LambdaMOO is a network-accessible, multi-user, programmable, interactive system well-suited to the construction of text-based adventure games, conferencing systems, and other collaborative software. Its most common use, however, is as a multi-participant, low-bandwidth virtual reality, and it is with this focus in mind that I describe it here.
Participants (usually referred to as players) connect to LambdaMOO using Telnet or some other, more specialized, client program. Upon connection, they are usually presented with a welcome message explaining how to either create a new character or connect to an existing one. Characters are the embodiment of players in the virtual reality that is LambdaMOO.
Having connected to a character, players then give one-line commands that are parsed and interpreted by LambdaMOO as appropriate. Such commands may cause changes in the virtual reality, such as the location of a character, or may simply report on the current state of that reality, such as the appearance of some object.
The job of interpreting those commands is shared between the two major components in the LambdaMOO system: the server and the database. The server is a program, written in a standard programming language, that manages the network connections, maintains queues of commands and other tasks to be executed, controls all access to the database, and executes other programs written in the MOO programming language. The database contains representations of all the objects in the virtual reality, including the MOO programs that the server executes to give those objects their specific behaviors.
Almost every command is parsed by the server into a call on a MOO procedure, or verb, that actually does the work. Thus, programming in the MOO language is a central part of making non-trivial extensions to the database and thus, the virtual reality.
In the next chapter, I describe the structure and contents of a LambdaMOO database. The following chapter gives a complete description of how the server performs its primary duty: parsing the commands typed by players. Next, I describe the complete syntax and semantics of the MOO programming language. Finally, I describe all of the database conventions assumed by the server.
Note: This manual describes only those aspects of LambdaMOO that are entirely independent of the contents of the database. It does not describe, for example, the commands or programming interfaces present in the LambdaCore database.
In this chapter, I begin by describing in detail the various kinds of data that can appear in a LambdaMOO database and that, therefore, MOO programs can manipulate. In a few places, I refer to the LambdaCore database. This is one particular LambdaMOO database, created every so often by extracting the "core" of the current database for the original LambdaMOO.
Note: The original LambdaMOO resides on the host
lambda.parc.xerox.com(the numeric address for which is192.216.54.2), on port 8888. Feel free to drop by! A copy of the most recent release of the LambdaCore database can be obtained by anonymous FTP from hostftp.parc.xerox.comin the directorypub/MOO.
There are only a few kinds of values that MOO programs can manipulate:
MOO supports the integers from -2^31 (that is, negative two to the power of 31) up to 2^31 - 1 (one less than two to the power of 31); that's from -2147483648 to 2147483647, enough for most purposes. In MOO programs, integers are written just as you see them here, an optional minus sign followed by a non-empty sequence of decimal digits. In particular, you may not put commas, periods, or spaces in the middle of large integers, as we sometimes do in English and other natural languages (e.g., `2,147,483,647').
Real numbers in MOO are represented as they are in almost all other programming languages, using so-called floating-point numbers. These have certain (large) limits on size and precision that make them useful for a wide range of applications. Floating-point numbers are written with an optional minus sign followed by a non-empty sequence of digits punctuated at some point with a decimal point (`.') and/or followed by a scientific-notation marker (the letter `E' or `e' followed by an optional sign and one or more digits). Here are some examples of floating-point numbers:
325.0 325. 3.25e2 0.325E3 325.E1 .0325e+4 32500e-2
All of these examples mean the same number. The third of these, as an example of scientific notation, should be read "3.25 times 10 to the power of 2".
Fine points: The MOO represents floating-point numbers using the local meaning of the C-language
doubletype, which is almost always equivalent to IEEE 754 double precision floating point. If so, then the smallest positive floating-point number is no larger than2.2250738585072014e-308and the largest floating-point number is1.7976931348623157e+308.IEEE infinities and NaN values are not allowed in MOO. The error
E_FLOATis raised whenever an infinity would otherwise be computed;E_INVARGis raised whenever a NaN would otherwise arise. The value0.0is always returned on underflow.
Character strings are arbitrarily-long sequences of normal, ASCII printing characters. When written as values in a program, strings are enclosed in double-quotes, like this:
"This is a character string."
To include a double-quote in the string, precede it with a backslash (`\'), like this:
"His name was \"Leroy\", but nobody ever called him that."
Finally, to include a backslash in a string, double it:
"Some people use backslash ('\\') to mean set difference."
MOO strings may not include special ASCII characters like carriage-return, line-feed, bell, etc. The only non-printing characters allowed are spaces and tabs.
Fine point: There is a special kind of string used for representing the arbitrary bytes used in general, binary input and output. In a binary string, any byte that isn't an ASCII printing character or the space character is represented as the three-character substring "~XX", where XX is the hexadecimal representation of the byte; the input character `~' is represented by the three-character substring "~7E". This special representation is used by the functions
encode_binary()anddecode_binary()and by the functionsnotify()andread()with network connections that are in binary mode. See the descriptions of theset_connection_option(),encode_binary(), anddecode_binary()functions for more details.
Objects are the backbone of the MOO database and, as such, deserve a great deal of discussion; the entire next section is devoted to them. For now, let it suffice to say that every object has a number, unique to that object. In programs, we write a reference to a particular object by putting a hash mark (`#') followed by the number, like this:
#495
Object numbers are always integers.
There are three special object numbers used for a variety of purposes:
#-1, #-2, and #-3, usually referred to in the
LambdaCore database as $nothing, $ambiguous_match, and
$failed_match, respectively.
Errors are, by far, the least frequently used values in MOO. In the normal case, when a program attempts an operation that is erroneous for some reason (for example, trying to add a number to a character string), the server stops running the program and prints out an error message. However, it is possible for a program to stipulate that such errors should not stop execution; instead, the server should just let the value of the operation be an error value. The program can then test for such a result and take some appropriate kind of recovery action. In programs, error values are written as words beginning with `E_'. The complete list of error values, along with their associated messages, is as follows:
E_NONE No error E_TYPE Type mismatch E_DIV Division by zero E_PERM Permission denied E_PROPNF Property not found E_VERBNF Verb not found E_VARNF Variable not found E_INVIND Invalid indirection E_RECMOVE Recursive move E_MAXREC Too many verb calls E_RANGE Range error E_ARGS Incorrect number of arguments E_NACC Move refused by destination E_INVARG Invalid argument E_QUOTA Resource limit exceeded E_FLOAT Floating-point arithmetic error
The final kind of value in MOO programs is lists. A list is a sequence of arbitrary MOO values, possibly including other lists. In programs, lists are written in mathematical set notation with each of the elements written out in order, separated by commas, the whole enclosed in curly braces (`{' and `}'). For example, a list of the names of the days of the week is written like this:
{"Sunday", "Monday", "Tuesday", "Wednesday",
"Thursday", "Friday", "Saturday"}
Note that it doesn't matter that we put a line-break in the middle of the list. This is true in general in MOO: anywhere that a space can go, a line-break can go, with the same meaning. The only exception is inside character strings, where line-breaks are not allowed.
Objects are, in a sense, the whole point of the MOO programming language. They are used to represent objects in the virtual reality, like people, rooms, exits, and other concrete things. Because of this, MOO makes a bigger deal out of creating objects than it does for other kinds of value, like integers.
Numbers always exist, in a sense; you have only to write them down in order to operate on them. With objects, it is different. The object with number `#958' does not exist just because you write down its number. An explicit operation, the `create()' function described later, is required to bring an object into existence. Symmetrically, once created, objects continue to exist until they are explicitly destroyed by the `recycle()' function (also described later).
The identifying number associated with an object is unique to that object. It was assigned when the object was created and will never be reused, even if the object is destroyed. Thus, if we create an object and it is assigned the number `#1076', the next object to be created will be assigned `#1077', even if `#1076' is destroyed in the meantime.
Every object is made up of three kinds of pieces that together define its behavior: attributes, properties, and verbs.
There are three fundamental attributes to every object:
The act of creating a character sets the player attribute of an object and
only a wizard (using the function set_player_flag()) can change that
setting. Only characters have the player bit set to 1.
The parent/child hierarchy is used for classifying objects into general classes
and then sharing behavior among all members of that class. For example, the
LambdaCore database contains an object representing a sort of "generic" room.
All other rooms are descendants (i.e., children or children's children,
or ...) of that one. The generic room defines those pieces of behavior
that are common to all rooms; other rooms specialize that behavior for their
own purposes. The notion of classes and specialization is the very essence of
what is meant by object-oriented programming. Only the functions
create(), recycle(), chparent(), and renumber() can
change the parent and children attributes.
A property is a named "slot" in an object that can hold an arbitrary MOO value. Every object has eight built-in properties whose values are constrained to be of particular types. In addition, an object can have any number of other properties, none of which have type constraints. The built-in properties are as follows:
name a string, the usual name for this object owner an object, the player who controls access to it location an object, where the object is in virtual reality contents a list of objects, the inverse of `location' programmer a bit, does the object have programmer rights? wizard a bit, does the object have wizard rights? r a bit, is the object publicly readable? w a bit, is the object publicly writable? f a bit, is the object fertile?
