3. Bigloo
A practical Scheme compiler
User manual for version 4.2a
September 2015 -- Modules
A modules is a compiler and interpreter entity. Modules have been first designed for the compiler that compiles modules and then, links them against libraries in order to produce executables. A module may be split into several files but a file cannot contain more than one module. A module is made of a module clause that is a list for which the car is the symbol module and followed by any Bigloo expression (that is definitions or expressions). The module clause names the module and defines the scope of the definitions. At last, the module clause is also the place where foreign bindings are defined and where classes are defined. Recent versions of Bigloo (since 2.7b) fully supports modules from the interpreter.

3.1 Program Structure

A Bigloo program is composed of one or more Bigloo modules where a module is defined by the following grammar:

<module>             ==> <module-declaration> <module-body>
<module-declaration> ==> the module declaration
<module-body>        ==> the module body
A module is not related to a specific file and can be spread over several files if that is convenient. In particular, there is no relationship between module names and file names. The module declaration (see Module Declaration) must be the first expression in the first of the files containing the module; other expressions form the body of the module. The module body (see Core Language) contains global variables, function definitions and top level expressions (see Expressions).


3.2 Module declaration

The module declaration form is

module name clause ...bigloo syntax
This form defines a module and must be the first in the file. The argument name is a symbol naming the module. If the same module name is used more than once, Bigloo signals an error. The runtime library is composed of modules that are read when a user module is compiled and hence, if a user module has the same name as one of the library modules, an error is signaled.

A simple module can be:

(module foo)

(display "this is a module")
The first line here is the complete module definition, the last line is the complete module body and together they form a complete Bigloo program. If these lines were stored in file zz.scm, invoking bigloo zz.scm would create the executable a.out which, when obeyed, would display this is a module on the terminal.

Note: Some special identifiers are reserved and can't be used to name modules. If such an identifier is used, the compiler will produce the message:

#(module t
#^
# *** ERROR:bigloo:TOP-LEVEL:Parse error
# Illegal module name -- (MODULE eval ...
The list of reserved identifiers may be enlarged for next release. For the current release that list is made of: eval, foreign and t.

Module clauses can be:

main namebigloo module clause
This clause defines the entry point for a stand alone application to be procedure name of arity one. Bigloo invokes this procedure at the beginning of execution providing the list, composed of the shell command line arguments, as its single argument.

(module foo
   (main start))

(define (start argv) (display argv) (newline))
Then if this program is compiled into foo and invoked using the command foo -t bar, the list which is the argument for the main procedure start would be ("foo" "-t" "bar").

The special form args-parse helps main function argument parsing (see Command Line Parsing).

include file-name ...bigloo module clause
This is a list of file-names to be included in the source file. Include files are not modules and may have a special syntax. Thus, besides containing Bigloo expressions, they can contain import and include clauses, which must be written in a single list whose first element is the keyword directives. Includes files can be used to include implementation-neutral Scheme expressions and definitions in a Bigloo module. Here is an example of an include file.

;; `foo.sch'
(define-struct point x y)
and the module that includes the `foo.sch' file:

;; `foo.scm'
(module foo
   (include "foo.sch"))

(print (point 1 2))
Include files, may contain module information. This is the role of the include directives clause here illustrated with the `bar.sch' example:

;; `bar.sch'
;; the directives
(directives (include "foobar.sch")
            (import  hux))

;; expressions (define (gee x) (print x))

import import ...bigloo module clause
An import is a list of the form:

<import>      ==> <iclause> ...
<iclause>     ==> (<bind-name> ... <bind-name> <module-name> <file-name> ...)
                | (<bind-name> ... <bind-name> <module-name>)
                | <module-name>
                | (<module-name> <file-name> ...)
<bind-name>   ==> <r5rs-ident>
                | <alias-name>
<alias-name>  ==> (<r5rs-ident> <r5rs-ident>)
<module-name> ==> <r5rs-ident>
<file-name>   ==> <string>
The first alternative in iclause imports the variable named bind-name which is defined in the module module-name, located in the files file-name .... The second does the same but without specifying the name of the file where the module is located. The third and the fourth form import all the exported variables of the module module-name.

