Chapter 3.  Scheme Language

Table of Contents

Types
Numbers
Characters
Symbols
Strings
Pairs and Lists
Vectors
Boxes
Parameters
Immutable types
Equivalence
Syntax and Lexical Structure
Comments
Shared Structures
Control Features
Syntactic Extension
Errors and Error Handling
Failure Continuations
Error Records
Examples
dynamic-wind
Symbolic Environments and Property Maps
Access Functions
Obtaining and Naming
Chained Symbolic Environments
Miscellaneous Functions

In this chapter we will examine the language that SISC interprets, which is a superset of the R5RS Scheme Standard.

Types

Numbers

The full Scheme number tower is supported:

  • Integers

  • Floating Point numbers

  • Rational numbers

  • Complex numbers

Depending on the numeric library compiled into SISC, floating point numbers have either 32 or 64 bit IEEE precision, or arbitrary [2] precision. Regardless, SISC's complex numbers have floating point components of the same precision as the reals. Integers have arbitrary precision in all numeric libraries, and rational numbers are built with arbitrary precision components.

Numeric constants

The precision specifying exponents (S, F, L, and D) are ignored in SISC, all inexact numbers are kept in the precision of the numeric library. The exponents are read and used to scale the real number as expected. In the case of arbitrary precision floats, specific precision constraints are maintained to prevent a runaway increase of precision. The constraints can be set by minFloatPrecision and maxFloatPrecision configuration parameters on startup. See the section called “Configuration Parameters”.

All four base specifiers (#x, #o, #d, #b) are supported for integers and rationals. Only decimal (#d), the default, is supported for floating point and complex numbers.

SISC will produce infinite or not-a-number quantities from some operations. Those quantities are represented and can be used in Scheme programs as #!+inf (positive infinity), #!-inf (negative infinity), and #!nan (not-a-number).

Exactness

Exactness and inexactness contagion behaves as expected. Rational's are made inexact through division. Floats are made exact by conversion to a rational number. SISC attempts as accurate a conversion as possible, by converting the decimal portion of the number to a ratio with a denominator of the form 10^n, where n is the scale of the floating point number. Then the fraction is reduced as usual.

Since complex numbers must have floating point components currently, conversion to an exact merely rounds the components to integers.

Characters

SISC's characters are double-byte wide. This means that they are capable of representing the full range of unicode characters. Unicode characters can be created with number->character; #\nnnnnn, where nnnnnn is an octal number in the range 000000 -> 177777; or #\uxxxx, where xxxx is a hexadecimal number in the range 0000 -> ffff. At least two zeros must be specified to distinguish from the '0' character when using an octal character literal. At least one zero must be specified to distinguish a hexadecimal character from the 'u' character.

SISC also provides additional named characters, to add to the Scheme standard's space and newline:

Table 3.1. Named character literals

Character NameUnicode Value (hex)
backspace0008
newline000a
nul0000
page000c
return000d
rubout007f
space0020
tab0009

Formally, SISC's lexer modifies the R5RS grammar with the following productions for character literals:

<character> --> #\ <any character> 
     | #\u <uinteger 16> 
     | #\ <uinteger 8>
     | #\ <character name>
<character name> --> backspace | newline | nul 
     | page | return | rubout | space | tab
      

Characters are not compared with respect to the locale of the running system. Character comparison is equivalent to numeric comparison of the character value as returned by char->integer.

There are a number of reasons why a full Unicode system is non-trivial, especially within the framework of the R5RS string and character functions. Such a discussion is outside the scope of this document. Unicode compliant processing may be made available in the future as a library, however.

Symbols

SISC's symbols are ordinarily case-insensitive. SISC maintains true pointer equality between symbols with like contents, unless the symbol is created uninterned. An uninterned symbol is one which is guaranteed to be pointer distinct from any other symbol in the Scheme system, even another with the same contents. Uninterned symbols can be generated with:

procedure: (string->uninterned-symbol string) => symbol

Converts the provided string into an uninterned, pointer distinct symbol.

Uninterned symbols, while always pointer-distinct, may still be equal? to another symbol if its representation matches another.

