# lm/gnu-smalltalk forked from bonzini/smalltalk

Switch branches/tags
Nothing to show
Fetching contributors…
Cannot retrieve contributors at this time
4961 lines (4088 sloc) 195 KB
 \input texinfo.tex @c -*- texinfo -*- @c %**start of header (This is for running Texinfo on a region.) @setfilename gst.info @settitle GNU Smalltalk User's Guide @setchapternewpage odd @c %**end of header (This is for running Texinfo on a region.) @c ******************************************* Values and macros ********* @include vers-gst.texi @ifclear UPDATE-MONTH @set UPDATE-MONTH @value{UPDATED} @end ifclear @macro bulletize{a} @item \a\ @end macro @ifinfo @set SMILE ;-) @end ifinfo @ifnotinfo @set SMILE @end ifnotinfo @c Preferred layout than @uref's @macro hlink{url, link} \link\@footnote{\link\, \url\} @end macro @macro mailto{mail} \mail\ @end macro @ifhtml @unmacro hlink @unmacro mailto @macro hlink{url, link} @uref{\url\, \link\} @end macro @macro mailto{mail} @uref{mailto:\mail\, , \mail\} @end macro @macro url{url} @uref{\url\} @end macro @end ifhtml @macro gst{} @sc{gnu} Smalltalk @end macro @macro gnu{} @sc{gnu} @end macro @dircategory Software development @direntry * Smalltalk: (gst). The @gst{} user's guide. @end direntry @copying @quotation Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and with no Back-Cover Texts. A copy of the license is included in the section entitled GNU Free Documentation License''. @end quotation @end copying @titlepage @title @sc{gnu} Smalltalk User's Guide @subtitle Version @value{VERSION} @subtitle @value{UPDATE-MONTH} @author by Steven B. Byrne, Paolo Bonzini, Andy Valencia. @comment The following two commands start the copyright page. @page @vskip 0pt plus 1filll @insertcopying @end titlepage @node Top, , , (DIR) @top @ifnottex This document describes installing and operating the @gst{} programming environment. @insertcopying @end ifnottex @menu * Overview:: What @gst{} is. * Using GNU Smalltalk:: Running @gst{}. * Features:: A description of @gst{}'s special features. * Packages:: An easy way to install Smalltalk code into an image. * Emacs:: @gst{} and Emacs. * C and Smalltalk:: @gst{}'s C/Smalltalk interoperability features. * Tutorial:: An introduction to Smalltalk and OOP. @detailmenu --- The detailed node listing --- Using GNU Smalltalk: * Invocation:: What you can specify on the command line. * Operation:: A step-by-step description of the startup process and a short description of how to interact with @gst{}. * Syntax:: A description of the input file syntax * Test suite:: How to run the test suite system. * Legal concerns:: Licensing of GNU Smalltalk Operation: * Command-line processing:: Picking an image path and a kernel path. * Loading or creating an image:: Loading an image or creating a new one. * Starting the system:: After the image is created or restored. Legal concerns: * GPL:: Complying with the GNU GPL. * LGPL:: Complying with the GNU LGPL. Features: * Extended streams:: Extensions to streams, and generators * Regular expressions:: String matching extensions * Namespaces:: Avoiding clashes between class names. * Disk file-IO:: Methods for reading and writing disk files. * Object dumping:: Methods that read and write objects in binary format. * Dynamic loading:: Picking external libraries and modules at run-time. * Documentation:: Automatic documentation generation. * Memory access:: The direct memory accessing classes and methods, plus broadcasts from the virtual machine. * GC:: The @gst{} memory manager. * Security:: Sandboxing and access control. * Special objects:: Methods to assign particular properties to objects. Packages * GTK and VisualGST: GUI. * Parser, STInST, Compiler: Smalltalk-in-Smalltalk. * DBI: Database. * I18N: Locales. * Seaside: Seaside. * Swazoo: Swazoo. * SUnit: SUnit. * Sockets, WebServer, NetClients: Network support. * XML, XPath, XSL: XML. * Other packages: Other packages. Emacs * Editing:: Autoindent and more for @gst{}. * Interactor:: Smalltalk interactor mode. C and Smalltalk: * External modules:: Linking your libraries to the virtual machine * C callout:: Calls from Smalltalk to C * C data types:: Manipulating C data from Smalltalk * Smalltalk types:: Manipulating Smalltalk data from C * Smalltalk callin:: Calls from C to Smalltalk * Object representation:: Manipulating your own Smalltalk objects * Incubator:: Protecting newly created objects from garbage collections * Other C functions:: Handling and creating OOPs * Using Smalltalk:: The Smalltalk environment as an extension library Tutorial: * Getting started:: Starting to explore @gst{} * Some classes:: Using some of the Smalltalk classes * The hierarchy:: The Smalltalk class hierarchy * Creating classes:: Creating a new class of objects * Creating subclasses:: Adding subclasses to another class * Code blocks (I):: Control structures in Smalltalk * Code blocks (II):: Guess what? More control structures * Debugging:: Things go bad in Smalltalk too! * More subclassing:: Coexisting in the class hierarchy * Streams:: Something really powerful * Exception handling:: More sophisticated error handling * Behind the scenes:: Some nice stuff from the Smalltalk innards * And now:: Some final words * The syntax:: For the most die-hard computer scientists @end detailmenu @end menu @node Overview @unnumbered Introduction @gst{} is an implementation that closely follows the Smalltalk-80 language as described in the book @cite{Smalltalk-80: the Language and its Implementation} by Adele Goldberg and David Robson, which will hereinafter be referred to as @cite{the Blue Book}. The Smalltalk programming language is an object oriented programming language. This means, for one thing, that when programming you are thinking of not only the data that an object contains, but also of the operations available on that object. The object's data representation capabilities and the operations available on the object are inseparable''; the set of things that you can do with an object is defined precisely by the set of operations, which Smalltalk calls @dfn{methods}, that are available for that object: each object belongs to a @dfn{class} (a datatype and the set of functions that operate on it) or, better, it is an @dfn{instance} of that class. You cannot even examine the contents of an object from the outside---to an outsider, the object is a black box that has some state and some operations available, but that's all you know: when you want to perform an operation on an object, you can only send it a @dfn{message}, and the object picks up the method that corresponds to that message. In the Smalltalk language, everything is an object. This includes not only numbers and all data structures, but even classes, methods, pieces of code within a method (@dfn{blocks} or @dfn{closures}), stack frames (@dfn{contexts}), etc. Even @code{if} and @code{while} structures are implemented as methods sent to particular objects. Unlike other Smalltalks (including Smalltalk-80), @gst{} emphasizes Smalltalk's rapid prototyping features rather than the graphical and easy-to-use nature of the programming environment (did you know that the first GUIs ever ran under Smalltalk?). The availability of a large body of system classes, once you master them, makes it pretty easy to write complex programs which are usually a task for the so called @dfn{scripting languages}. Therefore, even though we have a @sc{gui} environment based on GTK (@pxref{GUI, , GTK and VisualGST}), the goal of the @gst{} project is currently to produce a complete system to be used to write your scripts in a clear, aesthetically pleasing, and philosophically appealing programming language. An example of what can be obtained with Smalltalk in this novel way can be found in @ref{Top, , Class reference, gst-libs, the @gst{} Library Reference}. That part of the manual is entirely generated by a Smalltalk program, starting from the source code for the class libraries distributed together with the system. @node Using GNU Smalltalk @chapter Using @gst{} @menu * Invocation:: What you can specify on the command line. * Operation:: A step-by-step description of the startup process and a short description of how to interact with @gst{}. * Syntax:: A description of the input file syntax * Test suite:: How to run the test suite system. * Legal concerns:: Licensing of GNU Smalltalk @end menu @node Invocation @section Command line arguments The @gst{} virtual machine may be invoked via the following command: @example gst [ flags @dots{} ] [ file @dots{} ] @end example When you invoke @gst{}, it will ensure that the binary image file (called @file{gst.im}) is up to date; if not, it will build a new one as described in @ref{Loading or creating an image,, Loading an image or creating a new one}. Your first invocation should look something like this: @display "Global garbage collection... done" @gst{} ready st> @end display If you specify one or more @var{file}s, they will be read and executed in order, and Smalltalk will exit when end of file is reached. If you don't specify @var{file}, @gst{} reads standard input, issuing a @samp{st>} prompt if the standard input is a terminal. You may specify @option{-} for the name of a file to invoke an explicit read from standard input. @cindex saving @cindex quitting @cindex exiting @findex quit @findex snapshot To exit while at the @samp{st>} prompt, use @kbd{Ctrl-d}, or type @kbd{ObjectMemory quit} followed by @key{RET}. Use @kbd{ObjectMemory snapshot} first to save a new image that you can reload later, if you wish. As is standard for @acronym{GNU}-style options, specifying @option{--} stops the interpretation of options so that every argument that follows is considered a file name even if it begins with a @samp{-}. You can specify both short and long flags; for example, @option{--version} is exactly the same as @option{-v}, but is easier to remember. Short flags may be specified one at a time, or in a group. A short flag or a group of short flags always starts off with a single dash to indicate that what follows is a flag or set of flags instead of a file name; a long flag starts off with two consecutive dashes, without spaces between them. In the current implementation the flags can be intermixed with file names, but their effect is as if they were all specified first. The various flags are interpreted as follows: @table @option @item -a @itemx --smalltalk-args @findex arguments Treat all options afterward as arguments to be given to Smalltalk code retrievable with @code{Smalltalk arguments}, ignoring them as arguments to @gst{} itself. Examples: @multitable {@option{--verbose -aq -c}} {Options seen by @sc{gnu} Smalltalk} {@code{Smalltalk arguments}} @item command line @tab Options seen by @gst{} @tab @code{Smalltalk arguments} @item (empty) @tab (none) @tab @code{#()} @item @option{-Via foo bar} @tab @option{-Vi} @tab @code{#('foo' 'bar')} @item @option{-Vai test} @tab @option{-Vi} @tab @code{#('test')} @item @option{-Vaq} @tab @option{-Vq} @tab @code{#()} @item @option{--verbose -aq -c } @tab @option{--verbose -q} @tab @code{#('-c')} @end multitable @item -c @itemx --core-dump When a fatal signal occurs, produce a core dump before terminating. Without this option, only a backtrace is provided. @item -D @itemx --declaration-trace Print the class name, the method name, and the byte codes that the compiler generates as it compiles methods. Only applies to files that are named explicitly on the command line, unless the flag is given multiple times on the command line. @item -E @itemx --execution-trace Print the byte codes being executed as the interpreter operates. Only works for statements explicitly issued by the user (either interactively or from files given on the command line), unless the flag is given multiple times on the command line. @ignore This option is disabled when the dynamic translator (@pxref{Dynamic translator}) is enabled. @end ignore @item --kernel-directory Specify the directory from which the kernel source files will be loaded. This is used mostly while compiling @gst{} itself. Smalltalk code can retrieve this information with @code{Directory kernel}. @item --no-user-files Don't load any files from @file{~/.st/} (@pxref{Loading or creating an image,, Loading an image or creating a new one}).@footnote{The directory would be called @file{_st/} under MS-DOS. Under OSes that don't use home directories, it would be looked for in the current directory.} This is used mostly while compiling @gst{} itself, to ensure that the installed image is built only from files in the source tree. @item -K @var{file} @itemx --kernel-file @var{file} Load @var{file} in the usual way, but look for it relative to the kernel directory's parent directory, which is usually @file{/usr/local/share/smalltalk/}. See @option{--kernel-dir} above. @cindex shell scripts @item -f @itemx --file The following two command lines are equivalent: @example gst -f @var{file} @file{args...