Julia vs Standard ML

Julia runs almost as fast as (and in fact in some cases faster than) C code.

Dynamic and high level, but does not isolate the user from properly thinking about types. Can do explicit type signatures which is great for teaching structured thinking.

Out of the box Julia has a very good Read-Eval-Print-Loop, which both completes functions and types, as well as completion based on history of previous statements. It integrates well with the shell and has an excellent online help system.

Julia code is easy to read and avoid a lot of unnecessary special symbols and fluff. It uses newline to end statements and "end" to end blocks so there is no need for lots of semicolons and curly braces. It is regular in that unless it is a variable assignment, function name always comes first. No need to be confused about whether something is a method on an object or a free function.

Unlike Python and Ruby, since you can annotate the types a function operates on, you can overload function names, so that you can use the same function name for many data types. So you can keep simple descriptive function names and not have to invent artificial function names to separate them from the type they operate on.

The Julia language is written in itself to a much larger extent than most other languages, so a budding programmer can read through the depths of the standard library and learn exactly how things work all the way down to the low-level bit-twiddling details, which can be englightening.

Julia has built in n-dimensional arrays similar in functionality as Python's numpy.

You can have multiple functions with the same name, but doing different things depending on function arguments and argument types.

Julia requires no boilerplate code – a beginner needs to write only the parts they care about. This combined with the REPL provides the best learning experience available.

Julia provides a high level of abstraction, making platform-independent programming trivial and easing the learning curve.

You can perform operations on scalars, for example 2^2 or [1, 2, 3].^2.

Julia allows you to edit Julia code in the language itself and write powerful macros. It is a great introduction to metaprogramming features

The Julia REPL allows quickly testing how some code behaves and gives access to documentation and package management immediately in the REPL.

The module system that Standard ML uses gives the programmer the power to define custom data types whose internal implementation is invisible to other programmers using the module.

Since mutability is only confined to a special type of reference cells, implementing laziness in SML can be done in only 20 lines of code.

Since it uses strict evaluation, it enforces distinction between data and computations which in turn enables you to use induction on algebraic data types as a reasoning principle.

Secret messages can be sent across distant parts of a program without possibility of being intercepted by unintended recipients in the middle.