -l
/servers/alpha/ /tasks/ping/ /wrap/
/servers/alpha/ /tasks/status/ /wrap/
/servers/beta/ /tasks/ping/ /wrap/
/servers/beta/ /tasks/status/ /wrap/

Each column above is a node dimension and each row is called a node combo. Node combos are simply separate nodes that are loaded together and having their environment variables merged into the script context. Dimension 2 overrides variables in dimension 1. Dimension 3 overrides variables in dimension 2.

Dimensions could also refer to node targets within the same YAML file or module:

space -f a.yaml /servers/alpha/ /servers/beta/ -l
/servers/alpha/ /servers/beta/

In this case there is one node combo to run, which might not be what was originally intended. If one wants to run those two node targets separately, then it is necessary to use wildcards and regular expressions to fill out one dimension with targets.
The following example shows one dimension with two targets:

space -f a.yaml "/servers/(alpha|beta)/" -l

Regular expressions are Bash regular expressions.
Wildcards and regular expressions must always be contained inside double quotes.

Automatically adding dimensions using friend nodes

Just as namespaces can be automatically filled by defining the special node /_namespaces/, the same happens with Space dimensions.

If Space has less than three node dimensions, it will look into the first namespace’s YAML after the specific node /_dimensions/.
What the _dimensions node does is that it defines auto-fill for dimensions two and three, where applicable.
If it has one dimension then it will match an entry which has a first and second object.

In the following example, Space will automatically add the second dimension if it is not defined and when the first dimension matches any of the nodes defined.

    - first: /node1/ /node1b/ /node1c/.*/
      second: /node2/ /node2b/*/

A third auto population could also be added, like so:

    - first: /node1/ /node1b/ /node1c/.*/
      second: /node2/ /node2b/*/
      third: /node3/.*/

When matched and dimension filled, Space will not evaluate any further fill objects. Meaning that if first and second matches, third will be filled, and any next object of first/second/third will not be evaluated.


When using more than one dimension it is possible to set another order for nodes to be loaded into the environment.

_dimensions_order: 3 1 2

The above will cause environment variables from dimension 1 to override those in dimension 3, and those in 2 override both others.
It could be very useful to have the node with the RUN variable be loaded last, so it has access to all other environment variables when it is evaluated.


To auto fill dimensions we must set _dimensions_fill to “2” or “3”.
If _dimensions_fill: 2 is set then we fill for dimension two and three.
If _dimensions_fill: 3 is set then we fill only dimension three.

When filling a dimension, hidden nodes are only included when explicitly listed in the object’s pattern. That is, if any node part begins with an underscore:

    - first: /.*/_.*/.*/
      second: /.*/

Given a file a.yaml defined by:

            - A: 1-1 value
            - A: 1-1b value

            - A: 2-2 value
            - A: 2-2b value

            - A: 3-3 value
            - A: 3-3b value

    - first: /node1/
      second: /node2/.*/

    - first: /node1/
      second: /node2/sub2/
      third: /node3/sub3/

    - first: /node1/
      second: /node2/sub2b/
      third: /node3/sub3b/

    - first: /node1/.*/
      second: /node2/.*/
      third: /node3/.*/

Run a.yaml:

space -f a.yaml /node1/ -l
/node1/ /node2/sub2/ /node3/sub3/
/node1/ /node2/sub2b/ /node3/sub3b/

Run a.yaml with -l acting as a wildcard /.*/:

space -f a.yaml /node1/ -a -l
/node1/sub1/ /node2/sub2/ /node3/sub3/
/node1/sub1/ /node2/sub2/ /node3/sub3b/
/node1/sub1/ /node2/sub2b/ /node3/sub3/
/node1/sub1/ /node2/sub2b/ /node3/sub3b/
/node1/sub1b/ /node2/sub2/ /node3/sub3/
/node1/sub1b/ /node2/sub2/ /node3/sub3b/
/node1/sub1b/ /node2/sub2b/ /node3/sub3/
/node1/sub1b/ /node2/sub2b/ /node3/sub3b/

Above is the same as running:

space -f a.yaml "/node1/.*" -l

Multiple objects under /_dimensions/ is always a possibility.
Whenever defining _dimensions in namespace 1, extra dimensions will be implicitly added according to _dimensions_fill and if matches are made.

Previous: Running a Space Module

Next: Writing your own Space Module

Edit this page