Chapter 10: Miscellaneous features

Pronouns

Dialog allows the player to refer to recently mentioned objects using the words IT, HIM, HER, and THEM. For this to work, when the story author calls attention to an object as part of the prose, they must tell the library about it. This is done by querying the (notice $) predicate. It is generally only done for highly conspicuous objects mentioned at the end of room or object descriptions, like so:

Objects marked (plural $) are associated with the pronoun THEM, but it is also possible to notice an entire list of objects:

This allows e.g. the following exchange:

In addition to objects noticed by explicit queries to (notice $), the library will notice the first object mentioned by the player as part of an action, except when that object is a direction or a relation. Thus, when TAKE ENVELOPE is followed by OPEN IT, the word IT is generally understood as referring to the envelope.

Two its

Dialog treats the pronoun IT in a special way: The player's “it” (i.e. the first object mentioned in the most recent action) is tracked separately from the narrator's “it” (specified using notice). The narrator's “it” plays second fiddle to the player's “it”: Whenever a player's “it” is defined, the narrator's “it” is short-lived and only remains valid for the next non-command action. But this is sufficient to capture many ambiguous situations and prevent misunderstandings:

The dual-it feature is particularly effective, and invisible, when one of the actions is deemed more likely than the other. Consider the following playable example:

Now the following exchange is possible:

But so is this one:

List manipulation

Built-in predicates

The Dialog programming language provides two useful built-in predicates for working with lists. They are:

($Element is one of $List)

Succeeds if $Element appears in the $List. If a multi-query is made, and $Element is unbound, the predicate will backtrack over each member of the list.

(split $List by $Keywords into $Left and $Right)

If one of the values in the list $Keywords occurs in $List, $List is split into a left and a right half at that occurrence. A multi-query will find each occurrence in turn.

Library predicates

In addition to the built-in predicates, the Dialog standard library provides the following predicates for working with lists:

($List contains one of $Keys)

Succeeds if $List contains any element from the list $Keys.

($List recursively contains $Element)

Succeeds if $Element is nested anywhere inside $List.

($List contains sublist $Sublist)

Succeeds if it is possible to obtain the value $Sublist by starting with $List and removing zero or more elements from its beginning and end.

(length of $List into $N)

Counts the number of elements in $List.

(nth $List $N $Element)

Retrieves element number $N from the list. The elements are numbered starting with 1.

(last $List $Element)

Retrieves the last element of a list.

(take $N from $List into $Prefix)

Takes the first $N elements from a list.

(append $A $B $AB)

Unifies $AB with the concatenation of $A and $B.

(reverse $Input $Output)

Reverses the order of the elements in a list.

(remove duplicates $Input $Output)

Removes duplicate elements from a list.

(remove from $List matching $Keys into $Output)

Removes any elements found in the list $Keys from $List.

(split $List anywhere into $Left and $Right)

Similar to the built-in (split $ by $ into $ and $), but without keywords; splits the list anywhere. Useful for parsing e.g. GIVE ATTILA THE HUN A COOKIE.

Object tree manipulation

The Dialog standard library organizes objects into tree-like structures using a combination of the built-in ($ has parent $) dynamic predicate, and a regular predicate called ($ has relation $). The following two access predicates are useful:

($Obj is $Rel $Parent)

Succeeds when $Obj is a direct child of $Parent, and has the given relation to it. At least one of $Obj or $Parent should be bound. Can be used in (now)-statements when all parameters are bound.

($Obj is nowhere)

Succeeds when $Obj has no parent. Can be used in (now)-statements when $Obj is bound.

To check whether an object is nested under a given ancestor in the object tree, or to backtrack over every ancestor of an object, or to backtrack over every descendant of an object, use:

($Obj has ancestor $Ancestor)

At least one of the parameters must be bound.

The following predicate is similar to ($ has ancestor $), but also unifies the topmost relation (to the $Ancestor) with the middle parameter:

($Obj is nested $Rel $Ancestor)

At least one of $Obj and $Ancestor must be bound.

Every object that is in play should ultimately be nested inside a room object. To find it, query the following predicate:

($Obj is in room $Room)

$Obj is the input and must be bound; $Room is the output. The predicate fails if there is no surrounding room. If $Obj is itself a room, $Room is simply unified with it.

To check whether an object is part of another via any number of intermediary parts, or to backtrack over such relations, use the following predicate:

($Obj is part of $Ancestor)

At least one of $Obj and $Ancestor must be bound.

Items of clothing may have parts, and such parts should generally be regarded as being worn by whoever wears the parent item. To check whether an object is worn by a person according to this definition, use:

($Obj is recursively worn by $Person)

Both parameters must be bound.

When writing before-rules for actions, the following predicates may come in handy:

(ensure $Obj is held)

Attempt to take $Obj if it is an item, but not currently held. Attempt to remove $Obj if it is worn, or part of something worn.

