A new Standard ML dialect written in Java

Morel

Standard ML interpreter, with relational extensions, implemented in Java

(Until version 0.1, Morel was known as smlj.)

Requirements

Java version 8 or higher.

Get Morel

From Maven

Get Morel from Maven Central :

<dependency>
  <groupId>net.hydromatic</groupId>
  <artifactId>smlj</artifactId>
  <version>0.1.0</version>
</dependency>

Download and build

$ git clone git://github.com/julianhyde/morel.git
$ cd morel
$ ./mvnw install

On Windows, the last line is

> mvnw install

Run the shell

$ ./morel
morel version 0.1.0 (java version "11.0.4", JLine terminal, xterm-256color)
= "Hello, world!";
val it = "Hello, world!" : string
= exit
$

Documentation

• Morel language reference
• "How to" guide
• Change log
• Reading test reference logs can be instructive; try, for example, relational.sml.out

Status

Implemented:

• Literals
• Variables
• Comments ( (* block *) and (*) line )
• let (expression that lets you define local variables and functions)
• val (including val rec )
• fun (declare function)
• Operators: = <> < > <= >= ~ + - * / div mod ^ andalso orelse ::
• Built-in constants and functions: it true false nil abs not ignore
• Type derivation
• fn , function values, and function application
• if
• case
• Primitive, list, tuple and record types
• Type variables (polymorphism) (except in let and datatype )
• Enumerated types ( datatype )
• Tuples and unit, record and list values
• Patterns (destructuring) in val and case , matching constants, wildcards, lists, records and tuples
• Basis library functions:
  • Top : abs
  • List : null , length , @ (as prefix at for now), hd , tl , last , getItem , nth , take , drop , rev , concat , revAppend , app , map , mapPartial , find , filter , partition , foldl , foldr , exists , all , tabulate , collate
  • String : maxSize , size , sub , extract , substring , ^ , concat , concatWith , str , implode , explode , map , translate , isPrefix , isSubstring , isSuffix

Not implemented:

• type
• local
• raise , handle
• exception
• while
• References, and operators ! and :=
• Operators: before o
• User-defined operators ( infix , infixr )
• Type annotations in expressions and patterns

Bugs:

• The @ infix operator to concatenate lists is currently named at and is prefix;
• Built-in operators do not use the option type;
• Prevent user from overriding built-in constants and functions: true , false , nil , ref , it , :: ; they should not be reserved
• Access parameters and variables by offset into a fixed-size array; currently we address them by name, in a map that is copied far too often
• Runtime should throw when divide by zero
• Validator should give good user error when it cannot type an expression

Extensions

Morel has a few extensions to Standard ML: postfix labels, implicit labels in record expressions, and relational extensions. Postfix labels and implicit labels are intended to make relational expressions more concise and more similar to SQL but they can be used anywhere in Morel, not just in relational expressions.

Postfix labels

Morel allows '.' for field references. Thus e.deptno is equivalent to #deptno e .

(Postfix labels are implemented as syntactic sugar; both expressions become an application of label #deptno to expression e .)

Because '.' is left-associative, it is a more convenient syntax for chained references. In the standard syntax, e.address.zipcode would be written #zipcode (#address e) .

Implicit labels in record expressions

In standard ML, a record expression is of the form {label1 = exp1, label2 = exp2, ...} ; in Morel, you can omit label = if the expression is an identifier, label application, or field reference.

Thus

{#deptno e, e.name, d}

is short-hand for

{deptno = #deptno e, name = e.name, d = d}

Relational extensions

The from expression (and associated in , where and yield keywords) is a language extension to support relational algebra. It iterates over a list and generates another list.

In a sense, from is syntactic sugar. For example, given emps and depts , relations defined as lists of records as follows

val emps =
  [{id = 100, name = "Fred", deptno = 10},
   {id = 101, name = "Velma", deptno = 20},
   {id = 102, name = "Shaggy", deptno = 30};
   {id = 103, name = "Scooby", deptno = 30}];
val depts =
  [{deptno = 10, name = "Sales"},
   {deptno = 20, name = "Marketing"},
   {deptno = 30, name = "Engineering"},
   {deptno = 40, name = "Support"}];

the expression

from e in emps where e.deptno = 30 yield e.id

is equivalent to standard ML

map (fn e => (#id e)) (filter (fn e => (#deptno e) = 30) emps)

with the where and yield clauses emulating the filter and map higher-order functions without the need for lambdas ( fn ).

Relational expressions are an experiment bringing the features of query languages such as SQL into a functional language. We believe that a little syntactic sugar, backed by a relational query planner, makes ML into a powerful and convenient tool for querying large data sets. Conversely, we want to see how SQL would look if it supported lambdas, function-values, polymorphism, pattern-matching, and removed the syntactic distinction between tables and collection-valued columns.

You can iterate over more than one collection, and therefore generate a join or a cartesian product:

from e in emps, d in depts
  where e.deptno = d.deptno
  yield {e.id, e.deptno, ename = e.name, dname = d.name};

As in any ML expression, you can define functions within a from expression, and those functions can operate on lists. Thus we can implement equivalents of SQL's IN and EXISTS operators:

let
  fun in_ e [] = false
    | in_ e (h :: t) = e = h orelse (in_ e t)
in
  from e in emps
  where in_ e.deptno (from d in depts
                where d.name = "Engineering"
                yield d.deptno)
  yield e.name
end

let
  fun exists [] = false
    | exists hd :: tl = true
in
  from e in emps
  where exists (from d in depts
                where d.deptno = e.deptno
                andalso d.name = "Engineering")
  yield e.name
end

In the second query, note that the sub-query inside the exists is correlated (references the e variable from the enclosing query) and skips the yield clause (because it doesn't matter which columns the sub-query returns, just whether it returns any rows).

More information

• License: Apache License, Version 2.0
• Author: Julian Hyde ( @julianhyde )
• Blog: http://blog.hydromatic.net
• Project page: http://www.hydromatic.net/morel
• API: http://www.hydromatic.net/morel/apidocs
• Source code: https://github.com/julianhyde/morel
• Developers list: dev at calcite.apache.org ( archive , subscribe )
• Issues: https://github.com/julianhyde/morel/issues
• Release notes and history