The `name' property is used to identify the object in various printed messages. It can only be set by a wizard or by the owner of the object. For player objects, the `name' property can only be set by a wizard; this allows the wizards, for example, to check that no two players have the same name.
The `owner' identifies the object that has owner rights to this object, allowing them, for example, to change the `name' property. Only a wizard can change the value of this property.
The `location' and `contents' properties describe a hierarchy of
object containment in the virtual reality. Most objects are located
"inside" some other object and that other object is the value of the
`location' property. The `contents' property is a list of those
objects for which this object is their location. In order to maintain the
consistency of these properties, only the move() function is able to
change them.
The `wizard' and `programmer' bits are only applicable to characters, objects representing players. They control permission to use certain facilities in the server. They may only be set by a wizard.
The `r' bit controls whether or not players other than the owner of this object can obtain a list of the properties or verbs in the object. Symmetrically, the `w' bit controls whether or not non-owners can add or delete properties and/or verbs on this object. The `r' and `w' bits can only be set by a wizard or by the owner of the object.
The `f' bit specifies whether or not this object is fertile, whether
or not players other than the owner of this object can create new objects with
this one as the parent. It also controls whether or not non-owners can use the
chparent() built-in function to make this object the parent of an
existing object. The `f' bit can only be set by a wizard or by the owner
of the object.
All of the built-in properties on any object can, by default, be read by any player. It is possible, however, to override this behavior from within the database, making any of these properties readable only by wizards. See the chapter on server assumptions about the database for details.
As mentioned above, it is possible, and very useful, for objects to have other properties aside from the built-in ones. These can come from two sources.
First, an object has a property corresponding to every property in its parent object. To use the jargon of object-oriented programming, this is a kind of inheritance. If some object has a property named `foo', then so will all of its children and thus its children's children, and so on.
Second, an object may have a new property defined only on itself and its descendants. For example, an object representing a rock might have properties indicating its weight, chemical composition, and/or pointiness, depending upon the uses to which the rock was to be put in the virtual reality.
Every defined property (as opposed to those that are built-in) has an owner and a set of permissions for non-owners. The owner of the property can get and set the property's value and can change the non-owner permissions. Only a wizard can change the owner of a property.
The initial owner of a property is the player who added it; this is usually, but not always, the player who owns the object to which the property was added. This is because properties can only be added by the object owner or a wizard, unless the object is publicly writable (i.e., its `w' property is 1), which is rare. Thus, the owner of an object may not necessarily be the owner of every (or even any) property on that object.
The permissions on properties are drawn from this set: `r' (read), `w' (write), and `c' (change ownership in descendants). Read permission lets non-owners get the value of the property and, of course, write permission lets them set that value. The `c' permission bit is a little more complicated.
Recall that every object has all of the properties that its parent does and perhaps some more. Ordinarily, when a child object inherits a property from its parent, the owner of the child becomes the owner of that property. This is because the `c' permission bit is "on" by default. If the `c' bit is not on, then the inherited property has the same owner in the child as it does in the parent.
As an example of where this can be useful, the LambdaCore database ensures
that every player has a `password' property containing the encrypted
version of the player's connection password. For security reasons, we don't
want other players to be able to see even the encrypted version of the
password, so we turn off the `r' permission bit. To ensure that the
password is only set in a consistent way (i.e., to the encrypted version of a
player's password), we don't want to let anyone but a wizard change the
property. Thus, in the parent object for all players, we made a wizard the
owner of the password property and set the permissions to the empty string,
"". That is, non-owners cannot read or write the property and, because
the `c' bit is not set, the wizard who owns the property on the parent
class also owns it on all of the descendants of that class.
Another, perhaps more down-to-earth example arose when a character named Ford started building objects he called "radios" and another character, yduJ, wanted to own one. Ford kindly made the generic radio object fertile, allowing yduJ to create a child object of it, her own radio. Radios had a property called `channel' that identified something corresponding to the frequency to which the radio was tuned. Ford had written nice programs on radios (verbs, discussed below) for turning the channel selector on the front of the radio, which would make a corresponding change in the value of the `channel' property. However, whenever anyone tried to turn the channel selector on yduJ's radio, they got a permissions error. The problem concerned the ownership of the `channel' property.
As I explain later, programs run with the permissions of their author. So, in this case, Ford's nice verb for setting the channel ran with his permissions. But, since the `channel' property in the generic radio had the `c' permission bit set, the `channel' property on yduJ's radio was owned by her. Ford didn't have permission to change it! The fix was simple. Ford changed the permissions on the `channel' property of the generic radio to be just `r', without the `c' bit, and yduJ made a new radio. This time, when yduJ's radio inherited the `channel' property, yduJ did not inherit ownership of it; Ford remained the owner. Now the radio worked properly, because Ford's verb had permission to change the channel.
The final kind of piece making up an object is verbs. A verb is a named MOO program that is associated with a particular object. Most verbs implement commands that a player might type; for example, in the LambdaCore database, there is a verb on all objects representing containers that implements commands of the form `put object in container'. It is also possible for MOO programs to invoke the verbs defined on objects. Some verbs, in fact, are designed to be used only from inside MOO code; they do not correspond to any particular player command at all. Thus, verbs in MOO are like the `procedures' or `methods' found in some other programming languages.
As with properties, every verb has an owner and a set of permission bits. The owner of a verb can change its program, its permission bits, and its argument specifiers (discussed below). Only a wizard can change the owner of a verb. The owner of a verb also determines the permissions with which that verb runs; that is, the program in a verb can do whatever operations the owner of that verb is allowed to do and no others. Thus, for example, a verb owned by a wizard must be written very carefully, since wizards are allowed to do just about anything.
The permission bits on verbs are drawn from this set: `r' (read), `w' (write), `x' (execute), and `d' (debug). Read permission lets non-owners see the program for a verb and, symmetrically, write permission lets them change that program. The other two bits are not, properly speaking, permission bits at all; they have a universal effect, covering both the owner and non-owners.
The execute bit determines whether or not the verb can be invoked from within a MOO program (as opposed to from the command line, like the `put' verb on containers). If the `x' bit is not set, the verb cannot be called from inside a program. The `x' bit is usually set.
The setting of the debug bit determines what happens when the verb's program does something erroneous, like subtracting a number from a character string. If the `d' bit is set, then the server raises an error value; such raised errors can be caught by certain other pieces of MOO code. If the error is not caught, however, the server aborts execution of the command and, by default, prints an error message on the terminal of the player whose command is being executed. (See the chapter on server assumptions about the database for details on how uncaught errors are handled.) If the `d' bit is not set, then no error is raised, no message is printed, and the command is not aborted; instead the error value is returned as the result of the erroneous operation.
Note: the `d' bit exists only for historical reasons; it used to be the only way for MOO code to catch and handle errors. With the introduction of the
try-exceptstatement and the error-catching expression, the `d' bit is no longer useful. All new verbs should have the `d' bit set, using the newer facilities for error handling if desired. Over time, old verbs written assuming the `d' bit would not be set should be changed to use the new facilities instead.
In addition to an owner and some permission bits, every verb has three `argument specifiers', one each for the direct object, the preposition, and the indirect object. The direct and indirect specifiers are each drawn from this set: `this', `any', or `none'. The preposition specifier is `none', `any', or one of the items in this list:
with/using at/to in front of in/inside/into on top of/on/onto/upon out of/from inside/from over through under/underneath/beneath behind beside for/about is as off/off of
The argument specifiers are used in the process of parsing commands, described in the next chapter.
The MOO server is able to do a small amount of parsing on the commands that a player enters. In particular, it can break apart commands that follow one of the following forms:
verb verb direct-object verb direct-object preposition indirect-object
Real examples of these forms, meaningful in the LambdaCore database, are as follows:
look take yellow bird put yellow bird in cuckoo clock
Note that English articles (i.e., `the', `a', and `an') are not generally used in MOO commands; the parser does not know that they are not important parts of objects' names.
To have any of this make real sense, it is important to understand precisely how the server decides what to do when a player types a command.
First, the server checks whether or not the first non-blank character in the command is one of the following:
" : ;
If so, that character is replaced by the corresponding command below, followed by a space:
say emote eval
For example, the command
"Hi, there.
is treated exactly as if it were as follows:
say Hi, there.
The server next breaks up the command into words. In the simplest case, the command is broken into words at every run of space characters; for example, the command `foo bar baz' would be broken into the words `foo', `bar', and `baz'. To force the server to include spaces in a "word", all or part of a word can be enclosed in double-quotes. For example, the command
foo "bar mumble" baz" "fr"otz" bl"o"rt
is broken into the words `foo', `bar mumble', `baz frotz', and `blort'. Finally, to include a double-quote or a backslash in a word, they can be preceded by a backslash, just like in MOO strings.
Having thus broken the string into words, the server next checks to see if the first word names any of the six "built-in" commands: `.program', `PREFIX', `OUTPUTPREFIX', `SUFFIX', `OUTPUTSUFFIX', or the connection's defined flush command, if any (`.flush' by default). The first one of these is only available to programmers, the next four are intended for use by client programs, and the last can vary from database to database or even connection to connection; all six are described in the final chapter of this document, "Server Commands and Database Assumptions". If the first word isn't one of the above, then we get to the usual case: a normal MOO command.