Case Sensitivity

SISC also allows symbols to be created that are case-sensitive. This can be done one of two ways. The first is by setting the caseSensitive configuration parameter (see the section called “Configuration Parameters”. The second method is via a non-standard symbol syntax. If a symbol is enclosed in pipe ('|') characters, the reader will treat that individual symbol as cased. The syntax extends the R5RS grammar with the following production:

<cased symbol> --> |<identifier>|
        

Example 3.1. Case sensitive Symbol literals

(eq? 'a '|A|) ; => #f
(eq? 'a '|a|) ; => #t
(eq? '|A| '|a|) ; => #f
          

Printed Representation

Symbols may contain characters that are disallowed by R5RS using symbol->string. In such a case, the printed representation of that symbol will contain those characters, prefaced with the escape ('\') character. Likewise, such symbols may be created without symbol->string by escaping non-standard characters.

Symbols which contain characters that could only be present in a case-sensitive environment will be printed in one of two ways, depending on the value of the case-sensitive parameter. If true, the symbols will be printed as is, containing the upper and lower case letters. If false, the symbol will be printed surrounded by pipe characters.

Strings

Strings are built from Unicode characters, and are compared lexicographically in a manner derived from character comparison. In addition to using backslash to escape the double-quote (") character and the backspace character itself, SISC provides several escape codes to ease string literal construction.

Table 3.2. String escape codes

EscapeValue
\fInserts the formfeed character (unicode 000c)
\nInserts the newline character (unicode 000a)
\rInserts the rubout character (unicode 007f)
\tInserts the tab character (unicode 0009)
\uxxxxInserts the unicode character described by the hex number 'xxxx'. All four hex digits must be specified.
\\Inserts the backslash ('\') character
\"Inserts the double quote ('"') character

Pairs and Lists

A function is provided to determine if a given pair is a proper list.

procedure: (proper-list? datum) => #t/#f

Returns #t if the given argument is a proper-list. That is, if the argument is a pair, whose cdr is either the empty-list or also a proper-list, and which contains no references to itself (is not circular).

Vectors

SISC supports the length prefix method of creating Vector constants. For example, '#5(x) creates a vector constant containing five identical symbols. In addition, the length-prefix form is used when printing vectors, and if elements repeat at the end of a Vector, only the last unique element is printed. This form is referred to as the compact vector representation. The unprefixed form with all elements displayed is called the verbose representation.

Vectors are displayed differently depending on the call used. When called with display, in addition to the ordinary R5RS rules regarding the output of values displayed with display, the verbose representation is displayed. Using write, on the other hand produces the compact representation.

Displaying a vector with pretty-print may output either the verbose or compact representation of a vector. The behavior in this regard is controlled by the vectorLengthPrefixing configuration parameter (see the section called “Configuration Parameters”). If set to #t, pretty-print will emit the compact representation. If #f, the verbose representation is produced.

Boxes

SISC supports boxes, a container for a Scheme value. Boxing is often used to implement call-by-reference semantics. Boxes are created and accessed using the following three functions:

procedure: (box value) => box

Creates a box filled with the given value.

procedure: (unbox box) => value

Returns the value contained in the given box.

procedure: (set-box! box value) => undefined

Replaces the value contained in the given box with the value provided.

In addition to the box function for creating boxes, SISC provides an external representation for boxes and boxed values. It extends the R5RS grammar with the following:

<boxed value> --> #&<datum>
      

This syntax denotes a boxed value, with <datum> as the contained value.

Boxes are a distinct first class type. The box? predicate tests a value to see if is a box.

procedure: (box? value) => #t/#f

Returns #t only if the given value is a box.

Boxes, like pairs, are only equal in the sense of eq? and eqv? when a box is compared with itself. A box is equal to another in the sense of equal? if the value contained within the box is equal? to the value contained in the other.

Parameters

A parameter is a named dynamic variable that is accessed through a function. The function, when given no arguments, returns the current value of the parameter. When given an argument, the value of the parameter is set to the provided value.

SISC's parameters are fully compatible with those specified by SRFI-39. Consult the SRFI-39 specification at srfi.schemers.org for documentation on how to construct and use parameters. SRFI-39 does not specify the semantics for parameters in the presence of threads. SISC's parameters bind into the dynamic environment, which means their semantics are defined based on the semantics of dynamic environments' interactions with threads, specified in the section called “ Scheme Thread Semantics ”.

Immutable types

SISC follows the R5RS recommendation of immutable list, string, and vector constants. Quoted lists and vectors are immutable. Attempting to modify elements in these constants will raise an error. String constants are immutable as well when created with symbol->string.