} gst -q @var{file} -a @file{args...} @end example This is meant to be used in the so called sharp-bang'' sequence at the beginning of a file, as in @example #! /usr/bin/gst -f @r{@i{@dots{} Smalltalk source code @dots{}}} @end example @gst{} treats the first line as a comment, and the @option{-f} option ensures that the arguments are passed properly to the script. Use this instead to avoid hard-coding the path to @command{gst}:@footnote{The words in the shell command @command{exec} are all quoted, so GNU Smalltalk parses them as five separate comments.} @example #! /bin/sh "exec" "gst" "-f" "$0" "$@@" @r{@i{@dots{} Smalltalk source code @dots{}}} @end example @item -g @itemx --no-gc-messages Suppress garbage collection messages. @item -h @itemx --help Print out a brief summary of the command line syntax of @gst{}, including the definitions of all of the option flags, and then exit. @item -i @itemx --rebuild-image Always build and save a new image file; see @ref{Loading or creating an image,, Loading an image or creating a new one}. @item --maybe-rebuild-image Perform the image checks and rebuild as described in @ref{Loading or creating an image,, Loading an image or creating a new one}. This is the default when @option{-I} is not given. @cindex image path @item -I @var{file} @itemx --image-file @var{file} Use the image file named @var{file} as the image file to load instead of the default location, and set @var{file}'s directory part as the image path. This option completely bypasses checking the file dates on the kernel files; use @option{--maybe-rebuild-image} to restore the usual behavior, writing the newly built image to @var{file} if needed. @item -q @itemx --quiet @itemx --silent Suppress the printing of answered values from top-level expressions while @gst{} runs. @item -r @itemx --regression-test This is used by the regression testing system and is probably not of interest to the general user. It controls printing of certain information. @item -S @itemx --snapshot Save the image after loading files from the command line. Of course this snapshot'' is not saved if you include - (stdin) on the command line and exit by typing @kbd{Ctrl-c}. @item -v @itemx --version Print out the @gst{} version number, then exit. @item -V @itemx --verbose Print various diagnostic messages while executing (the name of each file as it's loaded, plus messages about the beginning of execution or how many byte codes were executed). @end table @node Operation @section Startup sequence @strong{Caveat}: @emph{The startup sequence is pretty complicated. If you are not interested in its customization, you can skip the first two sections below. These two sections also don't apply when using the command-line option @option{-I}, unless also using @option{--maybe-rebuild-image}.} You can abort @gst{} at any time during this procedure with @kbd{Ctrl-c}. @menu * Command-line processing:: Picking an image path and a kernel path. * Loading or creating an image:: Loading an image or creating a new one. * Starting the system:: After the image is created or restored. @end menu @node Command-line processing @subsection Picking an image path and a kernel path @cindex image path When @gst{} is invoked, it first chooses two paths, the image path'' and the kernel path''. The image path is set by considering these paths in succession: @itemize @item the directory part of the @option{--image-file} option if it is given; @item the value of the @env{SMALLTALK_IMAGE} environment variable if it is defined and readable; this step will disappear in a future release; @item the path compiled in the binary (usually, under Unix systems, @file{/usr/local/var/lib/smalltalk} or a similar path under @file{/var}) if it exists and it is readable; @item the current directory. The current directory is also used if the image has to be rebuilt but you cannot write to a directory chosen according to the previous criteria. @end itemize @cindex kernel path The kernel path'' is the directory in which to look for Smalltalk code compiled into the base image. The possibilities in this case are: @itemize @item the argument to the @option{--kernel-dir} option if it is given; @item the value of the @env{SMALLTALK_KERNEL} environment variable if it is defined and readable; this step will disappear in a future release; @item the path compiled in the binary (usually, under Unix systems, @file{/usr/local/share/smalltalk/kernel} or a similar data file path) if it exists and it is readable; @item a subdirectory named @file{kernel} of the image path. @end itemize @node Loading or creating an image @subsection Loading an image or creating a new one @cindex compatible images @cindex images, loading @gst{} can load images created on any system with the same pointer size as its host system by approximately the same version of @gst{}, even if they have different endianness. For example, images created on 32-bit PowerPC can be loaded with a 32-bit x86 @command{gst} @acronym{VM}, provided that the @gst{} versions are similar enough. Such images are called @dfn{compatible images}. It cannot load images created on systems with different pointer sizes; for example, our x86 @command{gst} cannot load an image created on x86-64. Unless the @option{-i} flag is used, @gst{} first tries to load the file named by @option{--image-file}, defaulting to @file{gst.im} in the image path. If this is found, @gst{} ensures the image is not stale'', meaning its write date is newer than the write dates of all of the kernel method definition files. It also ensures that the image is compatible'', as described above. If both tests pass, @gst{} loads the image and continues with @ref{Starting the system,, After the image is created or restored}. If that fails, a new image has to be created. The image path may now be changed to the current directory if the previous choice is not writeable. @cindex kernel, loading To build an image, @gst{} loads the set of files that make up the kernel, one at a time. The list can be found in @file{libgst/lib.c}, in the @code{standard_files} variable. You can override kernel files by placing your own copies in @file{~/.st/kernel/}.@footnote{The directory is called @file{_st/kernel} under MS-DOS. Under OSes that don't use home directories, it is looked for in the current directory.} For example, if you create a file @file{~/.st/kernel/Builtins.st}, it will be loaded instead of the @file{Builtins.st} in the kernel path. @cindex @file{pre.st} @cindex @file{site-pre.st} To aid with image customization and local bug fixes, @gst{} loads two more files (if present) before saving the image. The first is @file{site-pre.st}, found in the parent directory of the kernel directory. Unless users at a site change the kernel directory when running @command{gst}, @file{/usr/local/share/smalltalk/site-pre.st} provides a convenient place for site-wide customization. The second is @file{~/.st/pre.st}, which can be different for each user's home directory.@footnote{The file is looked up as @file{_st/pre.st} under MS-DOS and again, under OSes that don't use home directories it is looked for as @file{pre.st} in the current directory.}. Before the next steps, @gst{} takes a snapshot of the new memory image, saving it over the old image file if it can, or in the current directory otherwise. @node Starting the system @subsection After the image is created or restored @c so it's not a "function"... it's an operation @findex returnFromSnapshot @cindex @file{init.st} Next, @gst{} sends the @code{returnFromSnapshot} event to the dependents of the special class @code{ObjectMemory} (@pxref{Memory access}). Afterwards, it loads @file{~/.st/init.st} if available.@footnote{The same considerations made above hold here too. The file is called @file{_st/init.st} under MS-DOS, and is looked for in the current directory under OSes that don't use home directories.} @cindex startup, customizing @cindex customizing startup You can remember the difference between @file{pre.st} and @file{init.st} by remembering that @file{pre.st} is the @emph{pre}-snapshot file and @file{init.st} is the post-image-load @emph{init}ialization file. Finally, @gst{} loads files listed on the command line, or prompts for input at the terminal, as described in @ref{Invocation,, Command line arguments}. @node Syntax @section Syntax of @gst{} The language that @gst{} accepts is basically the same that other Smalltalk environment accept and the same syntax used in the @dfn{Blue Book}, also known as @cite{Smalltalk-80: The Language and Its Implementation}. The return operator, which is represented in the Blue Book as an up-arrow, is mapped to the ASCII caret symbol @code{^}; the assignment operator (left-arrow) is usually represented as @code{:=}@footnote{It also bears mentioning that there are two assignment operators: @code{_} and @code{:=}. Both are usable interchangeably, provided that they are surrounded by spaces. The @gst{} kernel code uses the @code{:=} form exclusively, but @code{_} is supported a) for compatibility with previous versions of @gst{} b) because this is the correct mapping between the assignment operator mentioned in the Blue Book and the current ASCII definition. In the ancient days (like the middle 70's), the ASCII underscore character was also printed as a back-arrow, and many terminals would display it that way, thus its current usage. Anyway, using @code{_} may lead to portability problems.}. Actually, the grammar of @gst{} is slightly different from the grammar of other Smalltalk environments in order to simplify interaction with the system in a command-line environment as well as in full-screen editors. Statements are executed one by one; multiple statements are separated by a period. At end-of-line, if a valid statement is complete, a period is implicit. For example, @example 8r300. 16rFFFF @end example @noindent prints out the decimal value of octal @code{300} and hex @code{FFFF}, each followed by a newline. Multiple statements share the same local variables, which are automatically declared. To delete the local variables, terminate a statement with @code{!} rather than @code{.} or newline. Here, @example a := 42 a! a @end example @noindent the first two @code{a}s are printed as @code{42}, but the third one is uninitialized and thus printed as @code{nil}. In order to evaluate multiple statements in a single block, wrap them into an @dfn{eval block} as follows: @example Eval [ a := 42. a printString ] @end example @noindent This won't print the intermediate result (the integer 42), only the final result (the string @code{'42'}). @example ObjectMemory quit @end example @noindent exits from the system. You can also type a @kbd{C-d} to exit from Smalltalk if it's reading statements from standard input. @gst{} provides three extensions to the language that make it simpler to write complete programs in an editor. However, it is also compatible with the @dfn{file out} syntax as shown in the @dfn{Green Book} (also known as @cite{Smalltalk-80: Bits of History, Words of Advice} by Glenn Krasner). A new class is created using this syntax: @display @var{superclass-name} @t{subclass:} @var{new-class-name} @t{[} @t{|} @var{instance variables} @t{|} @var{pragmas} @var{message-pattern-1} @t{[} @var{statements} @t{]} @var{message-pattern-2} @t{[} @var{statements} @t{]} @dots{} @var{class-variable-1} @t{:=} @var{expression}@t{.} @var{class-variable-2} @t{:=} @var{expression}@t{.