The server next gives code in the database a chance to handle the command. If
the verb $do_command() exists, it is called with the words of the
command passed as its arguments and argstr set to the raw command typed
by the user. If $do_command() does not exist, or if that verb-call
completes normally (i.e., without suspending or aborting) and returns a false
value, then the built-in command parser is invoked to handle the command as
described below. Otherwise, it is assumed that the database code handled the
command completely and no further action is taken by the server for that
command.
If the built-in command parser is invoked, the server tries to parse the command into a verb, direct object, preposition and indirect object. The first word is taken to be the verb. The server then tries to find one of the prepositional phrases listed at the end of the previous section, using the match that occurs earliest in the command. For example, in the very odd command `foo as bar to baz', the server would take `as' as the preposition, not `to'.
If the server succeeds in finding a preposition, it considers the words
between the verb and the preposition to be the direct object and those
after the preposition to be the indirect object. In both cases, the
sequence of words is turned into a string by putting one space between
each pair of words. Thus, in the odd command from the previous
paragraph, there are no words in the direct object (i.e., it is
considered to be the empty string, "") and the indirect object is
"bar to baz".
If there was no preposition, then the direct object is taken to be all of the words after the verb and the indirect object is the empty string.
The next step is to try to find MOO objects that are named by the direct and indirect object strings.
First, if an object string is empty, then the corresponding object is the
special object #-1 (aka $nothing in LambdaCore). If an object
string has the form of an object number (i.e., a hash mark (`#') followed
by digits), and the object with that number exists, then that is the named
object. If the object string is either "me" or "here", then the
player object itself or its location is used, respectively.
Otherwise, the server considers all of the objects whose location is either the player (i.e., the objects the player is "holding", so to speak) or the room the player is in (i.e., the objects in the same room as the player); it will try to match the object string against the various names for these objects.
The matching done by the server uses the `aliases' property of each of the objects it considers. The value of this property should be a list of strings, the various alternatives for naming the object. If it is not a list, or the object does not have an `aliases' property, then the empty list is used. In any case, the value of the `name' property is added to the list for the purposes of matching.
The server checks to see if the object string in the command is either exactly
equal to or a prefix of any alias; if there are any exact matches, the prefix
matches are ignored. If exactly one of the objects being considered has a
matching alias, that object is used. If more than one has a match, then the
special object #-2 (aka $ambiguous_match in LambdaCore) is used.
If there are no matches, then the special object #-3 (aka
$failed_match in LambdaCore) is used.
So, now the server has identified a verb string, a preposition string, and direct- and indirect-object strings and objects. It then looks at each of the verbs defined on each of the following four objects, in order:
For each of these verbs in turn, it tests if all of the the following are true:
I'll explain each of these criteria in turn.
Every verb has one or more names; all of the names are kept in a single string, separated by spaces. In the simplest case, a verb-name is just a word made up of any characters other than spaces and stars (i.e., ` ' and `*'). In this case, the verb-name matches only itself; that is, the name must be matched exactly.
If the name contains a single star, however, then the name matches any prefix of itself that is at least as long as the part before the star. For example, the verb-name `foo*bar' matches any of the strings `foo', `foob', `fooba', or `foobar'; note that the star itself is not considered part of the name.
If the verb name ends in a star, then it matches any string that begins with the part before the star. For example, the verb-name `foo*' matches any of the strings `foo', `foobar', `food', or `foogleman', among many others. As a special case, if the verb-name is `*' (i.e., a single star all by itself), then it matches anything at all.
Recall that the argument specifiers for the direct and indirect objects are
drawn from the set `none', `any', and `this'. If the specifier
is `none', then the corresponding object value must be #-1 (aka
$nothing in LambdaCore); that is, it must not have been specified. If
the specifier is `any', then the corresponding object value may be
anything at all. Finally, if the specifier is `this', then the
corresponding object value must be the same as the object on which we found
this verb; for example, if we are considering verbs on the player, then the
object value must be the player object.
Finally, recall that the argument specifier for the preposition is either `none', `any', or one of several sets of prepositional phrases, given above. A specifier of `none' matches only if there was no preposition found in the command. A specifier of `any' always matches, regardless of what preposition was found, if any. If the specifier is a set of prepositional phrases, then the one found must be in that set for the specifier to match.
So, the server considers several objects in turn, checking each of their verbs in turn, looking for the first one that meets all of the criteria just explained. If it finds one, then that is the verb whose program will be executed for this command. If not, then it looks for a verb named `huh' on the room that the player is in; if one is found, then that verb will be called. This feature is useful for implementing room-specific command parsing or error recovery. If the server can't even find a `huh' verb to run, it prints an error message like `I couldn't understand that.' and the command is considered complete.
At long last, we have a program to run in response to the command typed by the player. When the code for the program begins execution, the following built-in variables will have the indicated values:
player an object, the player who typed the command this an object, the object on which this verb was found caller an object, the same as `player' verb a string, the first word of the command argstr a string, everything after the first word of the command args a list of strings, the words in `argstr' dobjstr a string, the direct object string found during parsing dobj an object, the direct object value found during matching prepstr a string, the prepositional phrase found during parsing iobjstr a string, the indirect object string iobj an object, the indirect object value
The value returned by the program, if any, is ignored by the server.
MOO stands for "MUD, Object Oriented." MUD, in turn, has been said to stand for many different things, but I tend to think of it as "Multi-User Dungeon" in the spirit of those ancient precursors to MUDs, Adventure and Zork.
MOO, the programming language, is a relatively small and simple object-oriented language designed to be easy to learn for most non-programmers; most complex systems still require some significant programming ability to accomplish, however.
Having given you enough context to allow you to understand exactly what MOO code is doing, I now explain what MOO code looks like and what it means. I begin with the syntax and semantics of expressions, those pieces of code that have values. After that, I cover statements, the next level of structure up from expressions. Next, I discuss the concept of a task, the kind of running process initiated by players entering commands, among other causes. Finally, I list all of the built-in functions available to MOO code and describe what they do.
First, though, let me mention comments. You can include bits of text in your MOO program that are ignored by the server. The idea is to allow you to put in notes to yourself and others about what the code is doing. To do this, begin the text of the comment with the two characters `/*' and end it with the two characters `*/'; this is just like comments in the C programming language. Note that the server will completely ignore that text; it will not be saved in the database. Thus, such comments are only useful in files of code that you maintain outside the database.
To include a more persistent comment in your code, try using a character string literal as a statement. For example, the sentence about peanut butter in the following code is essentially ignored during execution but will be maintained in the database:
for x in (players())
"Grendel eats peanut butter!";
player:tell(x.name, " (", x, ")");
endfor
Expressions are those pieces of MOO code that generate values; for example, the MOO code
3 + 4
is an expression that generates (or "has" or "returns") the value 7. There are many kinds of expressions in MOO, all of them discussed below.
Most kinds of expressions can, under some circumstances, cause an error to be
generated. For example, the expression x / y will generate the error
E_DIV if y is equal to zero. When an expression generates an
error, the behavior of the server is controlled by setting of the `d'
(debug) bit on the verb containing that expression. If the `d' bit is not
set, then the error is effectively squelched immediately upon generation; the
error value is simply returned as the value of the expression that generated
it.
Note: this error-squelching behavior is very error prone, since it affects all errors, including ones the programmer may not have anticipated. The `d' bit exists only for historical reasons; it was once the only way for MOO programmers to catch and handle errors. The error-catching expression and the
try-exceptstatement, both described below, are far better ways of accomplishing the same thing.
If the `d' bit is set, as it usually is, then the error is raised
and can be caught and handled either by code surrounding the expression in
question or by verbs higher up on the chain of calls leading to the current
verb. If the error is not caught, then the server aborts the entire task and,
by default, prints a message to the current player. See the descriptions of
the error-catching expression and the try-except statement for
the details of how errors can be caught, and the chapter on server assumptions
about the database for details on the handling of uncaught errors.
The simplest kind of expression is a literal MOO value, just as described in the section on values at the beginning of this document. For example, the following are all expressions:
17
#893
"This is a character string."
E_TYPE
{"This", "is", "a", "list", "of", "words"}
In the case of lists, like the last example above, note that the list expression contains other expressions, several character strings in this case. In general, those expressions can be of any kind at all, not necessarily literal values. For example,
{3 + 4, 3 - 4, 3 * 4}
is an expression whose value is the list {7, -1, 12}.
As discussed earlier, it is possible to store values in properties on objects; the properties will keep those values forever, or until another value is explicitly put there. Quite often, though, it is useful to have a place to put a value for just a little while. MOO provides local variables for this purpose.
Variables are named places to hold values; you can get and set the value in a given variable as many times as you like. Variables are temporary, though; they only last while a particular verb is running; after it finishes, all of the variables given values there cease to exist and the values are forgotten.
Variables are also "local" to a particular verb; every verb has its own set of them. Thus, the variables set in one verb are not visible to the code of other verbs.
