The Mission Director (MD) is a subsystem of the game and interprets mission scripts, which are written in an XML-based language. The X Rebirth Mission Director is based on the MD in X3: Terran Conflict, with some major changes based on feedback from MD users.

An introduction to the original MD can be found in the Egosoft forums. There is also a PDF guide for the X3 Mission Director, which is partially used as a template for this document.

This document is primarily supposed to be a guide for MD users (people who use the MD to develop missions or write other MD scripts), not for MD programmers (people who work on the MD engine in C++).

Table of Contents

MD scripts

MD scripts are not necessarily missions. An MD file can contain a part of a mission, multiple missions, or no mission at all, as the MD is used for more than just missions.

MD files are XML files located in the game folder md. All XML files in that folder are loaded at game start. The file names are irrelevant, since the internally used script names are read from the XML root nodes. However, it’s recommended to keep file name and internal script name identical to avoid having to look up the names.

To edit MD scripts, an XML editing tool is needed. Microsoft Visual Studio (if available) or Microsoft Visual Web Developer (for free) are highly recommended because they have pretty good support for XML schemas (XSD). The provided Mission Director schema files help you create the XML file by displaying all available tags and attributes as you edit the XML.

This functionality is only available if the schema files md.xsd and common.xsd are in the correct folder. If you are editing the XML in the game folder directly, all is well and the files are loaded from the libraries folder. However, if you are editing in a separate folder, copy those XSD files from the libraries folder directly into the folder where your XML files are located.

MD script structure

In this section we will look at how to start the whole process by creating a new MD mission file and the basic steps in producing mission content with XML code. There will be a description of the key elements of the mission file.

The XML root node of an MD file is called “mdscript” and looks like this:

<?xml version="1.0" encoding="utf-8"?>
<mdscript name="ScriptName" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="md.xsd">

“ScriptName” is the name used for this script regardless of the file name. It has to start with an upper case letter and must be unique among all MD script names. It also should not contain spaces, so other MD scripts can use it as an identifier to access this script’s contents easily.

The only allowed sub-node of <mdscript> is <cues>, which can only contain <cue> sub-nodes:

<?xml version="1.0" encoding="utf-8"?>
<mdscript name="ScriptName" ...>
  <cues>
    <cue name="RootCue1"> [...]
    </cue>
    <cue name="RootCue2"> [...]
    </cue>
   </cues>
</mdscript>

Cues

Cues are the main ingredient of an MD script. A cue consists of a set of conditions and a set of actions. When the conditions are met, the cue is activated and the actions are performed. A cue can have child cues, or sub-cues: A sub-cue exists only when its parent cue has become active, so the activation of the parent cue initiates the condition checks of its child cues.

A cue can have the following states:

• Disabled: The parent cue has not become active yet, so this cue is basically non-existing.

• Waiting: Either this is a root cue, or the parent has become active. The cue is checking its conditions and will become active when they are met.

• Active: The cue is about to perform the actions. Child cues have entered the waiting state.
  

• Complete: The cue has finished performing its actions.

• Cancelled: The cue has been cancelled. This state cannot normally be reached but only if a cue actively cancels itself or another cue. No condition checks or actions are performed in this cue or any sub-(sub-)cue.

This is how a cue node looks like:

<cue name="CueName">
  <conditions> [...]
  </conditions>
  <delay exact="5s" />
  <actions> [...]
  </actions>
  <cues> [...]
  </cues>
</cue>

The rules for naming cues is the same for MD script names: The name starts with an upper case letter, and has to be unique within this file. So it is actually possible to use the same cue name in different scripts, which is different from the MD in X3.

Conditions

The <conditions> node can contain one or multiple conditions, all of which must be met to activate the cue. If the node is missing, the cue will become active unconditionally. The conditions are checked in sequence, and if a check fails, the following conditions are ignored. There are two types of conditions: Events and non-event conditions.

Non-event conditions are checked in regular intervals. They may be based on simple values or ranges, such as a particular in-game time having been reached or the player having a certain amount of money. They may also be based on more complex player information, such as what ships they own, whether the player is in a particular area or near a particular object.

Event conditions are triggered when the corresponding event happens, such as the event that a particular object has been targeted, attacked or destroyed. All event nodes have the prefix “event_” so you can easily determine a condition type. Non-event conditions can be used in combination with an event, so they will be checked whenever the event happens. If a condition uses an event, it must be in the first sub-node of the <conditions> node. It is even possible to define multiple alternative events that should activate the cue. The first sub-node should be <check_any> in this case, so only one of its sub-conditions has to be met.