Equivalence

SISC's storage model maintains true pointer equality between symbols, booleans, the end-of-file object, void, and the empty list. Thus two instances of any of those types is guaranteed to return #t from eq? if they would have produced #t from equal?.

Numbers and characters are not pointer equal ordinarily (unless actually occupying the same storage). SISC will return #t from eqv? if two numbers are both exact, or both inexact, and are numerically equal. Two characters are equivalent from eqv? if they occupy the same code-point in the unicode character set. This is the behavior specified by R5RS.

Strings, vectors, lists, and boxes are containers for other Scheme types. As such they are not pointer equal unless they are referenced by two variables that point to the same storage location (i.e. they are actually pointer equal). SISC holds that only equal? will return #t if two objects are the same type and their contents contain equivalent values with respect to equal?.

Syntax and Lexical Structure

Comments

In addition to the single line comments of the Scheme standard, SISC supports both s-expression commenting and nested, multiline comments. An s-expression comment is used to comment out an entire s-expression. To do this, the sharp sequence #; is used. It extends the R5RS grammar with the following production:

<expression-comment> --> #;<datum>
        

The reader, upon encountering this sharp sequence, will read and discard the next datum. The expression commented out must still be a valid s-expression, however.

Nested, multiline comments are as defined in SRFI-30. Briefly, a multiline comment begins with the sharp sequence #| and ends with the sequence |#. The comment may contain nested comments as well. Unfortunately, this extension cannot be represented in a stateless grammar for the lexical structure.

Shared Structures

Reader Syntax

SISC provides a parse-time syntax for creating data (primarily vectors and lists) that contain references to themselves or data which contains several pointer-equal elements. This can be useful to create streams, graphs, and other self-referencing structures while maintaining readability and avoiding complex construction code.

The reader syntax has two parts, defining a pointer, and later referencing the pointer to create the circular reference.

Below is an additional production in the R5RS formal syntax (specifically section 7.1.2, external representations) to support circular structures:

<pointer definition> --> #<uinteger 10>=<datum>
<pointer reference> --> #<uinteger 10>#
        

The first form instructs the reader to create a pointer identified by the specified integer, which maps to the datum that follows, and is active during the reading of the datum on the right-hand side of the definition.

If a second definition occurs during the reading of the datum with the same integral identifier, the previous definition is overwritten for the duration of the read. The definitions are not scoped in any way. The pointer identifiers should be kept unique by the programmer to prevent any unintended effects of identifier collisions.

The second form references a previously created pointer definition. It is an error to reference an undefined pointer. The reader will handle a valid reference by placing a pointer at the current read position back to the location of the definition.

At this point some examples might be helpful:

Example 3.2. Circular Structures

(define x '#0=(1 2 . #0#))
(caddr x)        ; => 1
(list-ref x 15)  ; => 2

(define y '(1 2 #1=#(3 4) . #1#))
(eq? (caddr y) (cdddr y)) ; => #t
          

Writing

Ordinarily, the display of cyclical data would cause a problem for a Read-Eval-Print-Loop. For this reason, the REPL will attempt to check the structure it is about to print for circularities before printing. If a cycle is found in the structure, the REPL will refuse to print if the printShared configuration parameter, described below, is false. In that case the REPL will issue a warning to the user that the structure contains a cycle. If a circular structure is printed with display, write, etc, and the printShared parameter is set to false, the environment may enter an infinite loop which may or may not cause the Scheme system to exit with an error.

The printShared configuration parameter (see the section called “Configuration Parameters”), if set to true enables SISC to scan data for circularity and data sharing before writing values. If such sharing is found, an alternate printer is invoked which will emit a representation compatible with the circular structure representation described in the previous section.

Alternately, SISC also supports SRFI-38, which describes the functions write-showing-shared and read-with-shared-structure.

Control Features

In addition to the R5RS standard control features, two additional forms, when and unless, are supported by SISC.

syntax: (when condition expression [expressions] ...) => value

Evaluates condition, an expression. If true, the expressions that follow are evaluated, in order, the value of the last being returned. If not true, the result is unspecified.

syntax: (unless condition expression [expressions] ...) => value

Evaluates condition, an expression. If false, the expressions that follow are evaluated, in order, the value of the last being returned. If true, the result is unspecified.