The name for a variable is made up entirely of letters, digits, and the underscore character (`_') and does not begin with a digit. The following are all valid variable names:
foo _foo this2that M68000 two_words This_is_a_very_long_multiword_variable_name
Note that, along with almost everything else in MOO, the case of the letters in variable names is insignificant. For example, these are all names for the same variable:
fubar Fubar FUBAR fUbAr
A variable name is itself an expression; its value is the value of the named
variable. When a verb begins, almost no variables have values yet; if you try
to use the value of a variable that doesn't have one, the error value
E_VARNF is raised. (MOO is unlike many other programming languages in
which one must `declare' each variable before using it; MOO has no such
declarations.) The following variables always have values:
INT FLOAT OBJ STR LIST ERR player this caller verb args argstr dobj dobjstr prepstr iobj iobjstr NUM
The values of some of these variables always start out the same:
INT
typeof(), below)
NUM
INT (for historical reasons)
FLOAT
LIST
STR
OBJ
ERR
For others, the general meaning of the value is consistent, though the value itself is different for different situations:
player
this
caller
verb
args
The rest of the so-called "built-in" variables are only really meaningful for the first verb called for a given command. Their semantics is given in the discussion of command parsing, above.
To change what value is stored in a variable, use an assignment expression:
variable = expression
For example, to change the variable named `x' to have the value 17, you would write `x = 17' as an expression. An assignment expression does two things:
Thus, the expression
13 + (x = 17)
changes the value of `x' to be 17 and returns 30.
All of the usual simple operations on numbers are available to MOO programs:
+ - * / %
These are, in order, addition, subtraction, multiplication, division, and remainder. In the following table, the expressions on the left have the corresponding values on the right:
5 + 2 => 7 5 - 2 => 3 5 * 2 => 10 5 / 2 => 2 5.0 / 2.0 => 2.5 5 % 2 => 1 5.0 % 2.0 => 1.0 5 % -2 => 1 -5 % 2 => -1 -5 % -2 => -1 -(5 + 2) => -7
Note that integer division in MOO throws away the remainder and that the result of the remainder operator (`%') has the same sign as the left-hand operand. Also, note that `-' can be used without a left-hand operand to negate a numeric expression.
Fine point: Integers and floating-point numbers cannot be mixed in any particular use of these arithmetic operators; unlike some other programming languages, MOO does not automatically coerce integers into floating-point numbers. You can use the
tofloat()function to perform an explicit conversion.
The `+' operator can also be used to append two strings. The expression
"foo" + "bar"
has the value
"foobar"
Unless both operands to an arithmetic operator are numbers of the same kind
(or, for `+', both strings), the error value E_TYPE is raised. If
the right-hand operand for the division or remainder operators (`/' or
`%') is zero, the error value E_DIV is raised.
MOO also supports the exponentiation operation, also known as "raising to a power," using the `^' operator:
3 ^ 4 => 81 3 ^ 4.5 error--> E_TYPE 3.5 ^ 4 => 150.0625 3.5 ^ 4.5 => 280.741230801382
Note that if the first operand is an integer, then the second operand must also be an integer. If the first operand is a floating-point number, then the second operand can be either kind of number. Although it is legal to raise an integer to a negative power, it is unlikely to be terribly useful.
Any two values can be compared for equality using `==' and `!='. The first of these returns 1 if the two values are equal and 0 otherwise; the second does the reverse:
3 == 4 => 0
3 != 4 => 1
3 == 3.0 => 0
"foo" == "Foo" => 1
#34 != #34 => 0
{1, #34, "foo"} == {1, #34, "FoO"} => 1
E_DIV == E_TYPE => 0
3 != "foo" => 1
Note that integers and floating-point numbers are never equal to one another, even in the `obvious' cases. Also note that comparison of strings (and list values containing strings) is case-insensitive; that is, it does not distinguish between the upper- and lower-case version of letters. To test two values for case-sensitive equality, use the `equal' function described later.
Warning: It is easy (and very annoying) to confuse the equality-testing operator (`==') with the assignment operator (`='), leading to nasty, hard-to-find bugs. Don't do this.
Numbers, object numbers, strings, and error values can also be compared for ordering purposes using the following operators:
< <= >= >
meaning "less than," "less than or equal," "greater than or equal," and "greater than," respectively. As with the equality operators, these return 1 when their operands are in the appropriate relation and 0 otherwise:
3 < 4 => 1 3 < 4.0 error--> E_TYPE #34 >= #32 => 1 "foo" <= "Boo" => 0 E_DIV > E_TYPE => 1
Note that, as with the equality operators, strings are compared
case-insensitively. To perform a case-sensitive string comparison, use the
`strcmp' function described later. Also note that the error values are
ordered as given in the table in the section on values. If the operands to
these four comparison operators are of different types (even integers and
floating-point numbers are considered different types), or if they are lists,
then E_TYPE is raised.
There is a notion in MOO of true and false values; every value is one or the other. The true values are as follows:
0.0,
All other values are false:
0.0 and -0.0,
There are four kinds of expressions and two kinds of statements that depend upon this classification of MOO values. In describing them, I sometimes refer to the truth value of a MOO value; this is just true or false, the category into which that MOO value is classified.
The conditional expression in MOO has the following form:
expression-1 ? expression-2 | expression-3
First, expression-1 is evaluated. If it returns a true value, then expression-2 is evaluated and whatever it returns is returned as the value of the conditional expression as a whole. If expression-1 returns a false value, then expression-3 is evaluated instead and its value is used as that of the conditional expression.
1 ? 2 | 3 => 2
0 ? 2 | 3 => 3
"foo" ? 17 | {#34} => 17
Note that only one of expression-2 and expression-3 is evaluated, never both.
To negate the truth value of a MOO value, use the `!' operator:
! expression
If the value of expression is true, `!' returns 0; otherwise, it returns 1:
! "foo" => 0 ! (3 >= 4) => 1
The negation operator is usually read as "not."
It is frequently useful to test more than one condition to see if some or all of them are true. MOO provides two operators for this:
expression-1 && expression-2 expression-1 || expression-2
These operators are usually read as "and" and "or," respectively.
The `&&' operator first evaluates expression-1. If it returns a true value, then expression-2 is evaluated and its value becomes the value of the `&&' expression as a whole; otherwise, the value of expression-1 is used as the value of the `&&' expression. Note that expression-2 is only evaluated if expression-1 returns a true value. The `&&' expression is equivalent to the conditional expression
expression-1 ? expression-2 | expression-1
except that expression-1 is only evaluated once.
The `||' operator works similarly, except that expression-2 is evaluated only if expression-1 returns a false value. It is equivalent to the conditional expression
expression-1 ? expression-1 | expression-2
except that, as with `&&', expression-1 is only evaluated once.
These two operators behave very much like "and" and "or" in English:
1 && 1 => 1 0 && 1 => 0 0 && 0 => 0 1 || 1 => 1 0 || 1 => 1 0 || 0 => 0 17 <= 23 && 23 <= 27 => 1
Both strings and lists can be seen as ordered sequences of MOO values. In the
case of strings, each is a sequence of single-character strings; that is, one
can view the string "bar" as a sequence of the strings "b",
"a", and "r". MOO allows you to refer to the elements of lists
and strings by number, by the index of that element in the list or
string. The first element in a list or string has index 1, the second has
index 2, and so on.
The indexing expression in MOO extracts a specified element from a list or string:
expression-1[expression-2]
First, expression-1 is evaluated; it must return a list or a string (the
sequence). Then, expression-2 is evaluated and must return an
integer (the index). If either of the expressions returns some other type
of value, E_TYPE is returned. The index must be between 1 and the
length of the sequence, inclusive; if it is not, then E_RANGE is raised.
The value of the indexing expression is the index'th element in the sequence.
Anywhere within expression-2, you can use the symbol $ as an
expression returning the length of the value of expression-1.
"fob"[2] => "o"
"fob"[1] => "f"
{#12, #23, #34}[$ - 1] => #23
Note that there are no legal indices for the empty string or list, since there are no integers between 1 and 0 (the length of the empty string or list).
Fine point: The
$expression actually returns the length of the value of the expression just before the nearest enclosing[...]indexing or subranging brackets. For example:"frob"[{3, 2, 4}[$]] => "b"
It often happens that one wants to change just one particular slot of a list or string, which is stored in a variable or a property. This can be done conveniently using an indexed assignment having one of the following forms:
variable[index-expr] = result-expr object-expr.name[index-expr] = result-expr object-expr.(name-expr)[index-expr] = result-expr $name[index-expr] = result-expr
The first form writes into a variable, and the last three forms write into a
property. The usual errors (E_TYPE, E_INVIND, E_PROPNF
and E_PERM for lack of read/write permission on the property) may be
raised, just as in reading and writing any object property; see the
discussion of object property expressions below for details. Correspondingly,
if variable does not yet have a value (i.e., it has never been assigned
to), E_VARNF will be raised.
If index-expr is not an integer, or if the value of variable or the
property is not a list or string, E_TYPE is raised. If
result-expr is a string, but not of length 1, E_INVARG is
raised. Now suppose index-expr evaluates to an integer k. If
k is outside the range of the list or string (i.e. smaller than 1 or
greater than the length of the list or string), E_RANGE is raised.