} @dots{} @t{]} @end display In short: @itemize @bullet @item Instance variables are defined with the same syntax as method temporary variables. @item Unlike other Smalltalks, method statements are inside brackets. @item Class variables are defined the same as variable assignments. @item Pragmas define class comment, class category, imported namespaces, and the shape of indexed instance variables. @example @end example @end itemize A similar syntax is used to define new methods in an existing class. @display @var{class-expression} @t{extend} @t{[} @dots{} @t{]} @end display The @var{class-expression} is an expression that evaluates to a class object, which is typically just the name of a class, although it can be the name of a class followed by the word @code{class}, which causes the method definitions that follow to apply to the named class itself, rather than to its instances. @example Number extend [ radiusToArea [ ^self squared * Float pi ] radiusToCircumference [ ^self * 2 * Float pi ] ] @end example A complete treatment of the Smalltalk syntax and of the class library can be found in the included tutorial and class reference (@pxref{Top, , Class Reference, gst-base, the @gst{} Library Reference}). More information on the implementation of the language can be found in the @cite{Blue Book}; the relevant parts are available, scanned, at @url{http://stephane.ducasse.free.fr/FreeBooks/BlueBook/Bluebook.pdf}. @node Test suite @section Running the test suite @gst{} comes with a set of files that provides a simple regression test suite. To run the test suite, you should be connected to the top-level Smalltalk directory. Type @example make check @end example You should see the names of the test suite files as they are processed, but that's it. Any other output indicates some problem. @node Legal concerns @section Licensing of @gst{} Different parts of @gst{} comes under two licenses: the virtual machine and the development environment (compiler and browser) come under the @gnu{} General Public License, while the system class libraries come under the Lesser General Public License. @menu * GPL:: Complying with the GNU GPL. * LGPL:: Complying with the GNU LGPL. @end menu @node GPL @subsection Complying with the @gnu{} @acronym{GPL} The @acronym{GPL} licensing of the virtual machine means that all derivatives of the virtual machine must be put under the same license. In other words, it is strictly forbidden to distribute programs that include the @gst{} virtual machine under a license that is not the GPL. This also includes any bindings to external libraries. For example, the bindings to Gtk+ are released under the @acronym{GPL}. In principle, the @acronym{GPL} would not extend to Smalltalk programs, since these are merely input data for the virtual machine. On the other hand, using bindings that are under the @acronym{GPL} via dynamic linking would constitute combining two parts (the Smalltalk program and the bindings) into one program. Therefore, we added a special exception to the @acronym{GPL} in order to avoid gray areas that could adversely hit both the project and its users: @quotation In addition, as a special exception, the Free Software Foundation give you permission to combine @gst{} with free software programs or libraries that are released under the @gnu{} @acronym{LGPL} and with independent programs running under the @gst{} virtual machine. You may copy and distribute such a system following the terms of the @gnu{} @acronym{GPL} for @gst{} and the licenses of the other code concerned, provided that you include the source code of that other code when and as the @gnu{} @acronym{GPL} requires distribution of source code. Note that people who make modified versions of @gst{} are not obligated to grant this special exception for their modified versions; it is their choice whether to do so. The @gnu{} General Public License gives permission to release a modified version without this exception; this exception also makes it possible to release a modified version which carries forward this exception. @end quotation @node LGPL @subsection Complying with the @gnu{} @acronym{LGPL} Smalltalk programs that run under @gst{} are linked with the system classes in @gst{} class library. Therefore, they must respect the terms of the Lesser General Public License@footnote{Of course, they may be more constrained by usage of @acronym{GPL} class libraries.}. The interpretation of this license for architectures different from that of the C language is often difficult; the accepted one for Smalltalk is as follows. The image file can be considered as an object file, falling under Subsection 6a of the license, as long as it allows a user to load an image, upgrade the library or otherwise apply modifications to it, and save a modified image: this is most conveniently obtained by allowing the user to use the read-eval-print loop that is embedded in the @gst{} virtual machine. In other words, provided that you leave access to the loop in a documented way, or that you provide a way to file in arbitrary files in an image and save the result to a new image, you are obeying Subsection 6a of the Lesser General Public License, which is reported here: @quotation a) Accompany the work with the complete corresponding machine-readable source code for the Library including whatever changes were used in the work (which must be distributed under Sections 1 and 2 above); and, if the work is an executable linked with the Library, with the complete machine-readable "work that uses the Library", as object code and/or source code, so that the user can modify the Library and then relink to produce a modified executable containing the modified Library. (It is understood that the user who changes the contents of definitions files in the Library will not necessarily be able to recompile the application to use the modified definitions.) @end quotation In the future, alternative mechanisms similar to shared libraries may be provided, so that it is possible to comply with the @gnu{} @acronym{LGPL} in other ways. @node Features @chapter Features of @gst{} In this section, the features which are specific to @gst{} are described. These features include support for calling C functions from within Smalltalk, accessing environment variables, and controlling various aspects of compilation and execution monitoring. Note that, in general, @gst{} is much more powerful than the original Smalltalk-80, as it contains a lot of methods that are common in today's Smalltalk implementation and are present in the ANSI Standard for Smalltalk, but were absent in the Blue Book. Examples include Collection's @code{allSatisfy:} and @code{anySatisfy:} methods and many methods in SystemDictionary (the Smalltalk dictionary's class). @menu * Extended streams:: Extensions to streams, and generators * Regular expressions:: String matching extensions * Namespaces:: Avoiding clashes between class names. * Disk file-IO:: Methods for reading and writing disk files. * Object dumping:: Methods that read and write objects in binary format. * Dynamic loading:: Picking external libraries and modules at run-time. * Documentation:: Automatic documentation generation. * Memory access:: The direct memory accessing classes and methods, plus broadcasts from the virtual machine. * GC:: The @gst{} memory manager. * Security:: Sandboxing and access control. * Special objects:: Methods to assign particular properties to objects. @end menu @node Extended streams @section Extended streams The basic image in @gst{} includes powerful extensions to the @emph{Stream} hierarchy found in ANSI Smalltalk (and Smalltalk-80). In particular: @itemize @bullet @item Read streams support all the iteration protocols available for collections. In some cases (like @code{fold:}, @code{detect:}, @code{inject:into:}) these are completely identical. For messages that return a new stream, such as @code{select:} and @code{collect:}, the blocks are evaluated lazily, as elements are requested from the stream using @code{next}. @item Read streams can be concatenated using @code{,} like SequenceableCollections. @item @dfn{Generators} are supported as a quick way to create a Stream. A generator is a kind of pluggable stream, in that a user-supplied blocks defines which values are in a stream. For example, here is an empty generator and two infinite generators: @example "Returns an empty stream" Generator on: [ :gen | ] "Return an infinite stream of 1's" Generator on: [ :gen | [ gen yield: 1 ] repeat ] "Return an infinite stream of integers counting up from 1" Generator inject: 1 into: [ :value | value + 1 ] @end example The block is put on hold'' and starts executing as soon as @code{#next} or @code{#atEnd} are sent to the generator. When the block sends @code{#yield:} to the generator, it is again put on hold and the argument becomes the next object in the stream. Generators use @dfn{continuations}, but they shield the users from their complexity by presenting the same simple interface as streams. @end itemize @node Regular expressions @section Regular expression matching @emph{Regular expressions}, or "regexes", are a sophisticated way to efficiently match patterns of text. If you are unfamiliar with regular expressions in general, see @ref{Regexps, Syntax of Regular Expressions, 20.5 Syntax of Regular Expressions, emacs, GNU Emacs Manual}, for a guide for those who have never used regular expressions. @gst{} supports regular expressions in the core image with methods on @code{String}. The @gst{} regular expression library is derived from GNU libc, with modifications made originally for Ruby to support Perl-like syntax. It will always use its included library, and never the ones installed on your system; this may change in the future in backwards-compatible ways. Regular expressions are currently 8-bit clean, meaning they can work with any ordinary String, but do not support full Unicode, even when package @code{I18N} is loaded. Broadly speaking, these regexes support Perl 5 syntax; register groups @samp{()} and repetition @samp{@{@}} must not be given with backslashes, and their counterpart literal characters should. For example, @samp{\@{@{1,3@}} matches @samp{@{}, @samp{@{@{}, @samp{@{@{@{}; correspondingly, @samp{(a)(\()} matches @samp{a(}, with @samp{a} and @samp{(} as the first and second register groups respectively. @gst{} also supports the regex modifiers @samp{imsx}, as in Perl. You can't put regex modifiers like @samp{im} after Smalltalk strings to specify them, because they aren't part of Smalltalk syntax. Instead, use the inline modifier syntax. For example, @samp{(?is:abc.)} is equivalent to @samp{[Aa][Bb][Cc](?:.|\n)}. In most cases, you should specify regular expressions as ordinary strings. @gst{} always caches compiled regexes, and uses a special high-efficiency caching when looking up literal strings (i.e. most regexes), to hide the compiled @code{Regex} objects from most code. For special cases where this caching is not good enough, simply send @code{#asRegex} to a string to retrieved a compiled form, which works in all places in the public API where you would specify a regex string. You should always rely on the cache until you have demonstrated that using Regex objects makes a noticeable performance difference in your code. Smalltalk strings only have one escape, the @samp{'} given by @samp{''}, so backslashes used in regular expression strings will be understood as backslashes, and a literal backslash can be given directly with @samp{\\}@footnote{Whereas it must be given as @samp{\\\\} in a literal Emacs Lisp string, for example.}. The methods on the compiled Regex object are private to this interface. As a public interface, @gst{} provides methods on String, in the category @samp{regex}. There are several methods for matching, replacing, pattern expansion, iterating over matches, and other useful things. The fundamental operator is @code{#searchRegex:}, usually written as @code{#=~}, reminiscent of Perl syntax. This method will always return a @code{RegexResults}, which you can query for whether the regex matched, the location Interval and contents of the match and any register groups as a collection, and other features. For example, here is a simple configuration file line parser: @example | file config | config := LookupTable new. file := (File name: 'myapp.conf') readStream. file linesDo: [:line | (line =~ '(\w+)\s*=\s*((?: ?\w+)+)') ifMatched: [:match | config at: (match at: 1) put: (match at: 2)]]. file close. config printNl. @end example As with Perl, @code{=~} will scan the entire string and answer the leftmost match if any is to be found, consuming as many characters as possible from that position. You can anchor the search with variant messages like @code{#matchRegex:}, or of course @code{^} and @code{$} with their usual semantics if you prefer. You shouldn't modify the string while you want a particular RegexResults object matched on it to remain valid, because changes to the matched text may propagate to the RegexResults object. @c (currently "will", but best to leave open) Analogously to the Perl @code{s} operator, @gst{} provides @code{#replacingRegex:with:}. Unlike Perl, @gst{} employs the pattern expansion syntax of the @code{#%} message here. For example, @code{'The ratio is 16/9.' replacingRegex: '(\d+)/(\d+)' with: '$%1\over%2$'} answers @code{'The ratio is$16\over9$.'}. In place of the @code{g} modifier, use the @code{#replacingAllRegex:with:} message instead. One other interesting String message is @code{#onOccurrencesOfRegex:do:}, which invokes its second argument, a block, on every successful match found in the receiver. Internally, every search will start at the end of the previous successful match. For example, this will print all the words in a stream: @example stream contents onOccurrencesOfRegex: '\w+' do: [:each | each match printNl] @end example @node Namespaces @section Namespaces @i{[This section (and the implementation of namespaces in @gst{}) is based on the paper @cite{Structured Symbolic Name Spaces in Smalltalk}, by Augustin Mrazik.]