Example for an event condition:

<conditions>
  <event_object_destroyed object="$target"/>
</conditions>

Example for an event condition with an additional (non-event) check:

<conditions>
  <event_player_killed_object/>
  <check_value value="event.param.isclass.turret"/>
</conditions>

Example for an event condition with two alternative events and a common additional check:

<conditions>
  <check_any>
    <event_cue_completed cue="Cue1"/>
    <check_all>
      <event_player_killed_object/>
      <check_value value="event.param.isclass.turret"/>
    </check_all>
  </check_any>
  <check_age min="$starttime"/>
</conditions>

For more information about expressions and event parameters, see below.

<check_all> and <check_any> can be used with non-event conditions as well, but if <check_any> is the first node of an event condition, all its sub-nodes have to define events. In case of <check_all>, only its first node must be an event (or yet another <check_any>), to make sure that exactly one event is required to activate the cue.

If a cue has a <conditions> node without any event, it must have one of the attributes onfail or checkinterval. For onfail, you can use the values “cancel” or “complete”. If the attribute is present, the condition check will happen only once and the cue will be activated on success, or completed/cancelled on failure (without performing the actions). With checkinterval, you can specify a constant time interval between condition checks. The conditions will be checked regularly forever until they are met, unless the cue’s state is changed explicitly by an external event.

Additionally, you can use the attribute checktime to set the time of the first condition check (also possible in combination with onfail). The checktime can be an expression with variables and is evaluated when the cue is enabled (when the condition checks would normally start, i.e. at game start for root cues, or after the parent cue becomes active).

Examples:

Check conditions every 5 seconds, but start checking only 1 hour after game start.

<cue name="Foo" checktime="1h" checkinterval="5s">
  <conditions>
  [...]
</cue>

Check conditions 3 seconds after the cue is enabled, and cancel the cue in case of failure.

<cue name="Foo" checktime="player.age + 3s" onfail="cancel">
  <conditions>
  [...]
</cue>

Actions

The <actions> node contains the actions that are performed one after another, without any delay inbetween. You can enforce a delay after activation of the cue and actual action performance, using a <delay> node right before the <actions>:

<delay min="10s" max="30s"/>

Note that during the delay the cue is already in the active state, and the sub-cues have been enabled! If you want to make sure that a sub-cue only becomes active after this cue is complete, there is a useful event condition for that:

<event_cue_completed cue="parent"/>

<actions> is optional. Leaving it out may be useful if you only want to enable sub-cues after the cue’s condition check. The state transition from active to complete will still take the <delay> node into account.

Note that the MD script language is not designed as a programming language. The actions are performed in sequence, although they can be nested to form more complex structures. Loops and conditionals exist to some extent, but not necessarily in the sense that a programmer might expect. Analogously to <check_all> and <check_any>, you can use <do_all> to perform all the contained sub-node actions, and <do_any> to perform only one of them. <do_all> is particularly useful when nested in a <do_any>.

Example, which selects one of the three texts randomly:

<actions>
 <do_any>
   <debug_text text="'Hello world'"/>
   <debug_text text="'Welcome to the MD'"/>
   <debug_text text="'And now for something completely different'"/>
 </do_any>
<actions>

Each child action in a <do_any> node can have a weight attribute, which can be used to control the random selection of an action node. The default weight of a child node is 1.

Also available is <do_if>, which completes the enclosed action(s) only if one provided value is non-null or matches another. Directly after a <do_if> node, you can add one or more <do_elseif> nodes to perform additional checks only in case the previous conditions were not met. The node <do_else> can be used directly after a <do_if> or a <do_elseif>. It is executed only if none of the conditions are met.

<do_while> also exists, but should be used carefully, since it is the only action that could cause an infinite loop, which freezes the game without any chance of recovery.

Every action can have a chance attribute, if you only want it to be performed with that chance, given as percentage. Otherwise it will simply be skipped. If chance is used on a conditional action such as <do_if>, the script will behave as if the condition check failed.

Libraries

Libraries are cues which are not created directly but only serve as templates for other cues. This allows for modularisation, so you can re-use library cues in many different missions.

Library cues are written like normal cues, they are also defined in a <cues> node, just with the difference that the XML tag is called library instead of cue:

<library name="LibFoo" checktime="1h" checkinterval="5s">
  <conditions>
  [...]
</library>

Although it is called library, it’s basically just a cue that doesn’t do anything. You can mix cues and libraries as you want, as root cues or sub-cues - the location within the file is unimportant. All that counts is the library name, which has to be unique within the MD script, like all other cue names.