Otherwise, the actual assignment takes place. For lists, the variable or the
property is assigned a new list that is identical to the original one except at
the k-th position, where the new list contains the result of
result-expr instead. For strings, the variable or the property is
assigned a new string that is identical to the original one, except the
k-th character is changed to be result-expr.
The assignment expression itself returns the value of result-expr. For
the following examples, assume that l initially contains the list
{1, 2, 3} and that s initially contains the string "foobar":
l[5] = 3 error--> E_RANGE
l["first"] = 4 error--> E_TYPE
s[3] = "baz" error--> E_INVARG
l[2] = l[2] + 3 => 5
l => {1, 5, 3}
l[2] = "foo" => "foo"
l => {1, "foo", 3}
s[2] = "u" => "u"
s => "fuobar"
s[$] = "z" => "z"
s => "fuobaz"
Note that the $ expression may also be used in indexed assignments with
the same meaning as before.
Fine point: After an indexed assignment, the variable or property contains a new list or string, a copy of the original list in all but the k-th place, where it contains a new value. In programming-language jargon, the original list is not mutated, and there is no aliasing. (Indeed, no MOO value is mutable and no aliasing ever occurs.)
In the list case, indexed assignment can be nested to many levels, to work on
nested lists. Assume that l initially contains the list
{{1, 2, 3}, {4, 5, 6}, "foo"}
in the following examples:
l[7] = 4 error--> E_RANGE
l[1][8] = 35 error--> E_RANGE
l[3][2] = 7 error--> E_TYPE
l[1][1][1] = 3 error--> E_TYPE
l[2][2] = -l[2][2] => -5
l => {{1, 2, 3}, {4, -5, 6}, "foo"}
l[2] = "bar" => "bar"
l => {{1, 2, 3}, "bar", "foo"}
l[2][$] = "z" => "z"
l => {{1, 2, 3}, "baz", "foo"}
The first two examples raise E_RANGE because 7 is out of the range of
l and 8 is out of the range of l[1]. The next two examples
raise E_TYPE because l[3] and l[1][1] are not lists.
The range expression extracts a specified subsequence from a list or string:
expression-1[expression-2..expression-3]
The three expressions are evaluated in order. Expression-1 must return a
list or string (the sequence) and the other two expressions must return
integers (the low and high indices, respectively); otherwise,
E_TYPE is raised. The $ expression can be used in either or both
of expression-2 and expression-3 just as before, meaning the length
of the value of expression-1.
If the low index is greater than the high index, then the empty string or list
is returned, depending on whether the sequence is a string or a list.
Otherwise, both indices must be between 1 and the length of the sequence;
E_RANGE is raised if they are not. A new list or string is returned
that contains just the elements of the sequence with indices between the low
and high bounds.
"foobar"[2..$] => "oobar"
"foobar"[3..3] => "o"
"foobar"[17..12] => ""
{"one", "two", "three"}[$ - 1..$] => {"two", "three"}
{"one", "two", "three"}[3..3] => {"three"}
{"one", "two", "three"}[17..12] => {}
The subrange assigment replaces a specified subsequence of a list or string with a supplied subsequence. The allowed forms are:
variable[start-index-expr..end-index-expr] = result-expr object-expr.name[start-index-expr..end-index-expr] = result-expr object-expr.(name-expr)[start-index-expr..end-index-expr] = result-expr $name[start-index-expr..end-index-expr] = result-expr
As with indexed assigments, the first form writes into a variable, and the last
three forms write into a property. The same errors (E_TYPE,
E_INVIND, E_PROPNF and E_PERM for lack of read/write
permission on the property) may be raised. If variable does not yet have
a value (i.e., it has never been assigned to), E_VARNF will be raised.
As before, the $ expression can be used in either start-index-expr
or end-index-expr, meaning the length of the original value of the
expression just before the [...] part.
If start-index-expr or end-index-expr is not an integer, if the value
of variable or the property is not a list or string, or result-expr
is not the same type as variable or the property, E_TYPE is
raised. E_RANGE is raised if end-index-expr is less than zero
or if start-index-expr is greater than the length of the list or string
plus one. Note: the length of result-expr does not need to be the same
as the length of the specified range.
In precise terms, the subrange assigment
v[start..end] = value
is equivalent to
v = {@v[1..start - 1], @value, @v[end + 1..$]}
if v is a list and to
v = v[1..start - 1] + value + v[end + 1..$]
if v is a string.
The assigment expression itself returns the value of result-expr. For
the following examples, assume that l initially contains the list
{1, 2, 3} and that s initially contains the string "foobar":
l[5..6] = {7, 8} error--> E_RANGE
l[2..3] = 4 error--> E_TYPE
l[#2..3] = {7} error--> E_TYPE
s[2..3] = {6} error--> E_TYPE
l[2..3] = {6, 7, 8, 9} => {6, 7, 8, 9}
l => {1, 6, 7, 8, 9}
l[2..1] = {10, "foo"} => {10, "foo"}
l => {1, 10, "foo", 6, 7, 8, 9}
l[3][2..$] = "u" => "u"
l => {1, 10, "fu", 6, 7, 8, 9}
s[7..12] = "baz" => "baz"
s => "foobarbaz"
s[1..3] = "fu" => "fu"
s => "fubarbaz"
s[1..0] = "test" => "test"
s => "testfubarbaz"
As was mentioned earlier, lists can be constructed by writing a comma-separated sequence of expressions inside curly braces:
{expression-1, expression-2, ..., expression-N}
The resulting list has the value of expression-1 as its first element, that of expression-2 as the second, etc.
{3 < 4, 3 <= 4, 3 >= 4, 3 > 4} => {1, 1, 0, 0}
Additionally, one may precede any of these expressions by the splicing
operator, `@'. Such an expression must return a list; rather than the
old list itself becoming an element of the new list, all of the elements of
the old list are included in the new list. This concept is easy to
understand, but hard to explain in words, so here are some examples. For
these examples, assume that the variable a has the value {2, 3,
4} and that b has the value {"Foo", "Bar"}:
{1, a, 5} => {1, {2, 3, 4}, 5}
{1, @a, 5} => {1, 2, 3, 4, 5}
{a, @a} => {{2, 3, 4}, 2, 3, 4}
{@a, @b} => {2, 3, 4, "Foo", "Bar"}
If the splicing operator (`@') precedes an expression whose value
is not a list, then E_TYPE is raised as the value of the list
construction as a whole.
The list membership expression tests whether or not a given MOO value is an element of a given list and, if so, with what index:
expression-1 in expression-2
Expression-2 must return a list; otherwise, E_TYPE is raised.
If the value of expression-1 is in that list, then the index of its first
occurrence in the list is returned; otherwise, the `in' expression returns
0.
2 in {5, 8, 2, 3} => 3
7 in {5, 8, 2, 3} => 0
"bar" in {"Foo", "Bar", "Baz"} => 2
Note that the list membership operator is case-insensitive in comparing strings, just like the comparison operators. To perform a case-sensitive list membership test, use the `is_member' function described later. Note also that since it returns zero only if the given value is not in the given list, the `in' expression can be used either as a membership test or as an element locator.
It is often the case in MOO programming that you will want to access the elements of a list individually, with each element stored in a separate variables. This desire arises, for example, at the beginning of almost every MOO verb, since the arguments to all verbs are delivered all bunched together in a single list. In such circumstances, you could write statements like these:
first = args[1]; second = args[2]; if (length(args) > 2) third = args[3]; else third = 0; endif
This approach gets pretty tedious, both to read and to write, and it's prone to errors if you mistype one of the indices. Also, you often want to check whether or not any extra list elements were present, adding to the tedium.
MOO provides a special kind of assignment expression, called scattering assignment made just for cases such as these. A scattering assignment expression looks like this:
{target, ...} = expr
where each target describes a place to store elements of the list that results from evaluating expr. A target has one of the following forms:
variable
?variable
?variable = default-expr
@variable
@ syntax in list construction, this variable is
assigned a list of all of the `leftover' list elements in this part of the list
after all of the other targets have been filled in. It is assigned the empty
list if there aren't any elements left over. This is called a rest
target, since it gets the rest of the elements. There may be at most one rest
target in each scattering assignment expression.
If there aren't enough list elements to fill all of the required targets, or if
there are more than enough to fill all of the required and optional targets but
there isn't a rest target to take the leftover ones, then E_ARGS is
raised.
