D is being designed with lessons learned from practical C++ usage rather than from a theoretical perspective. Even though it uses many C/C++ concepts it also discards some, and as such is not strictly backward compatible with C/C++ source code. It adds to the functionality of C++ by also implementing design by contract, unit testing, true modules, automatic memory management (garbage collection), first class arrays, associative arrays, dynamic arrays, array slicing, nested functions, inner classes, limited form of closures, anonymous functions, compile time function execution, lazy evaluation and has a reengineered template syntax. D retains C++'s ability to do low-level coding, and adds to it with support for an integrated inline assembler. C++ multiple inheritance is replaced by Java style single inheritance with interfaces and mixins. D's declaration, statement and expression syntax closely matches that of C++.
The inline assembler typifies the differences between D and application languages like Java and C#. An inline assembler lets programmers enter machine-specific assembly code in with standard D code—a technique often used by system programmers to access the low-level features of the processor needed to run programs that interface directly with the underlying hardware, such as operating systems and device drivers.
D has built-in support for documentation comments, but so far only the compiler supplied by Digital Mars implements a documentation generator.
D supports three main programming paradigms—imperative, object-oriented, and metaprogramming.
Imperative programming is almost identical to C. Functions, data, statements, declarations and expressions work just as C, and the C runtime library can be accessed directly.
OO programming in D is based on a single inheritance hierarchy, with all classes derived from class Object. Multiple inheritance is possible from interfaces (interfaces are a lot like C++ abstract classes).
Metaprogramming is supported by a combination of templates, compile time function execution, tuples, and string mixins.
Memory is usually managed with garbage collection, but specific objects can be finalized immediately when they go out of scope. Explicit memory management is possible using the overloaded operators new and delete, and by simply calling C's malloc and free directly. Garbage collection can be disabled for individual objects, or even for a full program, if more control over memory management is desired. The manual gives many examples of how to implement different highly optimized memory management schemes for when garbage collection is inadequate in a program.
C's application binary interface (ABI) is supported as well as all of C's fundamental and derived types, enabling direct access to existing C code and libraries. C's standard library is part of standard D. Unless you use very explicit namespaces it can be somewhat messy to access, as it is spread throughout the D modules that use it -- but the pure D standard library is usually sufficient unless interfacing with C code.
C++'s ABI is not fully supported, although D can access C++ code that is written to the C ABI, and can access C++ COM (Component Object Model) code. The D parser understands an extern (C++) calling convention for linking to C++ objects, but it is only implemented in the currently experimental D 2.0.
 D 2.0
D 2.0, a branch version of D that includes experimental features, was released on June 17, 2007. Some of these features include support for enforcing const-correctness, limited support for linking with code written in C++, and support for "real" closures.
Even though D is still under development, changes to the language are no longer made regularly since version 1.0 of January 2, 2007. The design is currently virtually frozen, and newer releases focus on resolving existing bugs. Version 1.0 is not completely compatible with older versions of the language and compiler. The official compiler by Walter Bright defines the language itself.
- DMD Compiler: the Digital Mars D compiler, the official D compiler by Walter Bright. The compiler front end is licensed under both the Artistic License and the GNU GPL; sources for the front end are distributed along with the compiler binaries. The compiler back end is proprietary.
- GDC: D 1.0 Compiler, built using the DMD compiler front end and the GCC back end.
 Development tools
D is still lacking support in many IDEs, which is a potential stumbling block for some users. Editors used include Entice Designer, emacs, vi and Smultron among others. A bundle is available for TextMate, and the Code::Blocks IDE includes partial support for the language. However, standard IDE features such as code completion or refactoring are not yet available.
There are at least two actively developed Eclipse plug-ins for D, Descent and Mmrnmhrm.
D applications can be debugged using any C/C++ debugger, like GDB or WinDBG, although support for various fundamental language features is extremely limited then. Debuggers with explicit support for D are Ddbg for Windows and ZeroBUGS for Linux. Ddbg can be used with various IDEs or from the command line, ZeroBUGS has its own GUI.