To use a library, use the attribute ref:

<cue name="Foo" ref="LibFoo"/>

This will create a cue with the name Foo that behaves just like the library cue LibFoo. In this example, LibFoo has to be a library in the same MD script file. To use a library LibFoo from another script, you have to qualify it with the script name, using the md prefix:

<cue name="Foo" ref="md.ScriptName.LibFoo"/>

When the ref attribute is provided, all other attributes (except for name) will be ignored and taken from the library cue instead. (By default a library creates its own namespace, as if namespace="static" were specified. See the section about namespaces.)

Also all sub-cues of the library will be created as sub-cues of the cue that uses it. They are defined in the library as <cue>, not as <library>. (Although you can define a library as a sub-cue of another library, the location in the file does not matter, as already stated above.) It is even possible to reference other libraries in sub-cues of a library!

In contrast to X3TC, a cue that references a library also has its own name (Foo in the example above), so other cues can access it in expressions by that name. Sub-cues of Foo cannot be accessed by their name though. Within the library itself, expressions can use all names of cues that belong to the library (the <library> and all sub-cues). They will be translated properly when the library is referenced. Examples:

<cue name="Foo" ref="LibFoo"/>
<cue name="Bar" ref="LibFoo"/>

<library name="LibFoo">
  <actions>
    <cancel_cue cue="this"/>             <!-- Cancels the cue referencing LibFoo -->
    <cancel_cue cue="LibFoo"/>           <!-- Cancels the cue referencing LibFoo -->
    <cancel_cue cue="Foo"/>              <!-- Error, Foo not found in library -->
    <cancel_cue cue="Baz"/>              <!-- Cancels Baz in the referencing cue -->
    <cancel_cue cue="md.Script.Foo"/>    <!-- Cancels Foo -->
    <cancel_cue cue="md.Script.LibFoo"/> <!-- Error, trying to cancel library -->
    <cancel_cue cue="md.Script.Baz"/>    <!-- Error, trying to cancel library sub-cue -->
  </actions>
  <cues>
    <cue name="Baz"> [...] <!-- Sub-cue is created in all cues referencing LibFoo -->
  </cues>
</library>

So when writing the library, you don’t have to worry about name confusion, just use the names of cues in your library and it will work as expected when the library is used. Names of cues that do not belong to the library will not be available in expressions (see Foo in the example above), however, names of other libraries in the file are available when referencing them in the ref attribute.

Library Parameters

A library can be parametrised, so that it can be adapted to the needs of a missions that uses it. You can define required and/or optional parameters for a library, and it will be validated at load time that the user of the library has provided all required parameters.

Parameters are defined like this:

<library name="Lib" onfail="cancel">
  <params>
    <param name="foo"/>
    <param name="bar" default="42"/>
    <param name="baz" default="player.age"/>
  </params>
  [...]
</library>

If a default value is supplied, the parameter is regarded as optional, otherwise it’s required. When providing the actual parameters in a referencing cue, note that there is no <params> node:

<cue name="Foo" ref="Lib">
 <param name="foo" value="race.argon"/>
 <param name="bar" value="0"/>
</cue>

The values (including default values) can be variable expressions and will be evaluated when the cue is enabled, i.e. when it starts checking the conditions. They will be available to the cue as variables, using the parameter name with a ‘$’ prefix. In the example above, the variables $foo, $bar, and $baz would be created.

<library name="Lib">
  <params>
    <param name="foo"/>
  </params>
  <actions>
    <debug_text text="$foo"/>
  </actions>
</library>

If your library is supposed to provide a result to the library user, it is recommended to store a predefined variable in the library cue with a standardised name, e.g. $result. The user will be able to read it via CueName.$result. This variable does not have to be defined as a parameter but should be documented in the library.

Instantiation

One of the possible cue attributes is instantiate. If you set it to true, this changes what happens when a cue's conditions are met. Normally, if a cue is not instantiated, the cue's actions are run (taking a delay node into account) and the cue is marked as completed. But with instantiate, a copy of the cue (and all its sub-cues) is made when the conditions are met, and it is this copy in which the actions are performed and it is the copy whose status is set to complete when they are finished - this means that the original cue (the so-called static cue) remains in the waiting state, and if the conditions are met again then the whole thing happens all over again.

An instantiating cue should only be used with conditions that are only going to be met once (or a fairly limited number of times), or with conditions that include an event condition. Instantiation should not be used in a cue which, say, just depends on the game time being greater than a specific value as this will result in a copy of the cue being made after each check interval, which could increase memory usage a lot. The most common use of an instantiated cue is in responding to events such as the player ship changing sector, to react every time that event happens.