Here are some examples of how this works. Assume first that the verb
me:foo() contains the following code:
b = c = e = 17;
{a, ?b, ?c = 8, @d, ?e = 9, f} = args;
return {a, b, c, d, e, f};
Then the following calls return the given values:
me:foo(1) error--> E_ARGS
me:foo(1, 2) => {1, 17, 8, {}, 9, 2}
me:foo(1, 2, 3) => {1, 2, 8, {}, 9, 3}
me:foo(1, 2, 3, 4) => {1, 2, 3, {}, 9, 4}
me:foo(1, 2, 3, 4, 5) => {1, 2, 3, {}, 4, 5}
me:foo(1, 2, 3, 4, 5, 6) => {1, 2, 3, {4}, 5, 6}
me:foo(1, 2, 3, 4, 5, 6, 7) => {1, 2, 3, {4, 5}, 6, 7}
me:foo(1, 2, 3, 4, 5, 6, 7, 8) => {1, 2, 3, {4, 5, 6}, 7, 8}
Using scattering assignment, the example at the begining of this section could be rewritten more simply, reliably, and readably:
{first, second, ?third = 0} = args;
It is good MOO programming style to use a scattering assignment at the top of nearly every verb, since it shows so clearly just what kinds of arguments the verb expects.
Usually, one can read the value of a property on an object with a simple expression:
expression.name
Expression must return an object number; if not, E_TYPE is
raised. If the object with that number does not exist, E_INVIND is
raised. Otherwise, if the object does not have a property with that name,
then E_PROPNF is raised. Otherwise, if the named property is not
readable by the owner of the current verb, then E_PERM is raised.
Finally, assuming that none of these terrible things happens, the value of the
named property on the given object is returned.
I said "usually" in the paragraph above because that simple expression only works if the name of the property obeys the same rules as for the names of variables (i.e., consists entirely of letters, digits, and underscores, and doesn't begin with a digit). Property names are not restricted to this set, though. Also, it is sometimes useful to be able to figure out what property to read by some computation. For these more general uses, the following syntax is also allowed:
expression-1.(expression-2)
As before, expression-1 must return an object number. Expression-2
must return a string, the name of the property to be read; E_TYPE
is raised otherwise. Using this syntax, any property can be read,
regardless of its name.
Note that, as with almost everything in MOO, case is not significant in the names of properties. Thus, the following expressions are all equivalent:
foo.bar
foo.Bar
foo.("bAr")
The LambdaCore database uses several properties on #0, the system
object, for various special purposes. For example, the value of
#0.room is the "generic room" object, #0.exit is the "generic
exit" object, etc. This allows MOO programs to refer to these useful objects
more easily (and more readably) than using their object numbers directly. To
make this usage even easier and more readable, the expression
$name
(where name obeys the rules for variable names) is an abbreviation for
#0.name
Thus, for example, the value $nothing mentioned earlier is really
#-1, the value of #0.nothing.
As with variables, one uses the assignment operator (`=') to change the value of a property. For example, the expression
14 + (#27.foo = 17)
changes the value of the `foo' property of the object numbered 27 to be
17 and then returns 31. Assignments to properties check that the owner of the
current verb has write permission on the given property, raising
E_PERM otherwise. Read permission is not required.
MOO provides a large number of useful functions for performing a wide variety of operations; a complete list, giving their names, arguments, and semantics, appears in a separate section later. As an example to give you the idea, there is a function named `length' that returns the length of a given string or list.
The syntax of a call to a function is as follows:
name(expr-1, expr-2, ..., expr-N)
where name is the name of one of the built-in functions. The
expressions between the parentheses, called arguments, are each
evaluated in turn and then given to the named function to use in its
appropriate way. Most functions require that a specific number of arguments
be given; otherwise, E_ARGS is raised. Most also require that
certain of the arguments have certain specified types (e.g., the
length() function requires a list or a string as its argument);
E_TYPE is raised if any argument has the wrong type.
As with list construction, the splicing operator `@' can precede
any argument expression. The value of such an expression must be a
list; E_TYPE is raised otherwise. The elements of this list
are passed as individual arguments, in place of the list as a whole.
Verbs can also call other verbs, usually using this syntax:
expr-0:name(expr-1, expr-2, ..., expr-N)
Expr-0 must return an object number; E_TYPE is raised otherwise.
If the object with that number does not exist, E_INVIND is raised. If
this task is too deeply nested in verbs calling verbs calling verbs, then
E_MAXREC is raised; the default limit is 50 levels, but this can be
changed from within the database; see the chapter on server assumptions about
the database for details. If neither the object nor any of its ancestors
defines a verb matching the given name, E_VERBNF is raised.
Otherwise, if none of these nasty things happens, the named verb on the given
object is called; the various built-in variables have the following initial
values in the called verb:
this
verb
args
caller
this in the calling verb
player
All other built-in variables (argstr, dobj, etc.) are initialized
with the same values they have in the calling verb.
As with the discussion of property references above, I said "usually" at the beginning of the previous paragraph because that syntax is only allowed when the name follows the rules for allowed variable names. Also as with property reference, there is a syntax allowing you to compute the name of the verb:
expr-0:(expr-00)(expr-1, expr-2, ..., expr-N)
The expression expr-00 must return a string; E_TYPE is raised
otherwise.
The splicing operator (`@') can be used with verb-call arguments, too, just as with the arguments to built-in functions.
In many databases, a number of important verbs are defined on #0, the
system object. As with the `$foo' notation for properties on
#0, the server defines a special syntax for calling verbs on #0:
$name(expr-1, expr-2, ..., expr-N)
(where name obeys the rules for variable names) is an abbreviation for
#0:name(expr-1, expr-2, ..., expr-N)
It is often useful to be able to catch an error that an expression raises, to keep the error from aborting the whole task, and to keep on running as if the expression had returned some other value normally. The following expression accomplishes this:
` expr-1 ! codes => expr-2 '
Note: the open- and close-quotation marks in the previous line are really part of the syntax; you must actually type them as part of your MOO program for this kind of expression.
The codes part is either the keyword ANY or else a
comma-separated list of expressions, just like an argument list. As in an
argument list, the splicing operator (`@') can be used here. The
=> expr-2 part of the error-catching expression is optional.
First, the codes part is evaluated, yielding a list of error codes that
should be caught if they're raised; if codes is ANY, then it is
equivalent to the list of all possible MOO values.
Next, expr-1 is evaluated. If it evaluates normally, without raising an error, then its value becomes the value of the entire error-catching expression. If evaluating expr-1 results in an error being raised, then call that error E. If E is in the list resulting from evaluating codes, then E is considered caught by this error-catching expression. In such a case, if expr-2 was given, it is evaluated to get the outcome of the entire error-catching expression; if expr-2 was omitted, then E becomes the value of the entire expression. If E is not in the list resulting from codes, then this expression does not catch the error at all and it continues to be raised, possibly to be caught by some piece of code either surrounding this expression or higher up on the verb-call stack.
Here are some examples of the use of this kind of expression:
`x + 1 ! E_TYPE => 0'
Returns x + 1 if x is an integer, returns 0 if x is
not an integer, and raises E_VARNF if x doesn't have a value.
`x.y ! E_PROPNF, E_PERM => 17'
Returns x.y if that doesn't cause an error, 17 if x
doesn't have a y property or that property isn't readable, and raises
some other kind of error (like E_INVIND) if x.y does.
`1 / 0 ! ANY'
Returns E_DIV.
As shown in a few examples above, MOO allows you to use parentheses to make it clear how you intend for complex expressions to be grouped. For example, the expression
3 * (4 + 5)
performs the addition of 4 and 5 before multiplying the result by 3.
If you leave out the parentheses, MOO will figure out how to group the expression according to certain rules. The first of these is that some operators have higher precedence than others; operators with higher precedence will more tightly bind to their operands than those with lower precedence. For example, multiplication has higher precedence than addition; thus, if the parentheses had been left out of the expression in the previous paragraph, MOO would have grouped it as follows:
(3 * 4) + 5
The table below gives the relative precedence of all of the MOO operators; operators on higher lines in the table have higher precedence and those on the same line have identical precedence:
! - (without a left operand) ^ * / % + - == != < <= > >= in && || ... ? ... | ... (the conditional expression) =
Thus, the horrendous expression
x = a < b && c > d + e * f ? w in y | - q - r
would be grouped as follows:
x = (((a < b) && (c > (d + (e * f)))) ? (w in y) | ((- q) - r))
It is best to keep expressions simpler than this and to use parentheses liberally to make your meaning clear to other humans.
Statements are MOO constructs that, in contrast to expressions, perform some useful, non-value-producing operation. For example, there are several kinds of statements, called `looping constructs', that repeatedly perform some set of operations. Fortunately, there are many fewer kinds of statements in MOO than there are kinds of expressions.
Statements do not return values, but some kinds of statements can, under certain circumstances described below, generate errors. If such an error is generated in a verb whose `d' (debug) bit is not set, then the error is ignored and the statement that generated it is simply skipped; execution proceeds with the next statement.
Note: this error-ignoring behavior is very error prone, since it affects all errors, including ones the programmer may not have anticipated. The `d' bit exists only for historical reasons; it was once the only way for MOO programmers to catch and handle errors. The error-catching expression and the
try-exceptstatement are far better ways of accomplishing the same thing.
If the `d' bit is set, as it usually is, then the error is raised
and can be caught and handled either by code surrounding the expression in
question or by verbs higher up on the chain of calls leading to the current
verb. If the error is not caught, then the server aborts the entire task and,
by default, prints a message to the current player. See the descriptions of
the error-catching expression and the try-except statement for
the details of how errors can be caught, and the chapter on server assumptions
about the database for details on the handling of uncaught errors.