} @subsection Introduction The Smalltalk-80 programming environment, upon which @gst{} is historically based, supports symbolic identification of objects in one global namespace---in the @code{Smalltalk} system dictionary. This means that each global variable in the system has its unique name which is used for symbolic identification of the particular object in the source code (e.g.@: in expressions or methods). The most important of these global variables are classes defining the behavior of objects. In development dealing with modelling of real systems, @dfn{polymorphic symbolic identification} is often needed. By this, we mean that it should be possible to use the same name for different classes or other global variables. Selection of the proper variable binding should be context-specific. By way of illustration, let us consider class @code{Statement} as an example which would mean totally different things in different domains: @table @asis @item @gst{} or other programming language An expression in the top level of a code body, possibly with special syntax available such as assignment or branching. @item Bank A customer's trace report of recent transactions. @item AI, logical derivation An assertion of a truth within a logical system. @end table This issue becomes inevitable if we start to work persistently, using @code{ObjectMemory snapshot} to save after each session for later resumption. For example, you might have the class @code{Statement} already in your image with the Bank'' meaning above (e.g.@: in the live bank support systems we all run in our images) and you might decide to start developing @acronym{YAC} [Yet Another C]. Upon starting to write parse nodes for the compiler, you would find that @code{#Statement} is boundk in the banking package. You could replace it with your parse node class, and the bank's @code{Statement} could remain in the system as an unbound class with full functionality; however, it could not be accessed anymore at the symbolic level in the source code. Whether this would be a problem or not would depend on whether any of the bank's code refers to the class @code{Statement}, and when these references occur. Objects which have to be identified in source code by their names are included in @code{Smalltalk}, the sole instance of @code{SystemDictionary}. Such objects may be identified simply by writing their names as you would any variable names. The code is compiled in the default environment, and if the variable is found in @code{Smalltalk}, without being shadowed by a class pool or local variables, its value is retrieved and used as the value of the expression. In this way @code{Smalltalk} represents the sole symbolic namespace. In the following text the symbolic namespace, as a concept, will be called simply @dfn{environment} to make the text more clear. @subsection Concepts To support polymorphic symbolical identification several environments will be needed. The same name may exist concurrently in several environments as a key, pointing to diverse objects in each. Symbolic navigation between these environments is needed. Before approaching the problem of the syntax and semantics to be implemented, we have to decide on structural relations to be established between environments. Since the environment must first be symbolically identified to direct access to its global variables, it must first itself be a global variable in another environment. @code{Smalltalk} is a great choice for the root environment, from which selection of other environments and their variables begins. From @code{Smalltalk} some of the existing sub-environments may be seen; from these other sub-environments may be seen, etc. This means that environments represent nodes in a graph where symbolic selections from one environment to another one represent branches. The symbolic identification should be unambiguous, although it will be polymorphic. This is why we should avoid cycles in the environment graph. Cycles in the graph could cause also other problems in the implementation, e.g.@: inability to use trivially recursive algorithms. Thus, in general, the environments must build a directed acyclic graph; @gst{} currently limits this to an n-ary tree, with the extra feature that environments can be used as pool dictionaries. Let us call the partial ordering relation which occurs between environments @dfn{inheritance}. Sub-environments inherit from their super-environments. The feature of inheritance in the meaning of object-orientation is associated with this relation: all associations of the super-environment are valid also in its sub-environments, unless they are locally redefined in the sub-environment. A super-environment includes all its sub-enviroments as @code{Association}s under their names. The sub-environment includes its super-environment under the symbol @code{#Super}. Most environments inherit from @code{Smalltalk}, the standard root environment, but they are not required to do so; this is similar to how most classes derive from @code{Object}, yet one can derive a class directly from @code{nil}. Since they all inherit @code{Smalltalk}'s global variables, it is not necessary to define @code{Smalltalk} as pointing to @code{Smalltalk}'s @code{Smalltalk} in each environment. The inheritance links to the super-environments are used in the lookup for a potentially inherited global variable. This includes lookups by a compiler searching for a variable binding and lookups via methods such as @code{#at:} and @code{#includesKey:}. @subsection Syntax Global objects of an environment, be they local or inherited, may be referenced by their symbol variable names used in the source code, e.g. @example John goHome @end example @noindent if the @code{#John -> aMan} association exists in the particular environment or one of its super-environments, all along the way to the root environment. If an object must be referenced from another environment (i.e.@: which is not one of its sub-environments) it has to be referenced either @emph{relatively} to the position of the current environment, using the @code{Super} symbol, or @emph{absolutely}, using the full pathname'' of the object, navigating from the tree root (usually @code{Smalltalk}) through the tree of sub-environments. For the identification of global objects in another environment, we use a pathname'' of symbols. The symbols are separated by periods; the look'' to appear is that of @example Smalltalk.Tasks.MyTask @end example @noindent and of @example Super.Super.Peter. @end example As is custom in Smalltalk, we are reminded by capitalization that we are accessing global objects. Another syntax returns the @dfn{variable binding}, the @code{Association} for a particular global. The first example above is equivalently: @example #@{Smalltalk.Tasks.MyTask@} value @end example The latter syntax, a @dfn{variable binding}, is also valid inside literal arrays. @subsection Implementation A superclass of @code{SystemDictionary} called @code{RootNamespace} is defined, and many of the features of the Smalltalk-80 @code{SystemDictionary} will be hosted by that class. @code{Namespace} and @code{RootNamespace} are in turn subclasses of @code{AbstractNamespace}. To handle inheritance, the following methods have to be defined or redefined in Namespace (@emph{not} in RootNamespace): @table @asis @item Accessors like @code{#at:ifAbsent:} and @code{#includesKey:} Inheritance must be implemented. When @code{Namespace}, trying to read a variable, finds an association in its own dictionary or a super-environment dictionary, it uses that; for @code{Dictionary}'s writes and when a new association must be created, @code{Namespace} creates it in its own dictionary. There are special methods like @code{#set:to:} for cases in which you want to modify a binding in a super-environment if that is the relevant variable's binding. @c this needs more clarity for #at:put: #set:to: disambig @item Enumerators like @code{#do:} and @code{#keys} This should return @strong{all} the objects in the namespace, including those which are inherited. @item Hierarchy access @code{AbstractNamespace} will also implement a new set of methods that allow one to navigate through the namespace hierarchy; these parallel those found in @code{Behavior} for the class hierarchy. @end table The most important task of the @code{Namespace} class is to provide organization for the most important global objects in the Smalltalk system---for the classes. This importance becomes even more crucial in a structure of multiple environments intended to change the semantics of code compiled for those classes. In Smalltalk the classes have the instance variable @code{name} which holds the name of the class. Each @dfn{defined class} is included in @code{Smalltalk}, or another environment, under this name. In a framework with several environments the class should know the environment in which it has been created and compiled. This is a new property of @code{Class} which must be defined and properly used in relevant methods. In the mother environment the class shall be included under its name. Any class, as with any other object, may be included concurrently in several environments, even under different symbols in the same or in diverse environments. We can consider these alias names'' of the particular class or other value. A class may be referenced under the other names or in other environments than its mother environment, e.g.@: for the purpose of instance creation or messages to the class, but it should not compile code in these environments, even if this compilation is requested from another environment. If the syntax is not correct in the mother environment, a compilation error occurs. This follows from the existence of class mother environments'', as a class is responsible for compiling its own methods. An important issue is also the name of the class answered by the class for the purpose of its identification in diverse tools (e.g.@: in a browser). This must be changed to reflect the environment in which it is shown, i.e.@: the method @samp{nameIn: environment} must be implemented and used in proper places. Other changes must be made to the Smalltalk system to achieve the full functionality of structured environments. In particular, changes have to be made to the behavior classes, the user interface, the compiler, and a few classes supporting persistance. One small detail of note is that evaluation in the @acronym{REPL} or @samp{Workspace}, implemented by compiling methods on @code{UndefinedObject}, make more sense if @code{UndefinedObject}'s environment is the current environment'' as reachable by @code{Namespace current}, even though its mother environment by any other sensibility is @code{Smalltalk}. @subsection Using namespaces Using namespaces is often merely a matter of adding a @samp{namespace} option to the @gst{} @acronym{XML} package description used by @code{PackageLoader}, or wrapping your code like this: @example Namespace current: NewNS [ @r{@dots{}} ] @end example Namespaces can be imported into classes like this: @example Stream subclass: EncodedStream [ ] @end example @noindent Alternatively, paths to classes (and other objects) in the namespaces will have to be specified completely. Importing a namespace into a class is similar to C++'s @code{using namespace} declaration within the class proper's definition. Finally, be careful when working with fundamental system classes. Although you can use code like @example Namespace current: NewNS [ Smalltalk.Set subclass: Set [ @r{@dots{}} ] ] @end example @noindent this approach won't work when applied to core classes. For example, you might be successful with a @code{Set} or @code{WriteStream} object, but subclassing @code{SmallInteger} this way can bite you in strange ways: integer literals will still belong to the @code{Smalltalk} dictionary's version of the class (this holds for @code{Array}s, @code{String}s, etc.