(recursively leave non-vehicles)

While the player character is the direct child of an object that is neither a room nor a vehicle, make an attempt to leave that object.

Common checks and complaints

The following predicates are quite useful when writing prevent-rules: They check a condition, and succeed with a message if the condition was met. Their names are mostly self-explanatory.

(when $Obj is out of sight)

(when $Obj is already held)

(when $Obj isn't directly held)

(when $Obj is not here)

(when $Obj is out of reach)

(when $Obj is part of something)

(when $Obj is closed)

(when $Obj blocks passage)

(when $Obj is already $Rel $Parent)

(when $Obj is fine where it is)

(when $Obj can't be taken)

Succeeds for non-items.

(when $Obj won't accept $Rel)

The given object does not accept new children with the given relation.

(when $Obj won't accept actor $Rel)

The given object does not accept the player character as a child with the given relation.

(when $Obj is $Rel $Parent)

“You will have to get off the chair first.”

Note that it is possible to override these common complaint messages on a per-object basis, without editing the standard library code. Just remember that the when-rule needs to check the appropriate condition before printing the response:

Asking simple questions

The standard library provides the following three utility predicates for obtaining simple user input:

(yesno)

This predicate prints a prompt character, “>“, and waits for player input. If the input is Y or YES, the predicate succeeds. If the input is N or NO, the predicate fails. Otherwise, it prints the message “Please answer yes or no” and tries again.

(get number from $First to $Last $Output)

This predicate prints a prompt, “($First - $Last)”, asking the player for a number in that range. The number can be entered using decimal digits, or spelled out in text. The predicate fails on non-numeric input, and on input that's outside the specified range.

(any key)

This predicate waits for a keypress.

Identical objects

The Dialog standard library allows you to model game worlds with multiple identical objects. Authors are generally discouraged from doing this, because the existance of identical objects—and the frequent need to disambiguate between them—makes it difficult to maintain a good story-telling voice. Identical objects have a tendency to reduce even the most engaging narrative to an old-school, manipulate-the-medium-sized-dry-goods affair.

What's more, the technique described here is relatively difficult to pull off correctly. It is not recommended for novice Dialog programmers.

Still interested? Great, let's begin!

First of all, identical objects should be marked as fungible. This will affect how they are printed in object lists, and how player input is disambiguated. For performance reasons, the fungibility feature is disabled by default; it is enabled with the following rule definition:

To create many identical objects, you would normally invent a common trait for them, and then define rules in terms of the trait:

To save typing, Dialog provides a mechanism for generating such object definitions automatically. The following is mostly equivalent:

The only difference is that the internal names (hashtags) of the objects are now numbers instead of source-code names. This can potentially make debugging harder.

Now that we have created a bunch of identical objects, we must tell the library that these objects are fungible, i.e. completely interchangeable. Fungibility is expressed as a pairwise relation:

Now, in response to LOOK IN BOWL, the library might print:

A common (plural name $) must be defined for each copy of the object for this to work.

Although the objects start out identical, that will change as soon as the player starts moving them around. The library considers two objects truly fungible if they are fungible (by the aforementioned predicate) and have the same location. Thus, the four marbles in the bowl are different from the one in your pocket, and attempting to e.g. EAT MARBLE will trigger a disambiguating question: Did you want to eat a marble that's in the bowl, or the marble that's held by yourself?

To ensure that such disambiguating questions work properly, we have to instruct the library to mention the location of each fungible object when printing actions:

Otherwise it would just ask: Did you want to eat a marble, or the marble?

Of course, location is not the only property that makes fungible objects distinguishable; any dynamic predicate could have that effect. The standard library takes care of the location, but it is up to the story author to check any other properties that might differ, and to modify the full name of the object as appropriate. Thus:

Finally, recall that handled items are narrated by the standard library using a predicate called (appearance $ $ $). This mechanism is extended to also include groups of more than one fungible object. Such groups are narrated by the standard library whether the objects are marked as handled or not. The first parameter of appearance will be bound to a list. It is strongly recommended that all fungible objects are marked as handled from the start:

Here is a complete example of fungible marbles inside fungible boxes:

Try to TAKE TWO MARBLES, DROP THEM, EAT THE MARBLE, EAT A MARBLE, GET BOXES, etc. Also recompile without (fungibility enabled) to see what happens.

Note that this is an inherently confusing situation. There will be disambiguating questions along the lines of:

Here, cases 1 and 2 refer to different open boxes.

As was pointed out at the beginning of this section, fungible objects tend make interactive stories less immersive, and more fiddly. On the other hand, a judicial use of fungibility could enhance certain scenes by making the interaction feel more natural, and that may sometimes be worth the extra work. It's your call.