6 Erlang Libraries
The full library descriptions are included with the open-source distribution and also on line. The highlights are:
SASL, EVA, INETS and SNMP provide Operations- and maintenance-related functionality: release handling, alarm and event handling, administration via web browser and SNMP.
ORBER, CosEvent, CosTransactions and IC contain support for everything CORBA related.
Mnesia and Mnemosyne provide a real-time distributed database which can be both in-RAM and on-disk. There is also a standard ODBC client driver.
There is an ASN.1 compiler which produces Erlang code.
A rich collection of modules provides everything from string, list and regular expression manipulation to random number generation and calendar lookups.
gen_server, gen_fsm, and supervision trees provide standard ways of implementing client/server subsystems, state machines and supervised fault-tolerant systems.
GUI and Graphics
wx, gs and webtool provide ways to build a gui.
Among other things, there are also HTTP, SSH and FTP servers.
Linked lists are a fundamental part of Erlang, as are tuples. The other standard data structures are:
|sets||sets, i.e. a collection of unique elements.|
|gb_sets||sets, but based on a general balanced data structure|
|gb_tree||a general balanced tree|
|dict||maps, also called associative arrays|
|ets||hash tables and ordered sets (trees)|
|dets||on-disk hash tables|
The contributions area at the erlang.org site includes modules for double-ended queues and balanced trees.
ordset and orddict are not really suitable for large amounts of data: they are no better that lists for many common operations, e.g. inserting an element is O(n) for both. dict and sets are usually preferred.
In practice, Erlang programs use lists (either natively or via dict) for data structures involving up to a few hundred elements and use ETS (the Erlang Term Store) or mnesia for anything larger. ETS uses hashing to allow near constant-time access to almost arbitrarily large amounts of data.
For a collection of data consisting of a few (tens or hundreds) items, lists often outperform both ETS and trees. For large numbers of small items, ETS tends to work best. For larger items, balanced trees can outperform ETS because they avoid copying the data.
Johan Bevemyr wrote one which works for unix machines. The code is freely available at the contributions area as serial-1.0.
There have been discussions on the mailing list about writing a more general driver which also works on windows machines. Nothing much has come of this (lack of interest?).
There are several.
There's also an Erlang binding to GTK and a binding to GD which is useful for dynamically creating JPEG and PNG images). There's a popular OpenGL modeller and some interesting work to directly access the X protocol.
This takes more work and patience than most people have. Kent Boortz described many of the hurdles on the mailing list.
There are other ways to spread your code, including distributing it from your own website or adding it to the jungerl.