Paradigm | Multi-paradigm: object-oriented, scripting |
---|---|
Designed by | Limin Fu |
First appeared | 2006 |
Stable release | dao-1.2
/ 2012-07-08 |
Typing discipline | static, dynamic, inferred |
OS | Cross-platform |
License | BSD |
Website | http://www.daovm.net |
Influenced by | |
C++, Lua, Python |
Dao is an object-oriented scripting language with dynamically typed variables supporting complex data structures. It has text processing abilities, such as regular expression matching. It provides many built-in numerical datatypes such as complex numbers and multi-dimensional numeric arrays, as well as their corresponding arithmetic operations. Support for multi-threaded programming is also an integrated part of Dao. The Dao interpreter is implemented as a lightweight virtual register machine (Dao VM) in standard C. The Dao VM can be extended with C or C++.
Advanced features of the language include:
- A mixed static and dynamic type system with automatic type inference capabilities
- A flexible macro system that allows definition of new syntax
- Concurrent and distributed programming with message Passing Interfaces
Basics
editHello World
editThe classic hello world program can be written as follows:
io.write( "Hello world!" )
Here io(=stdio) is the standard library to handle input and output.
The std.listmeth routine can be used to display the methods in a library. The methods in math, for example, can be displayed using:
std.listmeth( math )
Datatypes
editA Dao variable can be implicitly declared by assigning the result of an expression to a variable name.
a_number = 123
a_string = "abc"
a_list = { 1, 2, 3 }
a_map = { "CD"=>123, "AB"=>456 }
a_tuple1 = ( 123, "ABC" )
a_tuple2 = ( index => 123, name => "ABC" ) # tuple with named items
a_vector = [ 1, 2, 3 ]
a_matrix = [ 1, 2; 3, 4 ]
By default, a variable will have a fixed type that is inferred from the expression that is used to declare the variable; this is called implicit typing. The type of a variable can also be specified explicitly, using the following Pascal-style declaration syntax:
var_typed : type
var_typed : type = value
type can be one of the following built-in types: int, float, double, string, complex, list, map, tuple, array, buffer, routine, or some composition of these types.
For example:
a_list2 : list<list<float> >
a_list3 : list<list<string> > = {}
a_map2 : map<string,int> = { "ABC"=>123 }
a_tuple1 : tuple<int,string> = ( 123, "ABC" )
a_tuple2 : tuple<index:int,name:string> = ( index => 123, name => "ABC" )
type can also be a Dao class name or the name of a user-defined C type. Special keywords for typing include: any for any type, ? for undefined types and @X for a type holder that can be initialized to a certain type in a parameter list.
All typed variables undergo static checking.
Control flow
editIf-Elif-Else:
a = math.rand();
if( a > 0.75 ){
a -= 0.75
}elif( a > 0.5 ){
a -= 0.5
}else{
a -= 0.25
}
While:
i = 0;
while( i < 10 ) i++
For:
for( i=0; i<10; i++ ) io.writeln( i )
for( i=0 : 9 ) io.writeln( i )
for( i=0 : 2 : 9 ) io.writeln( i )
For-in:
a_list = { 1, 2, 3 }
for( it in a_list ) io.writeln( it )
a_list2 = { "a", "b", "c" }
for( it in a_list; it2 in a_list2 ) io.writeln( it, it2 )
For-in also works for hash.
Switch
a = ""
switch( a ){
case 1 : a = "1";
case 2, 3, 4: a = "2,3,4"; # multiple values as case
case 5 ... 9 : a = "5...9"; # value range as case
case "abc" : a = "abc"; # string as case
default : a = "default";
}
Functions
editThe declaration of a basic function (known in Dao as routines) looks like this:
routine RepeatString( s : string, times = 1 ) => string
{
rep = '';
for( i = 1 : times ) rep += s;
return rep;
}
Dao supports first-class functions. The definition of such functions is identical to that of normal function definition, with the following differences:
- there is no need for a function name, but the created function must be assigned to a variable;
- the default value expressions do not need to be constant expressions, as they are evaluated at run time when the function is created;
- the function body may contain variables defined in the "upper" function that creates it.