The simplest kind of statement is the null statement, consisting of just a semicolon:
;
It doesn't do anything at all, but it does it very quickly.
The next simplest statement is also one of the most common, the expression statement, consisting of any expression followed by a semicolon:
expression;
The given expression is evaluated and the resulting value is ignored. Commonly-used kinds of expressions for such statements include assignments and verb calls. Of course, there's no use for such a statement unless the evaluation of expression has some side-effect, such as changing the value of some variable or property, printing some text on someone's screen, etc.
The `if' statement allows you to decide whether or not to perform some statements based on the value of an arbitrary expression:
if (expression) statements endif
Expression is evaluated and, if it returns a true value, the statements are executed in order; otherwise, nothing more is done.
One frequently wants to perform one set of statements if some condition is true and some other set of statements otherwise. The optional `else' phrase in an `if' statement allows you to do this:
if (expression) statements-1 else statements-2 endif
This statement is executed just like the previous one, except that statements-1 are executed if expression returns a true value and statements-2 are executed otherwise.
Sometimes, one needs to test several conditions in a kind of nested fashion:
if (expression-1)
statements-1
else
if (expression-2)
statements-2
else
if (expression-3)
statements-3
else
statements-4
endif
endif
endif
Such code can easily become tedious to write and difficult to read. MOO provides a somewhat simpler notation for such cases:
if (expression-1) statements-1 elseif (expression-2) statements-2 elseif (expression-3) statements-3 else statements-4 endif
Note that `elseif' is written as a single word, without any spaces. This simpler version has the very same meaning as the original: evaluate expression-i for i equal to 1, 2, and 3, in turn, until one of them returns a true value; then execute the statements-i associated with that expression. If none of the expression-i return a true value, then execute statements-4.
Any number of `elseif' phrases can appear, each having this form:
elseif (expression) statements
The complete syntax of the `if' statement, therefore, is as follows:
if (expression) statements zero-or-more-elseif-phrases an-optional-else-phrase endif
MOO provides three different kinds of looping statements, allowing you to have a set of statements executed (1) once for each element of a given list, (2) once for each integer or object number in a given range, and (3) over and over until a given condition stops being true.
To perform some statements once for each element of a given list, use this syntax:
for variable in (expression) statements endfor
The expression is evaluated and should return a list; if it does not,
E_TYPE is raised. The statements are then executed once for
each element of that list in turn; each time, the given variable is
assigned the value of the element in question. For example, consider
the following statements:
odds = {1, 3, 5, 7, 9};
evens = {};
for n in (odds)
evens = {@evens, n + 1};
endfor
The value of the variable `evens' after executing these statements is the list
{2, 4, 6, 8, 10}
To perform a set of statements once for each integer or object number in a given range, use this syntax:
for variable in [expression-1..expression-2] statements endfor
The two expressions are evaluated in turn and should either both return integers
or both return object numbers; E_TYPE is raised otherwise. The
statements are then executed once for each integer (or object number, as
appropriate) greater than or equal to the value of expression-1 and less
than or equal to the result of expression-2, in increasing order. Each
time, the given variable is assigned the integer or object number in question.
For example, consider the following statements:
evens = {};
for n in [1..5]
evens = {@evens, 2 * n};
endfor
The value of the variable `evens' after executing these statements is just as in the previous example: the list
{2, 4, 6, 8, 10}
The following loop over object numbers prints out the number and name of every valid object in the database:
for o in [#0..max_object()]
if (valid(o))
notify(player, tostr(o, ": ", o.name));
endif
endfor
The final kind of loop in MOO executes a set of statements repeatedly as long as a given condition remains true:
while (expression) statements endwhile
The expression is evaluated and, if it returns a true value, the statements are executed; then, execution of the `while' statement begins all over again with the evaluation of the expression. That is, execution alternates between evaluating the expression and executing the statements until the expression returns a false value. The following example code has precisely the same effect as the loop just shown above:
evens = {};
n = 1;
while (n <= 5)
evens = {@evens, 2 * n};
n = n + 1;
endwhile
Fine point: It is also possible to give a `name' to a `while' loop, using this syntax:
while name (expression) statements endwhilewhich has precisely the same effect as
while (name = expression) statements endwhileThis naming facility is only really useful in conjunction with the `break' and `continue' statements, described in the next section.
With each kind of loop, it is possible that the statements in the body of the loop will never be executed at all. For iteration over lists, this happens when the list returned by the expression is empty. For iteration on integers, it happens when expression-1 returns a larger integer than expression-2. Finally, for the `while' loop, it happens if the expression returns a false value the very first time it is evaluated.
Sometimes, it is useful to exit a loop before it finishes all of its iterations. For example, if the loop is used to search for a particular kind of element of a list, then it might make sense to stop looping as soon as the right kind of element is found, even if there are more elements yet to see. The `break' statement is used for this purpose; it has the form
break;
or
break name;
Each `break' statement indicates a specific surrounding loop; if name is not given, the statement refers to the innermost one. If it is given, name must be the name appearing right after the `for' or `while' keyword of the desired enclosing loop. When the `break' statement is executed, the indicated loop is immediately terminated and executing continues just as if the loop had completed its iterations normally.
MOO also allows you to terminate just the current iteration of a loop, making it immediately go on to the next one, if any. The `continue' statement does this; it has precisely the same forms as the `break' statement:
continue;
or
continue name;
The MOO program in a verb is just a sequence of statements. Normally, when the verb is called, those statements are simply executed in order and then the integer 0 is returned as the value of the verb-call expression. Using the `return' statement, one can change this behavior. The `return' statement has one of the following two forms:
return;
or
return expression;
When it is executed, execution of the current verb is terminated immediately after evaluating the given expression, if any. The verb-call expression that started the execution of this verb then returns either the value of expression or the integer 0, if no expression was provided.
Normally, whenever a piece of MOO code raises an error, the entire task is
aborted and a message printed to the user. Often, such errors can be
anticipated in advance by the programmer and code written to deal with them in
a more graceful manner. The try-except statement allows you to
do this; the syntax is as follows:
try statements-0 except variable-1 (codes-1) statements-1 except variable-2 (codes-2) statements-2 ... endtry
where the variables may be omitted and each codes part is either
the keyword ANY or else a comma-separated list of expressions, just like
an argument list. As in an argument list, the splicing operator (`@')
can be used here. There can be anywhere from 1 to 255 except clauses.
First, each codes part is evaluated, yielding a list of error codes that
should be caught if they're raised; if a codes is ANY, then it is
equivalent to the list of all possible MOO values.
Next, statements-0 is executed; if it doesn't raise an error, then that's
all that happens for the entire try-except statement. Otherwise,
let E be the error it raises. From top to bottom, E is searched
for in the lists resulting from the various codes parts; if it isn't
found in any of them, then it continues to be raised, possibly to be caught by
some piece of code either surrounding this try-except statement
or higher up on the verb-call stack.
If E is found first in codes-i, then variable-i (if provided) is assigned a value containing information about the error being raised and statements-i is executed. The value assigned to variable-i is a list of four elements:
{code, message, value, traceback}
where code is E, the error being raised, message and
value are as provided by the code that raised the error, and
traceback is a list like that returned by the `callers()' function,
including line numbers. The traceback list contains entries for every
verb from the one that raised the error through the one containing this
try-except statement.
Unless otherwise mentioned, all of the built-in errors raised by expressions,
statements, and functions provide tostr(code) as message and
zero as value.
Here's an example of the use of this kind of statement:
try
result = object:(command)(@arguments);
player:tell("=> ", toliteral(result));
except v (ANY)
tb = v[4];
if (length(tb) == 1)
player:tell("** Illegal command: ", v[2]);
else
top = tb[1];
tb[1..1] = {};
player:tell(top[1], ":", top[2], ", line ", top[6], ":",
v[2]);
for fr in (tb)
player:tell("... called from ", fr[1], ":", fr[2],
", line ", fr[6]);
endfor
player:tell("(End of traceback)");
endif
endtry
Whenever an error is raised, it is usually the case that at least some MOO code
gets skipped over and never executed. Sometimes, it's important that a piece
of code always be executed, whether or not an error is raised. Use the
try-finally statement for these cases; it has the following
syntax:
try statements-1 finally statements-2 endtry
First, statements-1 is executed; if it completes without raising an
error, returning from this verb, or terminating the current iteration of a
surrounding loop (we call these possibilities transferring control), then
statements-2 is executed and that's all that happens for the entire
try-finally statement.
Otherwise, the process of transferring control is interrupted and statments-2 is executed. If statements-2 itself completes without transferring control, then the interrupted control transfer is resumed just where it left off. If statements-2 does transfer control, then the interrupted transfer is simply forgotten in favor of the new one.
In short, this statement ensures that statements-2 is executed after control leaves statements-1 for whatever reason; it can thus be used to make sure that some piece of cleanup code is run even if statements-1 doesn't simply run normally to completion.