Note: The need for specifying in which files modules are located comes from the fact that there is no automatic mapping between module names and files names. Such a mapping can be defined in a ``module access file'' (see Module Access File) or in the import clause itself, as in the first and fourth alternatives in iclause above.

Here is an example of an import clause:

(module foo
   (import 
      ;; import all bar exported bindings:
      bar
      ;; import the hux binding exported by
      ;; the module hux:
      (hux hux)       
      ;; import the fun1, fun2 and fun3 bindings exported by
      ;; the module mod:
      (fun1 fun2 fun3 mod)       
      ;; import the fun4 bindings that will be known in this module
      ;; under the alias name f
      ((f fun4) mod)
      ;; import all gee bindings. the gee module
      ;; is located in a file called `gee.scm':
      (gee "gee.scm")))

use use ...bigloo module clause
use has the same meaning as import except that modules which are used are not initialized (see Module Initialization). Used modules are read before imported modules.

with with ...bigloo module clause
This clause specifies a list of modules which have to be initialized at runtime and is used to force the initialization of modules which are never imported but which are required by an application (see Embedded Bigloo applications).

export export ...bigloo module clause
In order to make a module's global bindings available to other modules, they have to be exported. Export clauses are in charge of this task and an export is a list of the form:

<export>  ==> <eclause> ...
<eclause> ==> <ident>
            | (inline <ident> <ident> ...)
            | (generic <ident> <ident> <ident> ...)
            | (<ident> <ident> ...)
            | <class>
            | (macro <ident> <ident> ...)
            | (expander <ident>)
            | (syntax <ident>)
The first form of eclause allows the variable ident be exported, the second allows the function ident, always regarded as immutable when exported this way, to be exported and the third exports an inline-procedure (see Inline Procedures) whose name is extracted from the first ident after the word inline. The last two are both connected with Bigloo's object system. The generic clause exports generic functions (see Generic functions) and class clause exports classes (see Class declaration).

Note: Only bindings defined in module m can be exported by m (i.e. bindings imported by m cannot be exported by m).

Type information, specified in any ident in an export clause, is used by Bigloo. Where no type information is given, a default generic type named obj is used.

Note: The last formal argument of a multiple arity function can not be typed because this argument is bound to be a pair or null. This union cannot be denoted by any type.

Here is an example of the module foo that exports bindings:

(module foo
   (export 
      ;; export the bar mutable variable
      bar
      ;; export the hux function. this
      ;; function takes exactly two arguments
      (hux x y)       
      ;; export the inline function gee
      ;; that takes at least one argument.
      (inline gee x . z)))

static static ...bigloo module clause
A static clause has exactly the same syntax as an export clause. However, bindings declared static are local to the module. Since the default scope of all bindings is static, static module clauses are useful only for program documentation.


from from ...bigloo module clause
from clauses have the syntax of import clauses. The allow the re-exportation of imported bindings. That is, any module can export any bindings imported via a from clause.

As an example, suppose we have module bar:

(module bar
   (export (fun)))

(define (fun) "bar")
Now, suppose we have a module foo that imports bar, by the means of a from clause. Module foo is able to re-export the bar binding of module bar:

(module foo
   (from (fun bar "bar.scm")))
A third module, let's name it gee, importing module foo, can see the binding for function bar:

(module gee
   (import (foo "foo.scm")))

(print (fun))
This feature is very useful when compiling modules exporting functions with type annotations. In particular, one may write:

(module foo
  (export (class c1 x)))
Then,
(module bar
  (import foo)
  (from foo)
  (export (fun::c1)))

(define (fun) (instantiate::c1 (x 10)))
And,

(module gee
   (import bar)
   (main main))

(define (main x) (let ((o (fun))) (print o) (print (c1? o))))

load load ...bigloo module clause
A load is a list of the form:

<load>    ==> <lclause> ...
<lclause> ==> (<module-name> <file-name>)
            | <module-name>
This clause forces Bigloo to load the module specified in the lclause in the environment used by the macro expansion mechanism. This means that the user's macros can use all the bindings of all the loaded modules but the loaded bindings remains unknown to the compiler.

