SkySpark by SkyFoundry

14. Axon Language


This chapter fully covers the the syntax constructs of the Axon programming language. See Axon chapter for an overview of how Axon is used in the SkySpark.


The following scalar literals are supported:

Numbers may be optionally annotated with a unit - see Units.

Note that bins are represented specially as a dictionary literal with the bin and mime tags.

There is no literal representation for DateTime or Coord, instead use an Axon function to construct:


Lists store a linear sequence of objects zero indexed by a Number.

[]           // empty list
[6]          // single item list
[1, 2, 3]    // multiple items
[4, "four"]  // mixed item types

Also see Axon Usage.


Dicts are name/value maps (associative arrays). Any string key may be used used if quoted, but by convention most maps use valid identifiers as their keys.

{}                // empty dictionary
{foo:6}           // foo bound to 6
{foo:2+3}         // foo found to 5 (value can be any expr)
{foo}             // foo is bound to special "marker" value
{-foo}            // foo is bound to special "remove" value
{n:"Bob",age:35}  // mutiple value separated with comma
{"not tag"}       // use string literal for non-identifier keys

Also see Axon Usage.


Operators in order of precedence:

Math Operators

Mathematical operator rules for operands:

=== Addition ===
num + num            >> num
date + num           >> date (num must be number of days)
dateTime + num       >> dateTime (num must have duration unit)
dateRange + num      >> dateRange (num must have duration unit)
uri + str            >> uri concat
uri + uri            >> uri concat
str + obj            >> str concat
obj + str            >> str concat

=== Subtraction ===
num - num            >> num
date - num           >> date (duration must be number of days)
date - date          >> num (difference in number of days)
dateTime - num       >> dateTime (num must have duration unit)
dateTime - dateTime  >> num (difference in number of hours)
dateRange - num      >> dateRange (num must have duration unit)

=== Multiplication ===
num * num            >> num

=== Division ===
num / num            >> num

Any math operation with null always results in null:

null + 2        >> null  // null always carries into result

Any math operation with na results in na. However, null still has precedence over na:

na() + 2        >> na()  // na() always carries into result
na() + 2 + null >> null  // null trumps na()

See Units for how numbers with units are handled by comparison and arithmetic operators.

Boolean Operators

The not, and, and or operators work with boolean values according to standard truth tables for boolean logic. Axon also supports boolean operators with na. Boolean operations on na will always yield na, but note that short-circuiting rules still apply.

=== not ===
not true          >> false
not false         >> true
not na()          >> na()

=== and ===
true  and false   >> false
true  and true    >> true
true  and na()    >> na()
false and na()    >> false    // short-circuit

=== or ===
true  or false    >> true
false or false    >> false
false or na()     >> na()
true  or na()     >> true     // short-circuit

Even though the boolean operators support na, it is important to understand that na is not a boolean type. Therefore, conditional expressions that require boolean values, such as if, will fail if the expression evaluates to na.

x: na()

// throws CastErr because na() is not boolean
if (not x) "TRUE" else "FALSE"

// works because explicitly checking for true
if (true == (not x)) "TRUE" else "FALSE"  >> "FALSE"

Get and Trap Operators

The indexing operator target[key] is a shortcut for the get function. It can be used to index strings, lists, dicts, and grids:

str[index]         // get unicode char from string
str[start..end]    // slice/substring
list[index]        // get an item at zero based index
list[start..end]   // perform list slice
dict[key]          // lookup tag value by key
grid[num]          // get row at index
grid[start..end]   // slice of grid rows

Like Fantom, negative indices may be used to access from the end of the string.

The trap operator -> is a shortcut for the trap function. When used on a Dict, it looks up a tag value by name. You can also use the -> operator on a Ref which will automatically resolve the id using readById:

id->foo            >>>  readById(id)->foo
id->siteRef->tag   >>>  readById(readById(id)->siteRef)->foo

Both [] and -> are used to lookup a tag value by name in a dict. The difference is what happens when the key is not defined. The [] operator (or get function) returns null when the key is not defined. The -> operator (or trap function) raises an UnknownNameErr exception. If you expect the tag to exist, use the -> operator to get a meaningful exception as opposed to generic NullErr. If you expect the tag might not exist, then use [], but make sure to check for null.

Also note that null values or sparse grid cells are treated as if the tag does not exist. For example this code will raise an exception:



The colon operator is used to define a variable as name/value binding:

a: 5
b: 3
a + b  // yields 8

You can re-assign the value of a variable using the = operator:

a: 4
a = a + 1

The variable must already be defined via the : operator before it can be assigned, otherwise a runtime exception is raised.


Axon uses lexical scoping with closure support. Each function call creates one scope. Note that do blocks, if blocks, or try blocks do not create new scopes. For example consider this code:

if (true) do
  a: "foo"
return a

In this case "foo" will be returned by a. Once a variable is defined, it is visible anywhere in the function. This model is more akin to JavaScript or Ruby, then static languages like Java or C#.

Each nested function within a top-level function has visibility to its lexically scoped outer function's variables. This design is known as closures. Here is a simple example:

num: 0
f: ()=> num = num+10
return num

In this example the function f passed to times is accessing the num variable from the outer scope. The result will be 30. However, a nested function can define a new variable which hides the outer scope:

x: "outer"
f: () => do
  x: "inner"
  return x
return [f(), x]

In the example above the result will be "[inner, outer]". The variable x inside f hides the outer scope which remains unchanged.

