Smalltalk vs Haxe

You can modify the system as it's running. You're "swimming with the fish", instead of probing a black box by remote control.

Programmers must make detailed mental models of the system they are developing. Bugs usually happen when the mental model does not match the actual system. This is one of the greatest difficulties beginners have because most systems are so opaque. It takes a lot of effort to see what's really going on. But in Smalltalk this is much easier, thanks to the powerful tools included in the environment, like the object inspector.

All tools (Inspector, Browser, Debugger etc.) are written in Smalltalk and are live objects in the environment. All sources are present, so that the tools can easily be studied, changed and experimented with.
The same goes for the other components like the compiler, OS-Integration etc.

Smalltalk is one of few languages that are purely object oriented. This provides a solid and easy to understand base on which to learn object oriented programming, the most popular approach to programming.

The syntax was designed to be easy enough for children to learn. Beginners can learn the language rules very quickly and then focus on programming without fighting the syntax at the same time. Things that have to be baked into the grammar in other languages are simple message sends with block arguments in Smalltalk. Expressions have only three precedence levels to worry about.

Smalltalk had the original (and still the best) unit test system that inspired it in many other languages (like Java's JUnit). Working with interactive live objects in Smalltalk style TDD makes it easy to teach and learn TDD.

You can edit code and swap it in while the program is still running after an exception has already been signaled, or restart from anywhere in the call stack. You can inspect and modify the state of any object. Some Smalltalkers write unit tests and then program exclusively in the debugger.

You can explore how everything works easily.

Smalltalk provides an extremely fast code-compile-run-debug cycle. You don't have to stop and reset the world to tweak your program, since you can compile one method at a time while the environment is still running. This is great for beginners to experiment and prototype ideas.

MIT licensed implementations Pharo, Squeak, Cuis & Dolphin
and GPL licensed GnuSmalltalk.

A Smalltalk environment can save the state of a running program and later restore and resume execution. This includes the internal state of live objects, multiple thread stacks, and debugging sessions, making it easier for beginners to take the exact problem to an expert for assistance.

This leads to a very simple programming model (pure OO) that is still very powerful. A lot of stuff that is hard to implement in other languages is easier in Smalltalk.

Beginners are usually stuck making command-line applications in other languages, because GUIs are too hard. Smalltalk GUIs are easy enough to start with.

Also known as "blocks". These objects contain reusable snippets of code and as first-class objects they can be passed as arguments to other methods or blocks and also returned from them. "lexical closures" mean they retain access to the variables in the lexical environment they were written in, that is, in the surrounding code.

Smalltalk is the inventor of Just-in-Time compilation and the MVC concept, refactoring through their so-called refactoring browser and it was also one of the first adopters of a language virtual machine, closures, live programming, test driven development, an IDE and the development of GUI`s.

Hybrid languages (e.g., Java, C#, C++) make it easy to slip into procedural thinking. Smalltalk's pure OO approach makes it hard not to think in object-oriented terms. In addition, since the entire IDE and runtime components are there in the image for you to browse, you have plenty of examples of good OO design to learn from.

Many languages today use object orientation, while the most of them stock on the half way in that perspective.
Smalltalk sees literally everything as an object and this includes things like the classes and primitive data types. There is are zero control structures such as selection and iteration, since all is done by sending messages to objects.
It use a lot of concepts from Lisp in order to provide a nice experience for this pure kind of object orientation.
It provides immutable data structures, closures, anonymous functions and higher order functions, while all those functions are objects. This is what makes Smalltalk so simple, elegant, and easy.

All this counts for Pharo, while other implementations as Amber are probably feature complete to it.

Haxe allows you to develop for Web, iOS, Android, Blackberry, Windows, OSX, Linux and others, all at once, without the need to switch languages and maintain separate code bases.

This is possible because Haxe can compile to JavaScript, ActionScript, Flash AVM bytecode, C++, Neko, PHP, C# (.NET) and Java.

Support for even more platforms and languages is under development.

The language was designed to be very expressive with the smallest possible amount of syntactic sugar. There are actually fewer keywords than other languages with similar power.

Haxe can easily compile over 100,000 lines of code to JS in seconds on a mid-spec computer.

The language is very easy to learn for those with background in JavaScript or ActionScript 3.

Haxelib (common library repo) and other sources contain large codebases for anything from cryptography to communications. A lot of these are fully cross platform and work with the JavaScript target.

The JavaScript target can be used for everything from node.js server applications (with code completion) to games using either the Flash-like OpenFL library or direct canvas or WebGL programming.

Haxe has been around for more than 10 years (since 2005) and whilst not the most popular project, has had continuous growth.

Highly unlikely to disappear or for support to stop.

Friendly community

One big advantage over pure javascript, (or some other languages listed here) is that Haxe will pick up a whole range of errors when you compile, saving you the pain of having to try and debug them later. This includes everything from syntax errors ("Unexpected ;") to type errors ("Class user has no field username. Suggestion: username").

Code completion is built into the compiler and available to the IDE allowing for much smarter code completion that can actually parse and understand the syntax tree.

The output that is generated can be trimmed using "dead code elimination" to only include those functions and libraries that are strictly necessary. All code is very readable with only minimal extras for specific functionality.

Small output is good for frontend development as file size is a major concern.

After a type is inferred from its context, it cannot be changed to a new type, and type safety is done at compile time so it stays safe without the extra maintenance required for static typing.

Syntactic macros allow you to extend compiler features at the syntax tree step. Macros come into play after code is parsed into the abstract syntax tree, and macros allow you to transform it before the rest of the compilation completes.

This provides for immense power, while at the same time scoping the extensibility at a level that is powerful, but well constrained.

You can use the same classes on the server as you do on the client where applicable. This saves a lot of time.

Haxe has the ability to use "externs". These are haxe files which describe the usage of existing JS libraries. Get code completion and compile-time-checking for everything from jQuery to Node.js.

Even without externs, native JS code can still be used through untyped code.

Haxe has the power and expression to not need the npm dependancy hell that is common in js and typescript, bit it's still simple.

Using a remoting proxy you can get type safe server to client communications, allowing for remote method execution on the server as if they were part of the client side code.

Package tree is just directory tree, it's wonderful!

Haxe supports conditional compilation, so depending on compiler flags Haxe will include or exlcude sections of your code. Making it easy to have debug and release builds.

You can define constants in an abstract enum and when used in a switch/case statement Haxe checks for exhaustiveness, making sure every constant is covered - with no runtime implication.

Haxe compiler will check types when using externs for existing libraries.

You should use externs when calling non Haxe code, but if you just need to call one or two external JS functions, you can skip type checking by calling untyped code.