Syntactic Extension

SISC provides a hygienic macro system that fully conforms to the R5RS standard. The macro system is provided by the portable syntax-case macro expander. In addition to R5RS macros, the expander provides a more flexible macro definition tool called syntax-case. A full description of the capabilities of the expander is best found in the Chez Scheme Users Guide , specifically Section 9.2, Syntax-Case .

In addition, SISC supports non-hygienic, legacy macro support in two forms; define-macro and defmacro. These forms, found in older Scheme code written for R4RS compliant Scheme systems, should be used only for executing legacy code which relies on it. New code should use the safer and more flexible syntax-case or the standard syntax-rules macros.

syntax: (define-macro (name . args) body ...)

syntax: (define-macro name transformer)

In the first form, define-macro creates a macro transformer bound to name, which when applied will have raw s-expressions bound to one or more parameters (args). The (name . args) name and formal parameter form is identical to the short form for procedure definition with define.

The transformer's body will then, using the s-expressions bound to its arguments, return a new s-expression that is the result of the macro transformation.

The second form binds an arbitrary procedure to the syntactic keyword name, using that procedure to transform occurences of that named syntax during future evaluations.

syntax: (defmacro name args body ...)

defmacro is another macro definition form supported by some Scheme systems. Its semantics are equivalent to:

(define-macro (name . args) body ...)
                

Errors and Error Handling

Errors can be raised by primitives in libraries and Scheme-level code. SISC provides a sophisticated mechanism for handling these errors when they occur during program execution.

Failure Continuations

During the execution of any program, there is always a continuation that represents the rest of a computation. In addition, one can imagine all the activities that will occur as a result of an error. This sequence of actions is explicitly represented in SISC as a failure continuation.

Two values must be applied to a failure continuation. The first is an error record, a datastructure which describes the error (and may contain information about the name of the function that generated the error, a descriptive message about the error, etc.). The second is the continuation of the expression that raised the error. All errors raised in SISC automatically and implicitly obtain and apply these values to the active failure continuation. Applying the error record and error continuation to the failure continuation will not return to the continuation of the application, unless that continuation was captured and later invoked in a non-local entrance.

Creation

A programmer may wish to augment current failure continuation, choosing a different set of actions to occur for a body of code if it raises an error. To facilitate this, SISC provides the with-failure-continuation procedure.

procedure: (with-failure-continuation handler thunk) => value

procedure: (with/fc handler thunk) => value

with-failure-continuation takes as arguments a thunk (a zero-argument procedure) to be evaluated. The thunk will be evaluated in the continuation of the with/fc function, and with a failure continuation defined by the provided error handler. If during the evaluation of the thunk an error is raised, the first, two argument procedure is called with values describing the error and its context. If no error occurs, value of the thunk is applied to the continuation of the with/fc expression.

The error handler required as an argument to with-failure-continuation must accept two values. The first is a value containing information about the error that occurred. This is often an association list containing a number of attributes of the error. The second is a procedure encapsulating the continuation that was in place at the site of the error. This continuation is referred to as the error continuation

When an error occurs, the error handler may choose one of three courses in dealing with the error. First, the handler may choose to return an alternate value to be applied to the continuation of the with/fc expression. Second, the handler may restart the computation from the error site by invoking the error continuation with a value that should be returned in place of the expression that caused the error. Finally, the handler may choose to propagate the error (or a new error) to the failure continuation of the with/fc expression. This can be done with the throw function described in the section called “Raising Errors”.

Capture

The currently active failure continuation may be obtained explicitly using the call-with-failure-continuation procedure. This continuation may be applied to appropriate values at any time in the future.

procedure: (call-with-failure-continuation procedure) => value

procedure: (call/fc procedure) => value

Calls the given one-argument procedure with the currently active failure continuation.

Interaction with Ordinary Continuations

Failure continuations exist as an attribute of the ordinary continuations of Scheme expressions. Because of this, the invocation of a continuation may cause a different failure continuation to become active in the region of the captured continuation. Specifically, the failure continuation in place at the call/cc expression will be reinstated when that continuation is later invoked.

Similarly, invoking a continuation that escapes a region of code will cause any created failure continuations to be abandoned, unless the region is itself captured in a continuation and later invoked.

See also the section called “dynamic-wind.