Here's an example:
try start = time(); object:(command)(@arguments); finally end = time(); this:charge_user_for_seconds(player, end - start); endtry
It is sometimes useful to have some sequence of statements execute at a later time, without human intervention. For example, one might implement an object that, when thrown into the air, eventually falls back to the ground; the `throw' verb on that object should arrange to print a message about the object landing on the ground, but the message shouldn't be printed until some number of seconds have passed.
The `fork' statement is intended for just such situations and has the following syntax:
fork (expression) statements endfork
The `fork' statement first executes the expression, which must return a integer; call that integer n. It then creates a new MOO task that will, after at least n seconds, execute the statements. When the new task begins, all variables will have the values they had at the time the `fork' statement was executed. The task executing the `fork' statement immediately continues execution. The concept of tasks is discussed in detail in the next section.
By default, there is no limit to the number of tasks any player may fork, but such a limit can be imposed from within the database. See the chapter on server assumptions about the database for details.
Occasionally, one would like to be able to kill a forked task before it even starts; for example, some player might have caught the object that was thrown into the air, so no message should be printed about it hitting the ground. If a variable name is given after the `fork' keyword, like this:
fork name (expression) statements endfork
then that variable is assigned the task ID of the newly-created task.
The value of this variable is visible both to the task executing the fork
statement and to the statements in the newly-created task. This ID can be
passed to the kill_task() function to keep the task from running and
will be the value of task_id() once the task begins execution.
A task is an execution of a MOO program. There are five kinds of tasks in LambdaMOO:
suspend() function suspends the execution of the current task. A
snapshot is taken of whole state of the execution, and the execution will be
resumed later. These are called suspended tasks.
read() function also suspends the execution of the current task, in
this case waiting for the player to type a line of input. When the line is
received, the task resumes with the read() function returning the input
line as result. These are called reading tasks.
The last three kinds of tasks above are collectively known as queued tasks or background tasks, since they may not run immediately.
To prevent a maliciously- or incorrectly-written MOO program from running forever and monopolizing the server, limits are placed on the running time of every task. One limit is that no task is allowed to run longer than a certain number of seconds; command and server tasks get five seconds each while other tasks get only three seconds. This limit is, in practice, rarely reached. The reason is that there is also a limit on the number of operations a task may execute.
The server counts down ticks as any task executes. Roughly speaking, it counts one tick for every expression evaluation (other than variables and literals), one for every `if', `fork' or `return' statement, and one for every iteration of a loop. If the count gets all the way down to zero, the task is immediately and unceremoniously aborted. By default, command and server tasks begin with an store of 30,000 ticks; this is enough for almost all normal uses. Forked, suspended, and reading tasks are allotted 15,000 ticks each.
These limits on seconds and ticks may be changed from within the database, as can the behavior of the server after it aborts a task for running out; see the chapter on server assumptions about the database for details.
Because queued tasks may exist for long periods of time before they begin execution, there are functions to list the ones that you own and to kill them before they execute. These functions, among others, are discussed in the following section.
There are a large number of built-in functions available for use by MOO programmers. Each one is discussed in detail in this section. The presentation is broken up into subsections by grouping together functions with similar or related uses.
For most functions, the expected types of the arguments are given; if the
actual arguments are not of these types, E_TYPE is raised. Some
arguments can be of any type at all; in such cases, no type specification is
given for the argument. Also, for most functions, the type of the result of
the function is given. Some functions do not return a useful result; in such
cases, the specification `none' is used. A few functions can potentially
return any type of value at all; in such cases, the specification `value'
is used.
Most functions take a certain fixed number of required arguments and, in some
cases, one or two optional arguments. If a function is called with too many or
too few arguments, E_ARGS is raised.
Functions are always called by the program for some verb; that program is
running with the permissions of some player, usually the owner of the verb in
question (it is not always the owner, though; wizards can use
set_task_perms() to change the permissions `on the fly'). In the
function descriptions below, we refer to the player whose permissions are being
used as the programmer.
Many built-in functions are described below as raising E_PERM unless
the programmer meets certain specified criteria. It is possible to restrict
use of any function, however, so that only wizards can use it; see the chapter
on server assumptions about the database for details.
One of the most important facilities in an object-oriented programming language
is ability for a child object to make use of a parent's implementation of some
operation, even when the child provides its own definition for that operation.
The pass() function provides this facility in MOO.
this.description; this verb is used by the implementation of the
look command. In many cases, a programmer would like the description of
some object to include some non-constant part; for example, a sentence about
whether or not the object was `awake' or `sleeping'. This sentence should be
added onto the end of the normal description. The programmer would like to
have a means of calling the normal description verb and then appending
the sentence onto the end of that description. The function `pass()' is
for exactly such situations.
pass calls the verb with the same name as the current verb but as
defined on the parent of the object that defines the current verb. The
arguments given to pass are the ones given to the called verb and the
returned value of the called verb is returned from the call to pass.
The initial value of this in the called verb is the same as in the
calling verb.
Thus, in the example above, the child-object's description verb might
have the following implementation:
return pass() + " It is " + (this.awake ? "awake." | "sleeping.");
That is, it calls its parent's description verb and then appends to the
result a sentence whose content is computed based on the value of a property on
the object.
In almost all cases, you will want to call `pass()' with the same
arguments as were given to the current verb. This is easy to write in MOO;
just call pass(@args).
There are several functions for performing primitive operations on MOO values, and they can be cleanly split into two kinds: those that do various very general operations that apply to all types of values, and those that are specific to one particular type. There are so many operations concerned with objects that we do not list them in this section but rather give them their own section following this one.
INT, FLOAT, STR, LIST, OBJ, or ERR.
Thus, one usually writes code like this:
if (typeof(x) == LIST) ...
and not like this:
if (typeof(x) == 3) ...
because the former is much more readable than the latter.
tostr(17) => "17"
tostr(1.0/3.0) => "0.333333333333333"
tostr(#17) => "#17"
tostr("foo") => "foo"
tostr({1, 2}) => "{list}"
tostr(E_PERM) => "Permission denied"
tostr("3 + 4 = ", 3 + 4) => "3 + 4 = 7"
Note that tostr() does not do a good job of converting lists into
strings; all lists, including the empty list, are converted into the string
"{list}". The function toliteral(), below, is better for this
purpose.
toliteral(17) => "17"
toliteral(1.0/3.0) => "0.333333333333333"
toliteral(#17) => "#17"
toliteral("foo") => "\"foo\""
toliteral({1, 2}) => "{1, 2}"
toliteral(E_PERM) => "E_PERM"
<= as the errors themselves. Toint() raises
E_TYPE if value is a list. If value is a string but the
string does not contain a syntactically-correct number, then toint()
returns 0.
toint(34.7) => 34
toint(-34.7) => -34
toint(#34) => 34
toint("34") => 34
toint("34.7") => 34
toint(" - 34 ") => -34
toint(E_TYPE) => 1
toint() except
that for strings, the number may be preceded by `#'.
toobj("34") => #34
toobj("#34") => #34
toobj("foo") => #0
toobj({1, 2}) error--> E_TYPE
toint() and then converted as integers are. Tofloat() raises
E_TYPE if value is a list. If value is a string but the
string does not contain a syntactically-correct number, then tofloat()
returns 0.
tofloat(34) => 34.0
tofloat(#34) => 34.0
tofloat("34") => 34.0
tofloat("34.7") => 34.7
tofloat(E_TYPE) => 1.0
value1 == value2"
except that, unlike ==, the equal() function does not treat
upper- and lower-case characters in strings as equal.
"Foo" == "foo" => 1
equal("Foo", "foo") => 0
equal("Foo", "Foo") => 1
string_hash(toliteral(value)); see the
description of string_hash() for details.
E_INVARG is raised. An
integer is chosen randomly from the range [1..mod] and returned.
If mod is not provided, it defaults to the largest MOO integer,
2147483647.
E_TYPE is raised.
-x; otherwise, the result is x. The number x can
be either integer or floating-point; the result is of the same kind.
tostr() or toliteral(). Precision is the number of digits
to appear to the right of the decimal point, capped at 4 more than the maximum
available precision, a total of 19 on most machines; this makes it possible to
avoid rounding errors if the resulting string is subsequently read back as a
floating-point value. If scientific is false or not provided, the result
is a string in the form "MMMMMMM.DDDDDD", preceded by a minus sign if
and only if x is negative. If scientific is provided and true, the
result is a string in the form "M.DDDDDDe+EEE", again preceded by a
minus sign if and only if x is negative.
E_INVARG if x is
negative.
[-pi/2..pi/2] or [0..pi], respectively. Raises
E_INVARG if x is outside the range [-1.0..1.0].
[-pi/2..pi/2] if x is not provided, or of y/x
in the range [-pi..pi] if x is provided.
E_INVARG if
x is not positive.
ceil(); otherwise it is equivalent to floor().
length(); see the description in the next section.
length("foo") => 3
length("") => 0
strsub("%n is a fink.", "%n", "Fred") => "Fred is a fink."
strsub("foobar", "OB", "b") => "fobar"
strsub("foobar", "OB", "b", 1) => "foobar"
index() (rindex()) returns the index of the first
character of the first (last) occurrence of str2 in str1, or zero
if str2 does not occur in str1 at all. By default the search for
an occurrence of str2