Instances that are created via instantiate are called instantiated cues. But sub-cues of instances are also instances (sub-instances) - they are created when they enter the waiting state. An instance is removed again (thereby freeing its memory) when it is complete or cancelled, and when all its instance sub-cues have been removed before. The simplest case is an instantiating cue with no sub-cues: The instance is created, the actions are performed, and the instance is removed immediately on completion. A pitfall could be an instance with a sub-cue that is forever in the waiting state (e.g. waiting for an event from an already destroyed object). It can never be removed, so you should clean up such a cue yourself, e.g. by cancelling it explicitly.

Cleaning up instances explicitly

Cancelling a cue with <cancel_cue> also cancels all its sub-cues, and cancelling a static cue stops it from instantiating more cues - but it does not cancel its instances. Resetting a cue with <reset_cue> resets both sub-cues and instantiated cues, but has the (desired) side effect that condition checks will start again if the parent cue’s state allows it. Even a sub-instance that has been reset can return to the waiting state. Resetting an instantiated cue will stop it forever, because it is not supposed to be in the waiting state (only its static cue is). Resetting will also induce the clean-up reliably, but keep in mind that this is not the case for instance sub-cues.

Access to instances

In case of instances with sub-instances, you will often want to access a related instance from the current one. Like in the non-instance case, you can simply write the cue name in an expression to reference that cue. However, you should be aware of the pitfalls that are accompanied by this.

When you use a cue name from the same script in an expression, it will always be resolved to some cue - usually a static cue, even if it is still in the disabled state, but it can also be an instance, if it is “related” to the current one.

Related means that this cue and the referenced cue have a common ancestor instance, and the referenced cue is a direct (non-instantiated) descendant of that common ancestor.

Example chart:

This chart represents a script of 5 cues: Foo, Bar, SubBar, Baz and SubBaz. Continuous arrows denote parent-child relationship. Foo and Baz are instantiating cues (highlighted with red border). The static cues always exist, although static children of instantiating cues can never become active. Instances only exist as long as they are needed.

Example situations:

• In the static tree: Cue names in expressions are always resolved to the static cues.

• In the inst-2 tree: “SubBar” in an expression will be resolved to SubBar (inst 2).

• In the inst-1 tree: “SubBar” in an expression will be resolved to SubBar (static) (!) because the SubBar child of Bar (inst 1) does not exist yet, or not any more.

• In the inst-2a tree: “SubBaz” in an expression will be resolved to SubBaz (inst 2a)

• In the inst-2a tree: “Bar” in an expression will be resolved to Bar (inst 2) because Foo (inst 2) is a common ancestor.

• In the inst-2 tree: “SubBaz” in an expression will be resolved to SubBaz (static) (!) because SubBaz (inst 2a) is not a direct descendant of the common ancestor Foo (inst 2), instead Baz (inst 2a) has been instantiated.

In expressions, you can use the cue property static to access the static cue that instantiated a cue. This does not work for sub-cues of other cues, and the result is not necessarily a real static cue! In the example above, it would only work for cues with a dotted arrow pointing at them, and is resolved to the source of the arrow. In other cases the result is null.

To get the real static cue that always exists and serves as template for instances, use the property staticbase. This works for all cues, even for the static cues themselves.

In general, to access ancestors of the current cue, you can also use the keyword parent, also recursively as properties of other cues (such as parent.parent.parent).

You can store cue references in variables. But when storing an instance cue in a variable, and later accessing that variable, be aware that the instance may not exist any more. Use the property exists to check if an instance is still alive. (In contrast, non-instance cues always exist, but may be in the disabled or cancelled state.)

Pitfalls

Some additional common pitfalls with respect to instantiation are listed here. There may be more.

• Conditions with results: If the instantiating cue has conditions with results, those results are stored in variables - but in the variables of the static cue, not of the instance! So in the <actions> you have to access the variables via the static keyword:
  <debug_text text="static.$foo"/>
  It may even be necessary to copy the variables over to the instance because the static variables can be overwritten by the next condition check:
  <set_value name="$foo" exact="static.$foo"/>

• Resetting completed/cancelled instances: As explained above, sub-instances are only created when needed (when going to the waiting state) and are destroyed when they are not needed any more (when they are completed or cancelled, including all sub-cues). There are cases in which you want to access cues that don’t exist any more - it simply doesn’t work. In some cases you are safe: You can be sure that all your ancestors exist, and instantiating cues won’t be removed until they are cancelled. In some other cases you simply don’t know and have to check if the instance is already (or still) there.

• Lifetime of instances: Do not make assumptions about when an instance is removed! Just looking at it in the Debug Manager keeps it alive for the time being. So, sometimes you could still have a completed instance that wouldn’t exist under other circumstances.