If the module foo is defined by:

(module foo
   (export (foo x)))

(define (foo x) `(cons ,x ,x))
then,

(module gee
   (load (foo "foo.scm")))

(define-macro (gee x) `(cons ,(-fx x 1) ,(foo x)))

(gee 5) ==> (cons 4 (cons 5 5)) => (4 5 . 5)

eval eval...bigloo module clause
This form allows interactions between compiled code and interpreted code. (See the Section Eval command line options for a presentation of compilation flags that enable compilation tuning for eval.) Each eval has the following syntax:

<eval> ==> (export-all)
         | (export-module)
         | (export-exports)
         | (export <bind-name>)
         | (export (@ <bind-name> <module-name>))
         | (import <bind-name>)
         | (class <bind-name>)
         | (library lib1 ...)
The first clause, (export-all), exports all the variables bound in the module (i.e., the variables defined in the module and the imported variables). The second clause, (export-module), exports all the module's variables (those declared static and exported) to the interpreter; the third exports all the exports (i.e. the ones present inside an export clause) variables to the interpreter; the fourth and fifth clause each export one variable to the interpreter. The last clause imports a variable from the interpreter and all such imported variables are immutable (i.e. they cannot be the first argument of a set! expression with the compiled code). Variables that are exported to the evaluators must be exported. If a variable is exported to the evaluators but not exported within an export clause, the compiler will produce an error message. The library clause makes the variables and functions of a library accessible from the interpreter.

(module foo
   (export (fib x))
   (eval (export fib)
         (import bar)))

(define (fib x) ...) (print bar)
The clause (class <bind-name>) exports a class definition to the interpreter. This makes the class constructor, the class predicate and the slots access functions available from the interpreter. The form (instantiate::class ...) and (with-access::class ...) are also available from the interpreter.

extern extern ...bigloo module clause
Extern (aka foreign) clauses will be explained in the foreign interface (see C Interface).

java java ...bigloo module clause
Java clauses will be explained in the Java interface (see Java Interface).

option option ...bigloo module clause
This clause enables variables which affect compilation to be set from inside a module and since the expressions, option ..., are evaluated when compiling, no code is compiled for them. They are allowed to make side effects and to change the values of the global variables which describe how the compiler must compile. Usually they allow the control variables, which are described when Bigloo is invoked with the -help2 option, to be set as in the following example:

(module examplar
   (option (set! *debug* 3)
           (set! *verbose* 2)))

(print 'dummy)
Whatever arguments are passed on the command line, Bigloo will compile this module in both verbose mode and debug mode.

library library ...bigloo module clause
This clause enables libraries (see Bigloo Libraries) when compiling and linking Bigloo modules. The expressions library ... are symbols naming the libraries to be used.

Here is an example of a module declaration which makes use of a library named format:

(module test
   (library format)
   (main    test-format)
   (import  (test2 "test2.scm")))
Using a library does not automatically binds its variables and functions to the interpreter. In order to make these available to the interpreter an explicit use of an eval library clause must be used.

type type ...bigloo module clause
This forms is used to define builtin Bigloo types. It is not recommended to use it in user programs. So, it is left undocumented.




3.3 Module initialization

Initializing a module means evaluating, at runtime, its top level forms (global bindings are top level forms).

When a module, module1, imports a module, module2, module2 is initialized before module1. Modules are initialized only once, nothing being done if a module already met during initialization is met again. Library modules are initialized before user modules and imported modules are initialized in the same order as they appear in import clauses.