@: too), primitive operations will still answer standard Smalltalk @code{SmallIntegers}, and so on. Similarly, word-shaped will recognize 32-bit @code{Smalltalk.LargeInteger} objects, but not @code{LargeInteger}s belonging to your own namespace. Unfortunately, this problem is not easy to solve since Smalltalk has to know the @acronym{OOP}s of determinate class objects for speed---it would not be feasible to lookup the environment to which sender of a message belongs every time the @code{+} message was sent to an Integer. So, @gst{} namespaces cannot yet solve 100% of the problem of clashes between extensions to a class---for that you'll still have to rely on prefixes to method names. But they @emph{do} solve the problem of clashes between class names, or between class names and pool dictionary names. Namespaces are unrelated from packages; loading a package does not import the corresponding namespace. @node Disk file-IO @section Disk file-IO primitive messages Four classes (@code{FileDescriptor}, @code{FileStream}, @code{File}, @code{Directory}) allow you to create files and access the file system in a fully object-oriented way. @code{FileDescriptor} and @code{FileStream} are much more powerful than the corresponding C language facilities (the difference between the two is that, like the C @code{stdio} library, @code{FileStream} does buffering). For one thing, they allow you to write raw binary data in a portable endian-neutral format. But, more importantly, these classes transparently implement virtual filesystems and asynchronous I/O. Asynchronous I/O means that an input/output operation blocks the Smalltalk Process that is doing it, but not the others, which makes them very useful in the context of network programming. Virtual file systems mean that these objects can transparently extract files from archives such as @file{tar} and @file{gzip} files, through a mechanism that can be extended through either shell scripting or Smalltalk programming. For more information on these classes, look in the class reference, under the @code{VFS} namespace. @acronym{URL}s may be used as file names; though, unless you have loaded the @code{NetClients} package (@pxref{Network support}), only @code{file} @acronym{URL}s will be accepted. In addition, the three files, @code{stdin}, @code{stdout}, and @code{stderr} are declared as global instances of @code{FileStream} that are bound to the proper values as passed to the C virtual machine. They can be accessed as either @code{stdout} and @code{FileStream stdout}---the former is easier to type, but the latter can be clearer. Finally, @code{Object} defines four other methods: @code{print} and @code{printNl}, @code{store} and @code{storeNl}. These do a @code{printOn:} or @code{storeOn:} to the Transcript'' object; this object, which is the sole instance of class @code{TextCollector}, normally delegates write operations to @code{stdout}. If you load the VisualGST @sc{gui}, instead, the Transcript Window will be attached to the Transcript object (@pxref{GUI, , GTK and VisualGST}). The @code{fileIn:} message sent to the FileStream class, with a file name as a string argument, will cause that file to be loaded into Smalltalk. For example, @example FileStream fileIn: 'foo.st' ! @end example @noindent will cause @file{foo.st} to be loaded into @gst{}. @node Object dumping @section The @gst{} ObjectDumper Another @gst{}-specific class, the @code{ObjectDumper} class, allows you to dump objects in a portable, endian-neutral, binary format. Note that you can use the @code{ObjectDumper} on ByteArrays too, thanks to another @gst{}-specific class, @code{ByteStream}, which allows you to treat ByteArrays the same way you would treat disk files. For more information on the usage of the @code{ObjectDumper}, look in the class reference. @node Dynamic loading @section Dynamic loading The @code{DLD} class enhances the C callout mechanism to automatically look for unresolved functions in a series of program-specified libraries. To add a library to the list, evaluate code like the following: @example DLD addLibrary: 'libc' @end example The extension (@file{.so}, @file{.sl}, @file{.a}, @file{.dll} depending on your operating system) will be added automatically. You are advised not to specify it for portability reasons. You will then be able to use the standard C call-out mechanisms to define all the functions in the C run-time library. Note that this is a potential security problem (especially if your program is SUID root under Unix), so you might want to disable dynamic loading when using @gst{} as an extension language. To disable dynamic loading, configure @gst{} passing the @option{--disable-dld} switch. Note that a @code{DLD} class will be present even if dynamic loading is disabled (either because your system is not supported, or by the @option{--disable-dld} configure switch) but any attempt to perform dynamic linking will result in an error. @node Documentation @section Automatic documentation generator @gst{} includes an automatic documentation generator invoked via the @command{gst-doc} command. The code is actually part of the @code{ClassPublisher} package, and @command{gst-doc} takes care of reading the code to be documented and firing a @code{ClassPublisher}. Currently, @command{gst-doc} can only generate output in Texinfo format, though this will change in future releases. @command{gst-doc} can document code that is already in the image, or it can load external files and packages. Note that the latter approach will not work for files and packages that programmatically create code or file in other files/packages. @command{gst-doc} is invoked as follows: @example gst-doc [ @var{flag} ... ] @var{class} ... @end example The following options are supported: @table @option @item -p @var{package} @itemx --package=@var{package} Produce documentation for the classes inside the @var{package} package. @item -f @var{file} @itemx --file=@var{file} Produce documentation for the classes inside the @var{file} file. @item -I @itemx --image-file Produce documentation for the code that is already in the given image. @item -o @itemx --output=@var{file} Emit documentation in the named file. @end table @var{class} is either a class name, or a namespace name followed by @code{.*}. Documentation will be written for classes that are specified in the command line. @var{class} can be omitted if a @option{-f} or @option{-p} option is given. In this case, documentation will be written for all the classes in the package. @node Memory access @section Memory accessing methods @gst{} provides methods to query its own internal data structures. You may determine the real memory address of an object or the real memory address of the OOP table that points to a given object, by using messages to the @code{Memory} class, described below. @defmethod Object asOop Returns the index of the OOP for anObject. This index is immume from garbage collection and is the same value used by default as an hash value for anObject (it is returned by Object's implementation of @code{hash} and @code{identityHash}). @end defmethod @defmethod Integer asObject Converts the given OOP @emph{index} (not address) back to an object. Fails if no object is associated to the given index. @end defmethod @defmethod Integer asObjectNoFail Converts the given OOP @emph{index} (not address) back to an object. Returns nil if no object is associated to the given index. @end defmethod Other methods in ByteArray and Memory allow to read various C types (@code{doubleAt:}, @code{ucharAt:}, etc.). These are mostly obsoleted by @code{CObject} which, in newer versions of @gst{}, supports manually managed heap-backed memory as well as garbage collected ByteArray-backed memory. Another interesting class is ObjectMemory. This provides a few methods that enable one to tune the virtual machine's usage of memory; many methods that in the past were instance methods of Smalltalk or class methods of Memory are now class methods of ObjectMemory. In addition, and that's what the rest of this section is about, the virtual machines signals events to its dependents exactly through this class. The events that can be received are @table @dfn @item returnFromSnapshot This is sent every time an image is restarted, and substitutes the concept of an @dfn{init block} that was present in previous versions. @item aboutToQuit This is sent just before the interpreter is exiting, either because @code{ObjectMemory quit} was sent or because the specified files were all filed in. Exiting from within this event might cause an infinite loop, so be careful. @item aboutToSnapshot This is sent just before an image file is created. Exiting from within this event will leave any preexisting image untouched. @item finishedSnapshot This is sent just after an image file is created. Exiting from within this event will not make the image unusable. @end table @node GC @section Memory management in @gst{} The @gst{} virtual machine is equipped with a garbage collector, a facility that reclaims the space occupied by objects that are no longer accessible from the system roots. The collector is composed of several parts, each of which can be invoked by the virtual machine using various tunable strategies, or invoked manually by the programmer. These parts include a @dfn{generation scavenger}, a @dfn{mark & sweep} collectory with an incremental sweep phase, and a @dfn{compactor}. All these facilities work on different memory spaces and differs from the other in its scope, speed and disadvantages (which are hopefully balanced by the availability of different algorithms). What follows is a description of these algorithms and of the memory spaces they work in. @dfn{NewSpace} is the memory space where young objects live. It is composed of three sub-spaces: an object-creation space (@dfn{Eden}) and two @dfn{SurvivorSpaces}. When an object is first created, it is placed in Eden. When Eden starts to fill up (i.e., when the number of used bytes in Eden exceeds the scavenge threshold), objects that are housed in Eden or in the occupied SurvivorSpace and that are still reachable from the system roots are copied to the unoccupied SurvivorSpace. As an object survives different scavenging passes, it will be shuffled by the scavenger from the occupied SurvivorSpace to the unoccupied one. When the number of used bytes in SurvivorSpace is high enough that the scavenge pause might be excessively long, the scavenger will move some of the older surviving objects from NewSpace to @dfn{OldSpace}. In the garbage collection jargon, we say that such objects are being @dfn{tenured} to OldSpace. This garbage collection algorithm is designed to reclaim short-lived objects, that is those objects that expire while residing in NewSpace, and to decide when enough data is residing in NewSpace that it is useful to move some of it in OldSpace. A @dfn{copying} garbage collector is particularly efficient in an object population whose members are more likely to die than survive, because this kind of scavenger spends most of its time copying survivors, who will be few in number in such populations, rather than tracing corpses, who will be many in number. This fact makes copying collection especially well suited to NewSpace, where a percentage of 90% or more objects often fails to survive across a single scavenge. The particular structure of NewSpace has many advantages. On one hand, having a large Eden and two small SurvivorSpaces has a smaller memory footprint than having two equally big semi-spaces and allocating new objects directly from the occupied one (by default, @gst{} uses 420=300+60*2 kilobytes of memory, while a simpler configuration would use 720=360*2 kilobytes). On the other hand, it makes tenuring decisions particularly simple: the copying order is such that short-lived objects tend to be copied last, while objects that are being referred from OldSpace tend to be copied first: this is because the tenuring strategy of the scavenger is simply to treat the destination SurvivorSpace as a circular buffer, tenuring objects with a First-In-First-Out policy. An object might become part of the scavenger root set for several reasons: objects that have been tenured are roots if their data lives in an OldSpace page that has been written to since the last scavenge (more on this later), plus all objects can be roots if they are known to be referenced from C code or from the Smalltalk stacks. In turn, some of the old objects can be made to live in a special area, called @dfn{FixedSpace}. Objects that reside in FixedSpace are special in that their body is guaranteed to remain at a fixed address (in general, @gst{} only ensures that the header of the object remains at a fixed address in the Object Table). Because the garbage collector can and does move objects, passing objects to foreign code which uses the object's address as a fixed key, or which uses a ByteArray as a buffer, presents difficulties. One can use @code{CObject} to manipulate C data on the @code{malloc} heap, which indeed does not move, but this can be tedious and requires the same attentions to avoid memory leaks as coding in C. FixedSpace provides a much more convenient mechanism: once an object is deemed fixed, the object's body will never move through-out its life-time; the space it occupies will however still be returned automatically to the FixedSpace pool when the object is garbage collected. Note that because objects in FixedSpace cannot move, FixedSpace cannot