Error Records

An error record is the value usually propagated with an error in SISC. It is a datastructure containing such information as the location of the error, a descriptive message about the error, and possibly other error metadata.

Creating Error Records

Error records can be created in advance of actually raising an error with the make-error function. The function allows the programmer to create error records that contain a location and a message or value. No field of an error record is required.

procedure: (make-error [location] [message] [arguments] ...) => error-record

Constructs an error record. If present, a symbol, and not #f, the first argument is the location of the error, which may be a symbol equivalent to a function identifier. If present, the message is a format-string processed with the optional arguments that follow as by format in SRFI-28. The remaining arguments must only be present if the format-string is present as the message.

procedure: (make-error [location] error-value) => error-record

Constructs an error record. If present, a symbol, and not #f, the first argument is the location of the error. The second argument is an arbitrary Scheme value that will be the error value. This value will be accessible with the error-message function.

None of the fields of an error-record are required. One may create an error record with no information, an error record with only a location, or an error record with only a message or value. Below are some examples (for an explanation of the throw procedure see the section called “Raising Errors”).

(throw (make-error))
; => Error.

(throw (make-error 'foo))
; => Error in foo.

(throw (make-error "something ~a happened" 'bad))
; => Error: something bad happened

(throw (make-error 3))
; => Error: 3

(throw (make-error #f 'foo))
; => Error: foo

(throw (make-error 'foo "something ~a happened" 'bad))
; => Error in foo: something bad happened
          

In addition, an error record may be created that adds additional information to an error record that was already created. This is useful when an error was caught in an error handler, and one wishes to raise an error from the handler that contains additional information about the local location or error message as well as the error that was caught.

procedure: (make-nested-error local-error parent-error parent-error-continuation) => error-record

procedure: (make-nested-error local-error exception) => error-record

The first version creates an error record which has parent-error (and its associated parent-error-continuation) as the root cause of an error-record passed as local-error.

The second version creates an error record which has exception (see the section called “Exceptions”) as the root cause of an error passed as local-error.

An example of the creating, throwing, and display of a nested error follows.

(with-failure-continuation
  (lambda (m e)
    (throw (make-nested-error 
            (make-error 'foo "could not call bar.") m e)))
  (lambda ()
    (error 'bar "something went wrong.")))
;=> Error in foo: could not call bar.
;   Caused by Error in bar: something went wrong.
          

Accessors

An error record contains several useful pieces of information. The following functions allow the programmer to access that information.

procedure: (error-location error-record) => symbol

Obtains the location of the error, a symbol which may be a function identifier. If there is no location specified, #f is returned.

procedure: (error-message error-record) => value

Obtains the message of the error, which may be a string which is a descriptive message of the error, or an arbitrary value (as created by the second form of make-error). If there is no message specified, #f is returned.

procedure: (error-parent-error error-record) => error-record

Obtains the parent error of the error. This is the value of the second argument to the make-nested-error function. If there is no parent specified, #f is returned.

procedure: (error-parent-continuation error-record) => error-continuation

Obtains the parent error continuation of the error. This is the value of the third argument to the make-nested-error function. If there is no parent specified, #f is returned.

Raising Errors

The fundamental mechanism for raising an error in application code is provided by the throw procedure.

procedure: (throw error-record [error-continuation]) => does not return

procedure: (throw exception) => does not return

The first verison applies the given error record to the current failure continuation. If provided, the error continuation is designated by the optional parameter. If not, the continuation of the throw expression is used.

The second form applies the current failure continuation to the error record and error continuation extracted from the supplied exception (see the section called “Exceptions”).

If invoked from an error-handler with the values of the handler's formal parameters, throw has the effect of propagating the error in a manner that is equivalent to the absence of the modified failure-continuation.

throw could be defined in terms of call-with-failure-continuation as:

(define (throw error . args)
  (call-with-failure-continuation
     (lambda (fk)
       (if (null? args)
           (call-with-current-continuation (lambda (k) (fk error k)))
           (fk error (car args))))))
          

For convenience and compatibility with SRFI-23, the function error is provided. Its syntax is identical to make-error, but it immediately applies the resulting error record to the current failure continuation with the current continuation as the error continuation.

procedure: (error [location] [message] [arguments] ...) => does not return

Raises an error record whose location, if provided, is location, a symbol; and whose error message, if present, is message. If provided, the error message is a format-string that is processed, with the optional arguments, as with the format function in SRFI 28.

procedure: (error [location] error-value) => does not return

Raises an error record whose location, if present, is the symbol location, and and whose error-value is any arbitrary Scheme value.

error can be implemented in terms of throw and make-error:

(define (error . args)
  (throw (apply make-error args)))
          

Exceptions

Exceptions in SISC are a simple wrapper around an error record and an associated error continuation.