Here is a first example with two modules. First the module foo:

;; module foo
(module foo
   (main main)
   (import (bar "bar.scm")))

(define (main argv) (print "argv: " argv)) (print "foo")
Then the module bar
;; module bar
(module bar)

(print "bar")
These can be compiled into the executable a.out with:
$ bigloo -c foo.scm
$ bigloo -c bar.scm
$ bigloo foo.o bar.o
Execution of a.out produces:
$ a.out
   -| bar
      foo
      argv: (a.out)
The explanation is:
  • module foo contains the program entry point so this is where initialization begins.
  • because foo imports module bar, bar must be initialized before foo. This explains why the word bar is printed before anything else.
  • module initialization for foo is completed before main is called. This explains why word foo is printed before main is entered.
Let's consider another example with 3 modules:
;; module1
(module module1
   (main main)
   (import (module2 "module2.scm")))

(define (main argv) (print "argv: " argv))

(print "module1")
The second module:
;; module2
(module module2
   (import (module3 "module3.scm")))

(print "module2")
The third module:
;; module3
(module module3
   (import (module1 "module1.scm")))
 
(print "module3")
Compile with:
$ bigloo module1.scm -c
$ bigloo module2.scm -c
$ bigloo module3.scm -c
$ bigloo module1.o module2.o module3.o
Execution produces:

$ a.out
   -| module3
      module2
      module1
      argv: (a.out)
The order of module initialization can be explicitly specified using with and use clauses.

3.4 Qualified notation

Global variables can be referenced using implicit notation or using qualified notation. Implicit notation is used when variables are referenced just by their name whereas qualified notation is used when variables are referenced by their name and the name of the module which defines them. Qualified notation has the following syntax:

(@ <bind-name> <module-name>)
and is useful when several imported modules export a variable with the same name. Using qualified notations instead of short notation only affects compilation.

When several variables are defined under the same identifier, the compiler uses the two following rules in order to decide which variable is selected by an implicit reference: 1) the variable defined in a module has a higher precedence than all imported variables, 2) imported variables have a higher precedence than library variables.


3.5 Inline procedures

Bigloo allows procedures called inline and which differ from normal ones only in the type of code planted. An inline procedure is a first class object which can be manipulated in the same way as any other procedure but when Bigloo sees a reference to one, rather than generating a C function call to the function, the body of the inline procedure is open-coded. The definition of an inline is given in the following way:

define-inline (name args ...) bodybigloo syntax
define-inline (name args ... . arg) bodybigloo syntax
Apart from the initial word, this form has the same syntax as that used by define for procedures. Inline procedures are exportable which means that the compiler scans imported files to find the bodies of all inline procedures. Here is a small example of a module which exports an inline and a module which imports it.

;; the exporter module
(module exporter
        (export (inline make-list . objs)))

(define-inline (make-list . objs) objs)
;; the importer module
(module importer
        (import exporter))

(print (make-list 1 2 3 4 5))
Because of the open-coding of the exporter procedure, the above print statement is equivalent to:

(print (let ((objs (list 1 2 3 4 5)))
          objs))
Any procedure can be an inline. Also any exported procedure can be an inline provided all global variables and functions it uses are also exported.

Note: Bigloo can decide to inline procedures declared with define but this can be achieved only with local procedures whereas procedures declared with the define-inline form are open-coded even through module importation.

Note: Procedures declared inline are macro expanded with the macro defined in the module where they are invoked. That is, if module module1 declares an inline procedure p and module module2 imports it, p may have two different macro-expansions: one for module1 and one for module2.


3.6 Module access file

Bigloo is different from languages such as C where a module is defined by a file. For Bigloo, the module name is not necessarily the name of the file where the text of the module is written and modules can even be split across several files.

Since modules are defined independently of files, it is necessary to make a link between a module and its files and there are two ways of doing this. Choosing an import clause where the file-names are specified or creating a ``module access file''. Such a file must contain only one list, each element of the list being of the form:

(module-name "file-name" ... "file-name")

Use the -afile <file> option to specify the ``module accessfile'' when compiling. By default Bigloo checks if a file named .afile exists. If it exists it is loaded as a module access file.

See The Bigloo command line.

Note: The Bigloo distribution contains a tool, bglafile, that can automatically build a ``module access file''. See the manpage for bglafile for details.


3.7 Reading path

Imported, included or loaded files are sought first in the current directory and then in the directories, sequentially from start to end, of the list in the *load-path* variable. This variable, initially set to the empty list, can be reset by the -I option of the compiler.


This Html page has been produced by Skribe.
Last update Thu Sep 3 08:07:38 2015.