a = "ABC";
rout = routine( x, y : string, z = a+a ){
a += "_abc";
io.writeln( "lambda ", a )
io.writeln( "lambda ", y )
io.writeln( "lambda ", z )
}
rout( 1, "XXX" );
Classes
edit class MyNumber
{
routine MyNumber( value = 0 ){ Value = value }
routine setValue( value : int ){ Value = value }
routine getValue(){ return Value }
operator .Value=( value : int ){ Value = value } # field set operator
operator .Value(){ return Value } # field get operator
private
var Value = 0; # default value is zero
}
Macros
editDao has a unique macro system where macro is defined in a way similar to writing Backus–Naur Form (BNF) metasyntax notation. Such macro is composed of two parts, the first part is the source syntax pattern to be matched, and the second part is the syntax to be applied. The transformation from the source syntax pattern to the target syntax pattern is controlled by some predefined markups (for identifiers, expressions, blocks and their repetitions etc.).
Dao macros are defined in the following way:
syntax{
source_syntax_pattern
}as{
target_syntax_pattern
}
This example enables "while-do-end" without brackets:
syntax{
while $EXP do \[ $BL \] end
}as{
while( $EXP ){ \[ $BL \] }
}
Functional methods
editBesides the supporting of first-class function, Dao has additional support of functional style programming by providing a set of higher-order functions as built-in methods, such as map(), fold() and unfold() etc.
The basic syntax for such methods is the following,
method( parameter(s) )->|variable(s)|{ inlined_function }
Examples,
a = { 1, 2, 3 }
b = map( a ) -> { 10*x } # produce { 10, 20, 30 }
b = map( a ) -> |x| { 10*x } # equivalent to above
# map() can take more than one lists as parameters:
b = { 11, 22, 33 }
c = map( a, b ) -> |x,y| { x + y }
# function composition
c = map( a, b )->|x,y|{ x + y, x - y }->|u,v| { u * v }
Concurrent and distributed programming
editAsynchronous function call
editProbably the simplest way to create multi-threaded programs in Dao is to use asynchronous function calls (AFC). The way of using AFC is almost identical to that of normal function calls, with the exception that the keyword async must follow the call.
myfunc( myparams ) async;
myobj.mymeth( myparams ) async;
Any functions or methods can be invoked in such an asynchronous way.
Normally AFC is executed in a separated native thread, which can be either an idle thread available from the thread pool, or a new thread created on the fly.
Message Passing Interface
editWith Message Passing Interface APIs provided in library mpi, one can easily do concurrent and distributed programming in Dao. In the MPI library, there are three principal functions: spawn(), send() and receive(). With spawn(), one can create lightweight virtual machine processes or real operation system processes, in the local or remote computers; and with send() and receive(), a process can send messages to or receive messages from other processes.
The process names are of the following form:
virtual_process@real_process@@hostname
"@real_process@@hostname" identifies an operating system process on host "@@hostname", and "virtual_process@real_process@@hostname" identifies a virtual process within "@real_process@@hostname". Message passing can only happen among virtual processes. Each real process has a main virtual process named "self". Any part can be omitted. For example, when "virtual_process" is omitted, it means the "self" virtual process, and "@@hostname" alone identifies the "self" within the operating system process that binds to port 4115 ( D : 4, A : 1, O : 15 ).
# spawn a real process to run script "mpi_script.dao"
mpi.spawn( "@pid", "mpi_script.dao" );
# spawn a virtual process to run function "test"
mpi.spawn( "vmp@pid", "test" );
# send message to a real process
mpi.send( "@pid", "TO MAIN" );
# print a received message
io.writeln( mpi.receive() );
# spawn a real process on "@@localhost" to run script "mpi_script.dao"
mpi.spawn( "@pid2@@localhost", "mpi_script.dao" );
# spawn a virtual process on "@pid2@@localhost" to run function "test"
mpi.spawn( "vmp@pid2@@localhost", "test" );
# send message to the virtual process "vmp@pid2@@localhost"
mpi.send( "vmp@pid2@@localhost", "ANOTHER", 123.456 );
External links
edit
Category:Scripting languages
Category:Virtual machines
Category:Free compilers and interpreters
Category:Free development toolkits and libraries