be compacted and can therefore suffer from extensive fragmentation. For this reason, FixedSpace should be used carefully. FixedSpace however is rebuilt (of course) every time an image is brought up, so a kind of compaction of FixedSpace can be achieved by saving a snapshot, quitting, and then restarting the newly saved image. Memory for OldSpace and FixedSpace is allocated using a variation of the system allocator @code{malloc}: in fact, @gst{} uses the same allocator for its own internal needs, for OldSpace and for FixedSpace, but it ensures that a given memory page never hosts objects that reside in separate spaces. New pages are mapped into the address space as needed and devoted to OldSpace or FixedSpace segments; similarly, when unused they may be subsequently unmapped, or they might be left in place waiting to be reused by @code{malloc} or by another Smalltalk data space. Garbage that is created among old objects is taken care of by a mark & sweep collector which, unlike the scavenger which only reclaims objects in NewSpace, can only reclaim objects in OldSpace. Note that as objects are allocated, they will not only use the space that was previously occupied in the Eden by objects that have survived, but they will also reuse the entries in the global Object Table that have been freed by object that the scavenger could reclaim. This quest for free object table entries can be combined with the sweep phase of the OldSpace collector, which can then be done incrementally, limiting the disruptive part of OldSpace garbage collection to the mark phase. Several runs of the mark & sweep collector can lead to fragmentation (where objects are allocated from several pages, and then become garbage in an order such that a bunch of objects remain in each page and the system is not able to recycle them). For this reason, the system periodically tries to compact OldSpace. It does so simply by looping through every old object and copying it into a new OldSpace. Since the OldSpace allocator does not suffer from fragmentation until objects start to be freed nor after all objects are freed, at the end of the copy all the pages in the fragmented OldSpace will have been returned to the system (some of them might already have been used by the compacted OldSpace), and the new, compacted OldSpace is ready to be used as the system OldSpace. Growing the object heap (which is done when it is found to be quite full even after a mark & sweep collection) automatically triggers a compaction. You can run the compactor without marking live objects. Since the amount of garbage in OldSpace is usually quite limited, the overhead incurred by copying potentially dead objects is small enough that the compactor still runs considerably faster than a full garbage collection, and can still give the application some breathing room. Keeping OldSpace and FixedSpace in the same heap would then make compaction of OldSpace (whereby it is rebuilt from time to time in order to limit fragmentation) much less effective. Also, the @code{malloc} heap is not used for FixedSpace objects because @gst{} needs to track writes to OldSpace and FixedSpace in order to support efficient scavenging of young objects. To do so, the grey page table@footnote{The denomination @dfn{grey} comes from the lexicon of @dfn{tri-color marking}, which is an abstraction of every possible garbage collection algorithm: in tri-color marking, grey objects are those that are known to be reachable or that we are not interested in reclaiming, yet have not been scanned to mark the objects that they refer to as reachable.} contains one entry for each page in OldSpace or FixedSpace that is thought to contain at least a reference to an object housed in NewSpace. Every page in OldSpace is created as grey, and is considered grey until a scavenging pass finds out that it actually does not contain pointers to NewSpace. Then the page is recolored black@footnote{Black objects are those that are known to be reachable or that we are not interested in reclaiming, and are known to have references only to other black or grey objects (in case you're curious, the tri-color marking algorithm goes on like this: object not yet known to be reachable are white, and when all objects are either black or white, the white ones are garbage).}, and will stay black until it is written to or another object is allocated in it (either a new fixed object, or a young object being tenured). The grey page table is expanded and shrunk as needed by the virtual machine. Drawing an histogram of object sizes shows that there are only a few sources of large objects on average (i.e., objects greater than a page in size), but that enough of these objects are created dynamically that they must be handled specially. Such objects should not be allocated in NewSpace along with ordinary objects, since they would fill up NewSpace prematurely (or might not even fit in it), thus accelerating the scavenging rate, reducing performance and resulting in an increase in tenured garbage. Even though this is not an optimal solution because it effectively tenures these objects at the time they are created, a benefit can be obtained by allocating these objects directly in FixedSpace. The reason why FixedSpace is used is that these objects are big enough that they don't result in fragmentation@footnote{Remember that free pages are shared among the three heaps, that is, OldSpace, FixedSpace and the @code{malloc} heap. When a large object is freed, the memory that it used can be reused by @code{malloc} or by OldSpace allocation}; and using FixedSpace instead of OldSpace avoids that the compactor copies them because this would not provide any benefit in terms of reduced fragmentation. Smalltalk activation records are allocated from another special heap, the context pool. This is because it is often the case that they can be deallocated in a Last-In-First-Out (stack) fashion, thereby saving the work needed to allocate entries in the object table for them, and quickly reusing the memory that they use. When the activation record is accessed by Smalltalk, however, the activation record must be turned into a first-class @code{OOP}@footnote{This is short for @dfn{Ordinary Object Pointer}.}. Since even these objects are usually very short-lived, the data is however not copied to the Eden: the eviction of the object bodies from the context pool is delayed to the next scavenging, which will also empty the context pool just like it empties Eden. If few objects are allocated and the context pool turns full before the Eden, a scavenging is also triggered; this is however quite rare. Optionally, @gst{} can avoid the overhead of interpretation by executing a given Smalltalk method only after that method has been compiled into the underlying microprocessor's machine code. This machine-code generation is performed automatically, and the resulting machine code is then placed in @code{malloc}-managed memory. Once executed, a method's machine code is left there for subsequent execution. However, since it would require way too much memory to permanently house the machine-code version of every Smalltalk method, methods might be compiled more than once: when a translation is not used at the time that two garbage collection actions are taken (scavenges and global garbage collections count equally), the incremental sweeper discards it, so that it will be recomputed if and when necessary. @node Security @section Security in @gst{} @node Special objects @section Special kinds of objects A few methods in Object support the creation of particular objects. This include: @itemize @bullet @item finalizable objects @item weak and ephemeron objects (i.e. objects whose contents are considered specially, during the heap scanning phase of garbage collection). @item read-only objects (like literals found in methods) @item fixed objects (guaranteed not to move across garbage collections) @end itemize They are: @defmethod Object makeWeak Marks the object so that it is considered weak in subsequent garbage collection passes. The garbage collector will consider dead an object which has references only inside weak objects, and will replace references to such an almost-dead'' object with nils, and then send the @code{mourn} message to the object. @end defmethod @defmethod Object makeEphemeron Marks the object so that it is considered specially in subsequent garbage collection passes. Ephemeron objects are sent the message @code{mourn} when the first instance variable is not referenced or is referenced @emph{only through another instance variable in the ephemeron}. Ephemerons provide a very versatile base on which complex interactions with the garbage collector can be programmed (for example, finalization which is described below is implemented with ephemerons). @end defmethod @defmethod Object addToBeFinalized Marks the object so that, as soon as it becomes unreferenced, its @code{finalize} method is called. Before @code{finalize} is called, the VM implicitly removes the objects from the list of finalizable ones. If necessary, the @code{finalize} method can mark again the object as finalizable, but by default finalization will only occur once. Note that a finalizable object is kept in memory even when it has no references, because tricky finalizers might resuscitate'' the object; automatic marking of the object as not to be finalized has the nice side effect that the VM can simply delay the releasing of the memory associated to the object, instead of being forced to waste memory even after finalization happens. An object must be explicitly marked as to be finalized @emph{every time the image is loaded}; that is, finalizability is not preserved by an image save. This was done because in most cases finalization is used together with operating system resources that would be stale when the image is loaded again. For @code{CObject}s, in particular, freeing them would cause a segmentation violation. @end defmethod @defmethod Object removeToBeFinalized Removes the to-be-finalized mark from the object. As I noted above, the finalize code for the object does not have to do this explicitly. @end defmethod @defmethod Object finalize This method is called by the VM when there are no more references to the object (or, of course, if it only has references inside weak objects). @end defmethod @defmethod Object isReadOnly This method answers whether the VM will refuse to make changes to the objects when methods like @code{become:}, @code{basicAt:put:}, and possibly @code{at:put:} too (depending on the implementation of the method). Note that @gst{} won't try to intercept assignments to fixed instance variables, nor assignments via @code{instVarAt:put:}. Many objects (Characters, @code{nil}, @code{true}, @code{false}, method literals) are read-only by default. @end defmethod @defmethod Object makeReadOnly: aBoolean Changes the read-only or read-write status of the receiver to that indicated by @code{aBoolean}. @end defmethod @defmethod Object basicNewInFixedSpace Same as @code{#basicNew}, but the object won't move across garbage collections. @end defmethod @defmethod Object basicNewInFixedSpace: Same as @code{#basicNew:}, but the object won't move across garbage collections. @end defmethod @defmethod Object makeFixed Ensure that the receiver won't move across garbage collections. This can be used either if you decide after its creation that an object must be fixed, or if a class does not support using @code{#new} or @code{#new:} to create an object @end defmethod Note that, although particular applications will indeed have a need for fixed, read-only or finalizable objects, the @code{#makeWeak} primitive is seldom needed and weak objects are normally used only indirectly, through the so called @dfn{weak collections}. These are easier to use because they provide additional functionality (for example, @code{WeakArray} is able to determine whether an item has been garbage collected, and @code{WeakSet} implements hash table functionality); they are: @itemize @bullet @bulletize @code{WeakArray} @bulletize @code{WeakSet} @bulletize @code{WeakKeyDictionary} @bulletize @code{WeakValueLookupTable} @bulletize @code{WeakIdentitySet} @bulletize @code{WeakKeyIdentityDictionary} @bulletize @code{WeakValueIdentityDictionary} @end itemize Versions of @gst{} preceding 2.1 included a @code{WeakKeyLookupTable} class which has been replaced by @code{WeakKeyDictionary}; the usage is completely identical, but the implementation was changed to use a more efficient approach based on ephemeron objects. @node Packages @chapter Packages @gst{} includes a packaging system which allows one to file in components (often called @dfn{goodies} in Smalltalk lore) without caring of whether they need other goodies to