SML/NJ extends the SML'97 Basis Library with several additional top-level structures:
System info – the SysInfo structure provides information about the runtime system, such as the operating system kind, type and version and whether or not the machine supports multiprocessing.
Weak pointers – the Weak structure provides support for weak pointers.
Compiler internals – the Internals structure provides access to several of the compiler internals, including methods to instantiate and modify the signal table.
Unsafe access – the Unsafe structure provides unsafe access to data structures and runtime-system functions.
Visible compiler structures – SML/NJ also includes structures that provide access to the ML compiler, which contains substructures for execution profiling, control of compiler error-message printing and warnings, and customizable pretty printing.
Additionally, SML/NJ provides some syntactic constructs that are not standard features of SML'97:
Vector expressions and patterns – SML/NJ allows the creation of vectors with the #syntax and allows pattern-matching on them with analogous syntax.
Or-patterns – SML/NJ extends the syntax of SML'97 patterns to allow for a matching multiple patterns in a single rule, provided each pattern has the same type, using => exp.
Object language embedding – SML/NJ provides the quote/antiquote syntax that permits the embedding of expressions in an object language's concrete syntax within ML expressions and programs.
Higher-order modules – SML/NJ supports the parametrization of functors by allowing functors to be components of structures, in addition to Standard ML's parametric modules in the form of functors.
With the 110.81 release, the evolution of SML/NJ is geared towards evolving the Standard ML Basis library and supporting Successor ML features. In the current release, the Successor ML features can be enabled using the command-line option-Cparser.succ-ml=true. The Successor ML documents have been extracted from the SML/NJ '97 TeX files and made available as a GitHub repository of TeX documents which the community is expected to collaborate and grow the language. In addition, HaMLet has a complete implementation of Successor ML features. The practical knowledge so gained, is being incorporated in SML/NJ as it evolves through its releases.