Exceptions are created with

procedure: (make-exception error-record error-continuation) => exception

Constructs an exception from an error-record and an error-continuation, e.g. as obtained from the arguments of a handler procedure passed to with-fc.

Accessors and a type-test are provided by the following procedures:

procedure: (exception-error exception) => error-record

Returns the exception's error record.

procedure: (exception-continuation exception) => error-continuation

Returns the exception's error continuation.

procedure: (exception? value) => #t/#f

Returns #t if value is an exception object, #f otherwise.

Examples

At this point, a few examples may be helpful:

(+ 1 (/ 1 0) 3)
; => A divide by zero error is raised
	

Example 3.3. Return a new value

(with-failure-continuation
    (lambda (error-record error-k)
      'error)
  (lambda () (+ 1 (/ 1 0) 3)))
; => The symbol 'error
	

Example 3.4. Restart with a different value

(with-failure-continuation
    (lambda (error-record error-k)
      (error-k 2))
  (lambda () (+ 1 (/ 1 0) 3)))
; => 6
	

Example 3.5. Propagate the error

(with-failure-continuation
    (lambda (error-record error-k)
      (throw error-record error-k))
  (lambda () (+ 1 (/ 1 0) 3)))
; => A divide by zero error is raised
	

Example 3.6. Propagate a different error with the same error continuation

(with-failure-continuation
    (lambda (error-record error-k)
      (throw (make-error '/ "could not perform the division.") error-k))
  (lambda () (+ 1 (/ 1 0) 3)))
; => An error is raised: Error in /: could not perform the division.
	

Example 3.7. Raise a new error

(with-failure-continuation
    (lambda (error-record error-k)
      (error 'example-function "could not evaluate the expression."))
  (lambda () (+ 1 (/ 1 0) 3)))
; => An error is raised: Error in example-function: could not evaluate the expression.
	

Note that the difference between Example 3.6, “Propagate a different error with the same error continuation” and Example 3.7, “Raise a new error” is that in the former, the computation can still be restarted from the second argument of the addition if an outside handler catches the newly raised exception and applies the continuation. This is not true in the last example, as its a new error whose continuation is the same as the with-failure-continuation expression.

dynamic-wind

R5RS does not specify the behavior of dynamic-wind in the case where an error is raised while evaluating the during thunk. SISC chooses to view an error raised in that section as an instance of the dynamic extent being exited. In other words, if an error is raised in the dynamic extent of a dynamic-wind expression, SISC will ensure that the after thunk is evaluated before the error is propagated to the failure-continuation of the dynamic-wind expression.

Example 3.8. Errors and dynamic-wind

(define x 0)
(dynamic-wind (lambda () (set! x (+ x 1)))
              (lambda () (/ 1 0))
              (lambda () (set! x (+ x 1))))

; => A divide by zero error is raised, and the value of x is 2
        

If an error is raised in either the before or after thunks, no additional measures are taken. The error is propagated to the failure-continuation of the dynamic-wind as if the dynamic-wind call was an ordinary function application. Explicitly, if an error is raised from before, neither during nor after will be executed. If an error is raised in after, the results of evaluating before and during remain valid.

Also noteworthy is what happens if a continuation is invoked that exits from either the before or after thunks. Such a case is treated just as if a continuation was invoked during the evaluation of an operand to an application. This is to say that no additional steps will be taken by SISC. If before is escaped by a continuation invocation, neither during nor after will be executed. If after is escaped, the results of before and during remain valid.

In summary, extraordinary evaluation is only possible during the evaluation of the during thunk. The before and after thunks are evaluated with the dynamic environment and dynamic-wind stack of the call to dynamic-wind itself.

Symbolic Environments and Property Maps

Symbolic environments and property maps provide additional named global environments useful for storing program specific data without exposing it to the general purpose top-level environment.

