=begin = Ruby/DL Ruby/DL provides an interface to the dynamic linker such as dlopen() on UNIX and LoadLibrary() on Windows. = Building and Installing $ ruby extconf.rb # to create the Makefile $ make # to build the library 'dl.so' $ make libtest.so # to build the C library 'libtest.so' for the test script $ make test # to run the test script $ make install # to install the library $ make clean # to remove the created files without Makefile $ make distclean # to remove the all created files = Using Ruby/DL We should usually use DL::Importable module provided by "dl/import.rb". It has high-level functions to access library functions. We use DL::Importable module to extend a module as follows: require "dl/import" module LIBC extend DL::Importable end Now we can use methods dlload and extern in this module. We load the libraries using dlload, and define wrapper methods to library functions using extern respectively as follows: module LIBC extend DL::Importable dlload "libc.so.6","libm.so.6" extern "int strlen(char*)" end # Note that we should not include the module LIBC from some reason. We can call the library function strlen() using LIBC.strlen. If the first character of given function name is an uppercase, the first character of the defined method name becomes lowercase. We can also construct memory images of structures and unions using functions struct and union which are defined in "dl/struct.rb" as follows: require "dl/import" require "dl/struct" module LIBC extend DL::Importable Timeval = struct [ # define timeval structure. "long tv_sec", "long tv_uses", ] end val = LIBC::Timeval.malloc # allocate memory. Notice that the above example takes LIBC::Timeval.malloc to allocate memory, rather than LIBC::Timeval.new. It is because DL::Timeval.new is for wrapping an object, PtrData, which has already been created. We can define a callback using the module function "callback" as follows: module Foo extend DL::Importable def my_comp(str1,str2) str1 <=> str2 end COMPARE = callback "int my_comp(char*,char*)" end where Foo::COMPARE is a Symbol object which invokes the method "my_comp". DL::Importable module is very useful. However, we sometimes encounter a case that we must directly use low-level functions such as dlsym(). In such case, we would use DL module functions. They are described in next section. = DL module Module DL consists of three classes, a few module functions and constants. The class Symbol represents the symbol we can call. The class PtrData indicates a memory block such as a pointer in C. An object instantiated from the class Handle keeps a handle to opened library. == Constants * VERSION * MAJOR_VERSION * MINOR_VERSION * PATCH_VERSION * RTLD_GLOBAL * RTLD_LAZY * RTLD_NOW * MAX_ARG * MAX_CBARG * MAX_CBENT == Functions * handle = dlopen(lib){|handle| ... } * is quite equal to `Handle.new(lib)' * sym = set_callback(cbtype, entry){|args| ... } * sym = set_callback(cbtype, entry, proc) * makes entry-th pre-defined function to call the proc or given block. the entry-th pre-defined function is specified by cbtype and entry. cbtype is a prototype of the callback. see also the section `Type specifiers' about cbtype. * sym = get_callback(cbtype, entry) * returns the Proc object which is given by the above function `set_callback'. * ptr = malloc(size, [free = nil]) * allocates the size bytes, and returns the pointer as a PtrData object ptr. * ptr = strdup(str) * returns a PtrData object ptr which represents the pointer to a new string which is a duplicate of the string str. * size = sizeof(type) * returns the size of type. `sizeof("C") + sizeof("L")' is not equal to `sizeof("CL")'. the latter is assumed to returns the enough size of the structure `struct foo { char c; long l; }', but the size may not equal to `sizeof(foo)' of C. == Handle class * handle = Handle.new(lib){|handle| ... } * opens a library lib and returns a Handle object handle. if a block is given, the handle is automatically closed as the block ends. * Handle#close * closes the handle opened by the above Handle.new(lib). * sym = Handle#sym(func, prototype = "0"), sym = Handle#[func, prototype = nil] * obtains the pointer to a function called func and returns a Symbol object or a DataPtr object. prototype is a string which consists of type specifiers, it indicates the function's prototype. see also the section `Type specifiers'. == Symbol class * sym = Symbol.new(addr, type = nil, name = nil) * creates the Symbol object sym with the type type if type is not nil. addr is the address where the function is allocated. If type is nil, it returns a DataPtr object. * Symbol::char2type(char) * takes a character char that represents a type and returns the type specifier of the C language. * str = Symbol#proto() * returns the function prototype. * str = Symbol#name() * Returns the function name. * str = Symbol#cproto(), str = Symbol#to_s() * returns the prototype of the C language. * str = Symbol#inspect() * returns the inspectable string. * r,rs = Symbol#call(arg1,arg2,...,argN), r,rs = Symbol#[](arg1,arg2,...,argN) * calls the function with parameters arg1, arg2, ..., argN. and the result consists of the return value r and parameters rs. rs is an array. * ptr = Symbol#to_ptr * returns the corresponding PtrData object ptr. == PtrData class * ptr = PtrData.new(addr, [size = 0, free = nil]) * returns the PtrData object representing the pointer which indicates the address addr. GC frees the memory using the free function. * PtrData#free=(sym) * If you specify a symbol object sym, GC frees the memory using the function represented by sym. * sym = PtrData#free * returns a symbol object sym which is used when GC frees the memory. it usually configured by `PtrData#free=' or `PtrData.new'. * size = PtrData#size, PtrData#size=(size) * gets and sets allocated size of the memory. * ary = PtrData#to_a(type, [size]) * returns an array of the type which specified with type. type must be one of 'S','P','I','L','D' and 'F'. * str = PtrData#to_s([len]) * returns a string which length is len. if len is omitted, the end of the string is '\0'. * ptr = PtrData#ptr,+@ * returns the pointed value as a PtrData object ptr. * ptr = PtrData#ref,-@ * returns the reference as a PtrData object ptr. * ptr = PtrData#+ * returns the PtrData object * ptr = PtrData#- * returns the PtrData object * PtrData#struct!(type, *members) * defines the data type to get access to a structure member with a symbol. (see also PtrData#[]) * PtrData#union!(type, *members) * defines the data type to get access to a union member with a symbol. (see also PtrData#[]) * val = PtrData#[key], PtrData#[key, num = 0] * if the key is a string or symbol, this method returns the value of the structure/union member which has the type defined by PtrData# {struct!,union!}. if the key is a integer value and this object represents the pointer ptr, it returns the value of `(ptr + key).to_s(num)' * PtrData#[key,num]=val, PtrData#[key]=val * if the key is a string or symbol, this method substitute the value of the structure/union member with val. if the key is a integer value and val is a string, this method copies num bytes of val to the memory area ptr using memcpy(3). == Type specifiers the prototype consists of the following type specifiers, first element of prototype represents the type of return value, and remaining elements represent the type of each argument. C : char c : char * H : short h : short * I : int i : int * L : long l : long * F : float f : float * D : double d : double * S : const char * s : char * A : const type[] a : type[] (allocates new memory space) P : void * (same as 'p') p : void * (same as 'P') 0 : void function (this must be a first character of the prototype) the cbtype consists of type specifiers 0, C, I, H, L, F, D, S and P. for example: DL.callback('IPP'){|ptr1,ptr2| str1 = ptr1.ptr.to_s str2 = ptr2.ptr.to_s str1 <=> str2 } =end