 Problems and controversies
 Operator overloading
D operator overloads are significantly less powerful than the C++ counterparts. A popular example is the opIndex, which does not allow returning references. This makes assignments like obj[i] = 5; impossible. The D solution is the opIndexAssign operator, which only fixes this very case, but not variations like obj[i] += 5;. In addition, the C++ way of returning a reference allows for the usage of the returned type's overloaded assignment operator. This is currently not possible in D. D 2.0 will fix this by introducing an opIndexLvalue - like operator overload, and deprecating opIndexAssign.
 Division around and lack of functionality in the standard library
D's standard library is called Phobos, and is often perceived as being far too simplistic, in addition to having numerous quirks and other issues. The tango project is an attempt at fixing this by writing an alternative standard library. However, phobos and tango are currently incompatible due to different implementation of the Object class (which leads to GC difficulties). The existence of two de-facto standard libraries could lead to significant problems where some packages use phobos and others use tango.
 Lack of a clear goal
D is often stated as being a "fixed and improved C++". This can lead to featuritis due to the fact that new features are added just because they are perceived as being useful.
 Unfinished support for shared/dynamic libraries
Unix' ELF shared libraries are supported to an extent using the GDC compiler. On Windows systems, DLLs are supported. D's garbage collector allocated objects can be safely passed to C functions, since the garbage collector scans the stack for pointers. However, there are still limitations with DLLs in D including the fact that run-time type information of classes defined in the DLL are incompatible with those defined in the executable, and that any object created from within the DLL must be finalized before the DLL is unloaded.
 Example 1
This example program prints its command line arguments. The main function is the entry point of a D program, and args is an array of strings representing the command line arguments. A string in D is an array of characters, represented by char. Newer versions of the language define string as an alias for char, however, an explicit alias definition is necessary for compatibility with older versions.
import std.; mainstring args i, a; args writefln, i, a; ;
The foreach statement can iterate over any collection, in this case it is producing a sequence of indexes (i) and values (a) from the array args. The index i and the value a have their types inferred from the type of the array args.
 Example 2
This illustrates the use of associative arrays to build much more complex data structures.
import std.; mainstring args string string container; container ~= ; container ~= ; container ~= ; string person, string items; container display_item_countperson, items; ; display_item_countstring person, string items writeflnperson, , items., ;
 Example 3
This heavily annotated example highlights many of the differences from C++, while still retaining some C++ aspects.
#!/usr/bin/dmd -run import std.; mainstring args writefln ; argc, argv; args cl = new CmdLinargc, argv; writeflncl., cl., , cl.; delete cl; specs count, allocated; bigarray = ; specs argspecsstring args in assertargs. > ; outresult assertresult. == CmdLin.; assertresult. > ; body specs* s = new specs; s. = args.; s. = typeofargs.; arg; args s. += arg. * typeofarg.; *s; string argcmsg = ; string allocmsg = ; writeflnargcmsg ~ ~ allocmsg, argspecsargs.,argspecsargs.; ; CmdLin _argc; string _argv; _totalc; : this argc, string argv _argc = argc + ; _argv = argv; _totalc++; ~this argnum _argc; string argv _argv; wstring suffix wstring suffix; _argc : suffix = ; ; : suffix = ; ; : suffix = ; ; default: suffix = ; suffix; typeof_totalc total _totalc; invariant assert_argc > ; assert_totalc >= _argc;
 Example 4
This example demonstrates some of the power of D's compile-time features.
Factorial n n <= Factorial = ; Factorial = n * Factorial!n; factorial n n <= ; n * factorialn; fact_7 = Factorial!; fact_9 = factorial; import std.; pragmamsg, Format!, fact_7; pragmamsg, Format!, fact_9; assertfalse, ;