be loaded first. The packaging system is implemented by a Smalltalk class, @code{PackageLoader}, which looks for information about packages in various places: @itemize @item the kernel directory's parent directory; this is where an installed @file{packages.xml} resides, in a system-wide data directory such as @file{/usr/local/share/smalltalk}; @item the above directory's @file{site-packages} subdirectory, for example @file{/usr/local/share/smalltalk/site-packages}; @item in the file @file{.st/packages.xml}, hosting per-user packages; @item finally, there can be a @file{packages.xml} in the same directory as the current image. @end itemize Each of this directories can contain package descriptions in an XML file named (guess what) @file{packages.xml}, as well as standalone packages in files named @file{*.star} (short for @cite{Smalltalk archive}). Later in this section you will find information about @command{gst-package}, a program that helps you create @file{.star} files. There are two ways to load something using the packaging system. The first way is to use the PackageLoader's @code{fileInPackage:} and @code{fileInPackages:} methods. For example: @example PackageLoader fileInPackages: #('DBD-MySQL' 'DBD-SQLite'). PackageLoader fileInPackage: 'Sockets'. @end example The second way is to use the @file{gst-load} script which is installed together with the virtual machine. For example, you can do: @t{@ @ @ @ gst-load DBD-MySQL DBD-SQLite DBI} @noindent and @gst{} will automatically file in: @itemize @bullet @bulletize DBI, loaded first because it is needed by the other two packages @bulletize Sockets and Digest, not specified, but needed by DBD-MySQL @bulletize DBD-MySQL @bulletize DBD-SQLite @end itemize @noindent Notice how DBI has already been loaded. Then it will save the Smalltalk image, and finally exit. @file{gst-load} supports several options: @table @option @item -I @itemx --image-file Load the packages inside the given image. @item -i @itemx --rebuild-image Build an image from scratch and load the package into it. Useful when the image specified with @option{-I} does not exist yet. @item -q @itemx --quiet Hide the script's output. @item -v @itemx --verbose Show which files are loaded, one by one. @item -f @itemx --force If a package given on the command-line is already present, reload it. This does not apply to automatically selected prerequisites. @item -t @itemx --test Run the package testsuite before installing, and exit with a failure if the tests fail. Currently, the testsuites are placed in the image together with the package, but this may change in future versions. @item -n @item --dry-run Do not save the image after loading. @item --start[=ARG] Start the services identified by the package. If an argument is given, only one package can be specified on the command-line. If at least one package specifies a startup script, @code{gst-load} won't exit. @end table To provide support for this system, you have to give away with your @gst{} goodies a small file (usually called @file{package.xml}) which looks like this: @example DBD-SQLite DBI.SQLite DBI dbd-sqlite3 DBI.SQLite.SQLiteTestSuite SQLiteTests.st SQLite.st Connection.st ResultSet.st Statement.st Row.st ColumnInfo.st Table.st TableColumnInfo.st SQLiteTests.st ChangeLog @end example Other tags exist: @table @code @item url Specifies a URL at which a repository for the package can be found. The repository, when checked out, should contain a @file{package.xml} file at its root. The contents of this tag are not used for local packages; they are used when using the @option{--download} option to @command{gst-package}. @item library Loads a dynamic shared object and registers the functions in it so that they can all be called from Smalltalk code. The @code{GTK} package registers the GTK+ library in this way, so that the bindings can use them. @item callout Instructs to load the package only if the C function whose name is within the tag is available to be called from Smalltalk code. @item start Specifies a Smalltalk script that @file{gst-load} and @file{gst-remote} will execute in order to start the execution of the service implemented in the package. Before executing the script, @code{%1} is replaced with either @code{nil} or a String literal. @item stop Specifies a Smalltalk script that @file{gst-remote} will execute in order to shut down the service implemented in the package. Before executing the script, @code{%1} is replaced with either @code{nil} or a String literal. @item dir Should include a @code{name} attribute. The @code{file}, @code{filein} and @code{built-file} tags that are nested within a @code{dir} tag are prepended with the directory specified by the attribute. @item test Specifies a subpackage that is only loaded by @file{gst-sunit} in order to test the package. The subpackage may include arbitrary tags (including @code{file}, @code{filein} and @code{sunit}) but not @code{name}. @item provides In some cases, a single functionality can be provided by multiple modules. For example, @gst{} includes two browsers but only one should be loaded at any time. To this end, a dummy package @code{Browser} is created pointing to the default browser (@code{VisualGST}), but both browsers use @code{provides} so that if the old BLOX browser is in the image, loading @code{Browser} will have no effect. @end table To install your package, you only have to do @example gst-package path/to/package.xml @end example @command{gst-package} is a Smalltalk script which will create a @file{.star} archive in the current image directory, with the files specified in the @code{file}, @code{filein} and @code{built-file} tags. By default the package is placed in the system-wide package directory; you can use the option @option{--target-directory} to create the @file{.star} file elsewhere. Instead of a local @file{package.xml} file, you can give: @itemize @bullet @item a local @file{.star} file or a @code{URL} to such a file. The file will be downloaded if necessary, and copied to the target directory; @item a URL to a @file{package.xml} file. The @code{url} tag in the file will be used to find a source code repository (@command{git} or @command{svn}) or as a redirect to another @file{package.xml} file. @end itemize @noindent There is also a short form for specifying @file{package.xml} file on @gst{}'s web site, so that the following two commands are equivalent: @example gst-package http://smalltalk.gnu.org/project/Iliad/package.xml gst-package --download Iliad @end example When downloading remote @file{package.xml} files, @command{gst-package} also performs a special check to detect multiple packages in the same repository. If the following conditions are met: @itemize @bullet @item a package named @code{@var{package}} has a prerequisite @code{@var{package}-@var{subpackage}}; @item there is a toplevel subdirectory @var{subpackage} in the repository; @item the subdirectory has a @file{package.xml} file in it @end itemize @noindent then the @file{@var{subpackage}/package.xml} will be installed as well. @command{gst-package} does not check if the file actually defines a package with the correct name, but this may change in future versions. Alternatively, @code{gst-package} can be used to create a skeleton @gnu{} style source tree. This includes a @file{configure.ac} that will find the installation path of @gst{}, and a @file{Makefile.am} to support all the standard Makefile targets (including @command{make install} and @command{make dist}). To do so, go in the directory that is to become the top of the source tree and type. @example gst-package --prepare path1/package.xml path2/package.xml @end example In this case the generated configure script and Makefile will use more features of @command{gst-package}, which are yet to be documented. The @gst{} makefile similarly uses @command{gst-package} to install packages and to prepare the distribution tarballs. The rest of this chapter discusses some of the packages provided with @gst{}. @menu * GTK and VisualGST: GUI. * Parser, STInST, Compiler: Smalltalk-in-Smalltalk. * DBI: Database. * I18N: Locales. * Seaside: Seaside. * Swazoo: Swazoo. * SUnit: SUnit. * Sockets, WebServer, NetClients: Network support. * XML, XPath, XSL: XML. * Other packages: Other packages. @end menu @node GUI @section GTK and VisualGST @gst{} comes with GTK bindings and with a browser based on it. The system can be started as @command{gst-browser} and will allow the programmer to view the source code for existing classes, to modify existing classes and methods, to get detailed information about the classes and methods, and to evaluate code within the browser. In addition, simple debugging and unit testing tools are provided. An Inspector window allows the programmer to graphically inspect and modify the representation of an object and a walkback inspector was designed which will display a backtrace when the program encounters an error. SUnit tests (@pxref{SUnit}) can be run from the browser in order to easily support test driven development. The Transcript global object is redirected to print to the transcript window instead of printing to stdout, and the transcript window as well as the workspaces, unlike the console read-eval-print loop, support variables that live across multiple evaluations: @example a := 2 "Do-it" a + 2 "Print-it: 4 will be shown" @end example To start the browser you can simply type: @example gst-browser @end example This will load any requested packages, then, if all goes well, a @emph{launcher} window combining all the basic tools will appear on your display. @node Smalltalk-in-Smalltalk @section The Smalltalk-in-Smalltalk library The Smalltalk-in-Smalltalk library is a set of classes for looking at Smalltalk code, constructing models of Smalltalk classes that can later be created for real, analyzing and performing changes to the image, finding smelly code and automatically doing repetitive changes. This package incredibly enhances the reflective capabilities of Smalltalk. @ignore Being quite big (20000 source code lines) this package is split into three different packages: @code{Parser} loads the parser only, @code{STInST} loads various other tools (which compose the Refactoring Browser'' package by John Brant and Don Roberts and will be the foundation for @gst{}'s next generation browser), @code{STInSTTest} performs comprehensive unit tests@footnote{ The tests can take @strong{hours} to complete!} (@pxref{SUnit}). Porting of the @code{STInST} package will be completed in @gst{} 2.2. @end ignore A fundamental part of the system is the recursive-descent parser which creates parse nodes in the form of instances of subclasses of @code{RBProgramNode}. The parser's extreme flexibility can be exploited in three ways, all of which are demonstrated by source code available in the distribution: @itemize @bullet @item First, actions are not hard-coded in the parser itself: the parser creates a parse tree, then hands it to methods in @code{RBParser} that can be overridden in different @code{RBParser} subclasses. This is done by the compiler itself, in which a subclass of @code{RBParser} (class @code{STFileInParser}) hands the parse trees to the @code{STCompiler} class. @item Second, an implementation of the visitor'' pattern is provided to help in dealing with parse trees created along the way; this approach is demonstrated by the Smalltalk code pretty-printer in class @code{RBFormatter}, by the syntax highlighting engine included with the browser, and by the compiler. @item The parser is able to perform complex tree searches and rewrites, through the ParseTreeSearcher and ParseTreeRewriter classes. @ignore This mechanism is exploited by most of the tools loaded by the @code{STInST} package. @end ignore @end itemize In addition, two applications were created on top of this library which are specific to @gst{}. The first is a compiler for Smalltalk methods written in Smalltalk itself, whose source code provides good insights into the @gst{} virtual machine. The second is the automatic documentation extractor. @code{gst-doc} is able to create documentation even if the library cannot be loaded (for example, if loading it requires a running X server). To do so it uses @code{STClassLoader} from the @file{Parser} package to load and interpret Smalltalk source code, creating objects for the classes and methods being read in; then, polymorphism allows one to treat these exactly like usual classes. @node Database @section Database connectivity @gst{} includes support for connecting to databases. Currently this support is limited to retrieving result sets from @acronym{SQL} selection queries and executing @acronym{SQL} data manipulation queries; in the future however a full object model will be available that hides the