Variable resolution is searched in the following order:

  1. current function's variables
  2. lexically scoped outer function's variables
  3. top-level namespace


Lambdas are function definitions which take zero or more parameters and calculate a resulting value:

x => x * x         // parens optional if single param
(x) => x * x       // but you can use them
() => "some val"   // no params
(x, y) => x + y    // multiple params

You'll often use the def operator to create named functions:

add: (x, y) => x + y
add(2, 3)  >>  5

You can define a default value for parameters using the colon:

f: (a, b:2) => a + b


You call a function with the () operator.

The number of parameters a function expects is called its arity. You must pass enough arguments to a function to satisify its arity. If function parameters have defaults, then those arguments may be omitted. It is also permissible to pass more arguments to a function - the additional arguments are ignored:

f: (a, b:2, c:3) => a + b + c
f()               >> error!
f(10)             >> 15
f(10, 11)         >> 24
f(10, 11, 13)     >> 34
f(10, 11, 13, 14) >> 34

If you leave off the () operator, then the expression evaluates to the function itself. This is useful for passing functions around to perform higher order functional programing.

Dot Calls

It is common to pipe the results of one function to another function to build up pipelines much like the Unix | operator. Consider the following expression:


In the expression above, we evaluate the function a, then pass the results to b, which in turn gets passed to c. We can write the above using the . dot call operator:

a().b.c        // may omit parens if no arguments

Here are some more examples:

year(today())                   >> today().year
toStr(year(today()))            >> today().year.toStr
hisRead(readAll(kw), 2009)      >> readAll(kw).hisRead(2009)

Trailing Lambda

Many functions are designed to take other functions as arguments. These functions are called higher order functions. Consider a simple example:

// default is to sort by alphabetic order
list: ["cape", "ape", "batch"]
list.sort  >>  ["ape", "batch", "cape"]

// we can pass a function to sort by string length
list.sort((x, y) => x.size <=> y.size)  >>  ["ape", "cape", batch"]

This pattern is so common that code can start to become confusing with the nesting parenthesis (much like Lisp code). Axon borrows a pattern from Fantom and Ruby which allows you to pull the last argument outside of the parenthesis if it is a lambda expression:

list.sort() (x, y) => x.size <=> y.size  >>  ["ape", "cape", batch"]

Note that when using dot calls you must include empty () parens before specifying the lambda argument. The exception to this rule is if the lamdba takes a single argument with no parens:

list: [3, 4, 1] x => -x      // ok x => -x    // ok (x) => -x  // ok (x) => -x    // not ok

More examples:

eachDay(2010-07, day => echo(day))  // inside parens
eachDay(2010-07) day => echo(day)   // trailing outside parens

Partial Application

When calling a function you can use the _ symbol to perform partial application. Partially applied functions evaluate to another function with one or more parameters bound. For example:

add: (a, b) => a + b
inc: add(_, 1)
inc(3) >> 4

In the example above the expression add(_, 1) created a new function with one parameter which was essentially a shortcut for this:

inc: (x) => add(x, 1)


Anywhere an expression is expected you can declare a block which is series of expressions evaluated in order. Blocks are declared with the do and end keywords. Blocks are most often used with lambdas:

f: (a, b) => do
  c: a + b

f(3, 4)  >>  49  which is (3+4) * (3+4)

The entire block evaluates to the last expression, or you can use the return keyword to short circuit and immediately return a value:

f: (x) => do
  if (x < 10) return "small"
  if (x > 90) return "big"

f(3)   >>  "small"
f(93)  >>  "big"
f(33)  >>  "medium"

Note that return returns from the inner-most function, not necessarily the top level function.

If Expr

The if expression is used to evaluate a block only if a given condition is true. You can also use an else clause to evaluate a block if the condition is false.

Like everything in Axon, if is itself an expression which evaluates to the result of true clause or false clause. If the condition is false and there is no else clause then the if expression evaluates to null:

if (5.isOdd) "odd"              >> "odd"
if (4.isOdd) "odd"              >> null
if (4.isOdd) "odd" else "even"  >> "even"

You can use a do block if you need to evaluate multiple expressions inside the if expression:

if (cond1) do
end else if (cond2) do
end else do

As a convenience you can omit the end keyword if it immediately preceeds the else keyword:

if (cond1) do
else if (cond2) do
else do

Throw Expr

The throw expression is used to raise an exception. Exceptions in Axon are dicts which have the err marker tag and the dis display message tag. To raise an exception use the throw keyword followed by an expression which evaluates to a dict:

throw {dis:"deep doo-doo!", ts:now()}

Note your dictionary can define any tags you might want to use in exception handling. As a convenience you can also use a string:

throw {dis:"deep doo-doo!"}  // as dict
throw "deep doo-doo!"        // convenience for above

The err tag is automatically added for you.

Try/Catch Expr

The try/catch expression is used to trap exceptions. Lets look at a simple example:


You can declare a variable to store the exception's dict as follows:

catch (ex)

The body of the try and catch clauses can be any expression or a do block.

try do
end catch (ex) do

As a convenience you can omit the end keyword if it immediately preceed the catch keyword:

try do
catch (ex) do

Like everything in Axon, try/catch is an expression which evaluates to a result. If no exception is thrown, then the whole try/catch expression evaluates to the try body clause, otherwise the expression evaluates to the catch block:

result: try getCurVal() catch "bad val"

In the expression above if no exception is thrown by getCurVal, then its result is assigned to the variable result. If an exception is thrown, then the string "bad val" is assigned to result.