A property map is dictionary structure tied to the interaction environment which maps symbolic names to Scheme values. First-class symbolic environments provide a similar mapping, but can be used as first class values (including as an argument to eval). Symbolic environments are used to implement SISC's global (top level)and report environments.

Access Functions

Access to symbolic environments is performed through the getprop and putprop functions. All symbolic environment operations are thread safe.

procedure: (getprop binding-name plist-name [default-value]) => value

procedure: (getprop binding-name environment [default-value]) => value

Attempts a lookup of binding-name in an environment.

In the first form, the the binding is resolved in the interaction-environment's property list named plist-name, a symbol. If the environment is not found or the binding doesn't exist, default-value is returned if provided, otherwise #f is returned.

In the second form, the binding is resolved in a first-class symbolic environment.

procedure: (putprop binding-name plist-name value) => undefined

procedure: (putprop binding-name environment value) => undefined

Sets the value of a binding named with the symbol binding-name in a property list or first class symbolic environment.

In the first form, the binding is resolved using a symbolic name (plist-name) in the interaction environment's property lists. If the map does not yet exist, it is created as an empty map.

In the second form, the binding is resolved in the provided first class symbolic environment. If the binding does not yet exist in the given environment, it is created. If a binding previously existed, its previous value is discarded.

Obtaining and Naming

Symbolic environments are a first class datatype in SISC. The top-level environment itself is merely a special cased symbolic environment. To obtain the top-level environment as a first class value, one can use the interaction-environment function that is an optional procedure in R5RS. Another useful environment is the R5RS report environment available by calling:

(scheme-report-environment 5)
        

Each call to scheme-report-environment returns a new environment that contains only the bindings available in the Scheme report. Finally, the initial environment available to the programmer when SISC starts can be retrieved using the sisc-initial-environment function:

procedure: (sisc-initial-environment) => environment

Returns the initial SISC interaction environment.

Like scheme-report-environment, each call to sisc-initial-environment returns a distinct environment which contains only the bindings initially available when SISC starts. An interesting use of this would be to define one or more distinct initial-environments, bound to toplevel variables. One could then define Scheme code and data in each environment that can use the full SISC language but cannot see any bindings in other environments.

Finally, R5RS states that it is an error to modify the contents of a top-level variable that has not yet been created. SISC adheres to the standard, and raises an error when any unbound variable in a symbolic environment (including the top-level) is modified using set!. This differs from some Scheme systems that will silently create the binding and set it to the new value.

Chained Symbolic Environments

SISC contains a mechanism for creating a symbolic environment which is chained to another environment, such that new and modified bindings are created in the new, child environment, but bindings may also be resolved from the parent if not present in the child. SISC uses this functionality to protect the contents of the R5RS and SISC initial environments from modification. One can use it in a similar way, protecting the bindings in the parent for sandboxing or other purposes.

procedure: (make-child-environment parent-environment) => environment

Creates a new environment, initially empty of its own bindings, but which chains to the provided parent-environment when resolving a binding.

procedure: (parent-environment environment) => environment

Obtains the parent environment of a symbolic environment. If the given environment has no parent (e.g. is not chained), #f is returned.

Miscellaneous Functions

The remaining functions in this chapter are not easily classified, but nevertheless are useful and worth describing.

procedure: (circular? datum) => #t/#f

Returns #t if the given datum is circular. A datum is circular if it is a compound datum (lists or vectors for example), and one of its elements is a reference to itself, or a reference to a sub-element which creates a cycle.

procedure: (compose [function] ...) => procedure

compose takes zero or more functions of one argument and returns a new function of one argument that will apply to that argument to the selected functions in reverse order. If no functions are provided, the identity function is returned.

For example, the function caddr could be simply defined as:

(define caddr (compose car cdr cdr))
            

procedure: (iota n) => pair

The iota function produces a list whose elements are the integers 0 to n-1 inclusive.

syntax: (time [iterations] expression) => list

Evaluates the given expression iterations times, or if iterations is not provided, only once. When complete, a list is returned of the following form:

(result (n ms))
              

where result is the Scheme value that resulted from the last evaluation of the expression, and n is the number of milliseconds taken to evaluate the expression. If more than one iteration occurred, then the average number of milliseconds elapsed during each iteration is returned.



[2] Essentially arbitrary, see the section called “Limits” for a discussion of the physical limits of number representation