A new implementation of the LambdaMOO server
This is experimental software. While it is now mostly functional, it is not yet fully complete.
Until such time as the EtaMOO database format is well tested and considered stable, please make and keep LambdaMOO-format backup copies of your EtaMOO databases.
EtaMOO is a new implementation of the LambdaMOO server written in Haskell.
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.
EtaMOO differs from LambdaMOO in a few significant ways:
EtaMOO is multi-threaded. MOO tasks run concurrently, producing network output and changes to the database in isolated transactions that are committed only when not in conflict with any other transaction. (In cases of conflict, transactions are automatically retried.) Separate threads are also used for network connection management, so for example name lookups do not block the entire server.
EtaMOO uses LMDB as a persistent backing store for the MOO database. Changes are committed on an ongoing basis for instantaneous crash recovery; checkpoints merely perform a quick synchronization, and are otherwise unnecessary. EtaMOO provides mechanisms for importing and exporting LambdaMOO-format databases to and from the EtaMOO-native format.
EtaMOO is Unicode-aware, and will eventually include support for Unicode MOO strings via compile-time build option.
EtaMOO supports 64-bit MOO integers via compile-time build option.
EtaMOO natively supports string-key association lists with efficient
lookup and update operations; the list index syntax has been extended to
allow alist
[
key
]
and alist
[
key
] =
value
for
string key
s whenever alist
is a well-formed association list.
EtaMOO supports several additional hashing algorithms besides MD5,
including SHA-1, SHA-2, SHA-3, and RIPEMD-160, via optional argument to
string_hash()
, binary_hash()
, and value_hash()
. Hash digests may
also optionally be returned as binary strings.
EtaMOO internally handles binary strings in an efficient manner, and only
translates to and from the special MOO binary string syntax upon demand.
For example, passing a binary string read from the network directly to
decode_binary()
does not suffer a round trip through the binary string
representation.
EtaMOO supports fractional second delays in suspend()
and fork
.
EtaMOO supports IPv6.
The implementation of EtaMOO otherwise closely follows the specifications of the LambdaMOO Programmer's Manual, and should be compatible with most LambdaMOO databases as of about version 1.8.3 of the LambdaMOO server code.
EtaMOO is built with Cabal, the Haskell package manager. In the simplest case, running:
cabal install EtaMOO
should automatically download, build, and install the etamoo
executable
after doing the same for all of its Haskell dependencies.
Cabal itself is part of the Haskell Platform which is available for many distributions and platforms.
There are a few options you can give to cabal install
to customize your
build:
Option | Feature |
---|---|
-j |
Build in parallel using multiple processors |
-f llvm |
Use GHC's LLVM backend to compile the code |
-f 64bit |
Enable 64-bit MOO integers |
EtaMOO has non-Haskell dependencies on three external libraries: liblmdb for
database persistence, libpcre (with UTF-8 support enabled) for regular
expression matching, and, possibly, libcrypt (often part of the standard
libraries) for the MOO crypt()
built-in function. You should ensure you have
these available before installing EtaMOO (e.g. on Debian-derived systems,
sudo apt-get install liblmdb-dev libpcre3-dev
).
etamoo
is nearly a drop-in replacement for the LambdaMOO moo
executable;
the main difference is that etamoo
takes a single database path, rather than
both input and output paths. You can run etamoo --help
for a command-line
synopsis.
EtaMOO uses a native binary database format that allows quick loading and
checkpointing, and instantaneous crash recovery. You can create a native
database from a LambdaMOO-format database by using etamoo --import
. You can
also go the other way and convert an EtaMOO database back to a
LambdaMOO-format database with etamoo --export
.
If you don't already have a database, you can find LambdaMOO-format cores for various MOOs online -- for example there is the venerable LambdaCore, or you can request a character on Waterpoint and then perform a live JHCore extraction. (Note that Waterpoint's core extraction process requires running an actual LambdaMOO server executable on the precore database to obtain the final core database; EtaMOO cannot yet do this itself.)
By default, EtaMOO will make use of all available CPUs for maximum
parallelism. If you'd rather limit the number of processors EtaMOO uses, you
can use the command-line option +RTS -N
n
-RTS
where n
is the
number of processors to use.
If you want to enable statistics from the memory_usage()
built-in function,
you will need to add +RTS -T -RTS
to the command line options.
The following LambdaMOO features are currently unsupported:
.program
intrinsic commandverb_cache_stats()
and log_cache_stats()
built-in functionsNP_SINGLE
and NP_LOCAL
network protocols (i.e. stdin/stdout,
UNIX-domain sockets, and/or named pipes; only TCP/IP is supported)OUT_OF_BAND_PREFIX
and OUT_OF_BAND_QUOTE_PREFIX
(these are
currently fixed as #$#
and #$"
, respectively)IGNORE_PROP_PROTECTED
compilation option$server_options.name_lookup_timeout
See also the DIFFERENCES.md
file for other differences between EtaMOO and
LambdaMOO.
Documentation is available for the various types, data structures, and functions used internally by EtaMOO.