usage of @acronym{SQL}. Classes that are independent of the database management system that is in use reside in package @code{DBI}, while the drivers proper reside in separate packages which have @code{DBI} as a prerequisite; currently, drivers are supplied for @emph{MySQL} and @emph{PostgreSQL}, in packages @code{DBD-MySQL} and @code{DBD-PostgreSQL} respectively. Using the library is fairly simple. To execute a query you need to create a connection to the database, create a statement on the connection, and execute your query. For example, let's say I want to connect to the @file{test} database on the localhost. My user name is @code{doe} and my password is @code{mypass}. @example | connection statement result | connection := DBI.Connection connect: 'dbi:MySQL:dbname=test;hostname=localhost' user: 'doe' password: 'mypass'). @end example You can see that the @acronym{DBMS}-specific classes live in a sub-namespace of @code{DBI}, while @acronym{DBMS}-independent classes live in @code{DBI}. Here is how I execute a query. @example statement := connection execute: 'insert into aTable (aField) values (123)'. @end example The result that is returned is a @code{ResultSet}. For write queries the object returns the number of rows affected. For read queries (such as selection queries) the result set supports standard stream protocol (@code{next}, @code{atEnd} to read rows off the result stream) and can also supply collection of column information. These are instances of @code{ColumnInfo}) and describe the type, size, and other characteristics of the returned column. A common usage of a ResultSet would be: @example | resultSet values | [resultSet atEnd] whileFalse: [values add: (resultSet next at: 'columnName') ]. @end example @node Locales @section Internationalization and localization support Different countries and cultures have varying conventions for how to communicate. These conventions range from very simple ones, such as the format for representing dates and times, to very complex ones, such as the language spoken. Provided the programs are written to obey the choice of conventions, they will follow the conventions preferred by the user. @gst{} provides two packages to ease you in doing so. The @code{I18N} package covers both @dfn{internationalization} and @dfn{multilingualization}; the lighter-weight @code{Iconv} package covers only the latter, as it is a prerequisite for correct internationalization. @dfn{Multilingualizing} software means programming it to be able to support languages from every part of the world. In particular, it includes understanding multi-byte character sets (such as UTF-8) and Unicode characters whose @dfn{code point} (the equivalent of the ASCII value) is above 127. To this end, @gst{} provides the @code{UnicodeString} class that stores its data as 32-bit Unicode values. In addition, @code{Character} will provide support for all the over one million available code points in Unicode. Loading the @code{I18N} package improves this support through the @code{EncodedStream} class@footnote{All the classes mentioned in this section reside in the @code{I18N} namespace.}, which interprets and transcodes non-ASCII Unicode characters. This support is mostly transparent, because the base classes @code{Character}, @code{UnicodeCharacter} and @code{UnicodeString} are enhanced to use it. Sending @code{asString} or @code{printString} to an instance of @code{Character} and @code{UnicodeString} will convert Unicode characters so that they are printed correctly in the current locale. For example, @samp{$<279> printNl} will print a small Latin letter @samp{e} with a dot above, when the @code{I18N} package is loaded. Dually, you can convert @code{String} or @code{ByteArray} objects to Unicode with a single method call. If the current locale's encoding is UTF-8, @samp{#[196 151] asUnicodeString} will return a Unicode string with the same character as above, the small Latin letter @samp{e} with a dot above. The implementation of multilingualization support is not yet complete. For example, methods such as @code{asLowercase}, @code{asUppercase}, @code{isLetter} do not yet recognize Unicode characters. You need to exercise some care, or your program will be buggy when Unicode characters are used. In particular, Characters must @strong{not} be compared with @code{==}@footnote{Character equality with @code{=} will be as fast as with @code{==}.} and should be printed on a Stream with @code{display:} rather than @code{nextPut:}. Also, Characters need to be created with the class method @code{codePoint:} if you are referring to their Unicode value; @code{codePoint:} is also the only method to create characters that is accepted by the ANSI Standard for Smalltalk. The method @code{value:}, instead, should be used if you are referring to a byte in a particular encoding. This subtle difference means that, for example, the last two of the following examples will fail: @example "Correct. Use #value: with Strings, #codePoint: with UnicodeString." String with: (Character value: 65) String with: (Character value: 128) UnicodeString with: (Character codePoint: 65) UnicodeString with: (Character codePoint: 128) "Correct. Only works for characters in the 0-127 range, which may be considered as defensive programming." String with: (Character codePoint: 65) "Dubious, and only works for characters in the 0-127 range. With UnicodeString, probably you always want #codePoint:." UnicodeString with: (Character value: 65) "Fails, we try to use a high character in a String" String with: (Character codePoint: 128) "Fails, we try to use an encoding in a Unicode string" UnicodeString with: (Character value: 128) @end example @dfn{Internationalizing} software, instead, means programming it to be able to adapt to the user's favorite conventions. These conventions can get pretty complex; for example, the user might specify the locale espana-castellano' for most purposes, but specify the locale usa-english' for currency formatting: this might make sense if the user is a Spanish-speaking American, working in Spanish, but representing monetary amounts in US dollars. You can see that this system is simple but, at the same time, very complete. This manual, however, is not the right place for a thorough discussion of how an user would set up his system for these conventions; for more information, refer to your operating system's manual or to the @gnu{} C library's manual. @gst{} inherits from @sc{iso} C the concept of a @dfn{locale}, that is, a collection of conventions, one convention for each purpose, and maps each of these purposes to a Smalltalk class defined by the @code{I18N} package, and these classes form a small hierarchy with class @code{Locale} as its roots: @itemize @bullet @ignore @item @code{LcCollate} defines the collating sequence for the local language and character set. @end ignore @item @code{LcNumeric} formats numbers; @code{LcMonetary} and @code{LcMonetaryISO} format currency amounts. @item @code{LcTime} formats dates and times. @item @code{LcMessages} translates your program's output. Of course, the package can't automatically translate your program's output messages into other languages; the only way you can support output in the user's favorite language is to translate these messages by hand. The package does, though, provide methods to easily handle translations into multiple languages. @end itemize Basic usage of the @code{I18N} package involves a single selector, the question mark (@code{?}), which is a rarely used yet valid character for a Smalltalk binary message. The meaning of the question mark selector is How do you say @dots{} under your convention?''. You can send @code{?} to either a specific instance of a subclass of @code{Locale}, or to the class itself; in this case, rules for the default locale (which is specified via environment variables) apply. You might say, for example, @code{LcTime ? Date today} or, for example, @code{germanMonetaryLocale ? account balance}. This syntax can be at first confusing, but turns out to be convenient because of its consistency and overall simplicity. Here is how @code{?} works for different classes: @ignore @defmethod LcCollate ? aString Answer an instance of LcCollationKey; code like @code{LcCollate ? string1 < string2} will compare the two strings under the rules of the default locale @end defmethod @end ignore @defmethod LcTime ? aString Format a date, a time or a timestamp (@code{DateTime} object). @end defmethod @defmethod LcNumber ? aString Format a number. @end defmethod @defmethod LcMonetary ? aString Format a monetary value together with its currency symbol. @end defmethod @defmethod LcMonetaryISO ? aString Format a monetary value together with its @sc{iso} currency symbol. @end defmethod @defmethod LcMessages ? aString Answer an @code{LcMessagesDomain} that retrieves translations from the specified file. @end defmethod @defmethod LcMessagesDomain ? aString Retrieve the translation of the given string.@footnote{The @code{?} method does not apply to the LcMessagesDomain class itself, but only to its instances. This is because LcMessagesDomain is not a subclass of Locale.} @end defmethod These two packages provides much more functionality, including more advanced formatting options support for Unicode, and conversion to and from several character sets. For more information, refer to @ref{I18N, , Multilingual and international support with Iconv and I18N, gst-libs, the @gst{} Library Reference}. As an aside, the representation of locales that the package uses is exactly the same as the C library, which has many advantages: the burden of mantaining locale data is removed from @gst{}'s mantainers; the need of having two copies of the same data is removed from @gst{}'s users; and finally, uniformity of the conventions assumed by different internationalized programs is guaranteed to the end user. In addition, the representation of translated strings is the standard @sc{mo} file format adopted by the @gnu{} @code{gettext} library. @node Seaside @section The Seaside web framework Seaside is a framework to build highly interactive web applications quickly, reusably and maintainably. Features of Seaside include callback-based request handling, hierarchical (component-based) page design, and modal session management to easily implement complex workflows. A simple Seaside component looks like this: @example Seaside.WAComponent subclass: MyCounter [ | count | MyCounter class >> canBeRoot [ ^true ] initialize [ super initialize. count := 0. ] states [ ^@{ self @} ] renderContentOn: html [ html heading: count. html anchor callback: [ count := count + 1 ]; with: '++'. html space. html anchor callback: [ count := count - 1 ]; with: '--'. ] ] MyCounter registerAsApplication: 'mycounter' @end example Most of the time, you will run Seaside in a background virtual machine. First of all, you should load the Seaside packages into a new image like this: @example $gst-load -iI seaside.im Seaside Seaside-Development Seaside-Examples @end example @noindent Then, you can start Seaside with either of these commands @example$ gst-load -I seaside.im --start Seaside $gst-remote -I seaside.im --daemon --start=Seaside @end example @noindent which will start serving pages at @url{http://localhost:8080/seaside}. The former starts the server in foreground, the latter instead runs a virtual machine that you can control using further invocations of @command{gst-remote}. For example, you can stop serving Seaside pages, and bring down the server, respectively with these commands: @example$ gst-remote --stop=Seaside $gst-remote --kill @end example @node Swazoo @section The Swazoo web server Swazoo (Smalltalk Web Application Zoo) is a free Smalltalk HTTP server supporting both static web serving and a fully-featured web request resolution framework. The server can be started using @example$ gst-load --start@i{[=@var{ARG}]} Swazoo @end example @noindent or loaded into a background @gst{} virtual machine with @example $gst-remote --start=Swazoo@i{[:@var{ARG}]} @end example Usually, the first time you start Swazoo @var{ARG} is @code{swazoodemo} (which starts a simple Hello, World!'' servlet) or a path to a configuration file like this one: @example @end example After this initial step, @var{ARG} can take the following meanings: @itemize @bullet @item if omitted altogether, all the sites registered on the server are started; @item if a number, all the sites registered on the server on that port are started; @item if a configuration file name, the server configuration is @emph{replaced} with the one loaded from that file; @item if any other string, the site named @var{ARG} is started. @end itemize In addition, a background server can be stopped using @example$ gst-remote --stop=Swazoo@i{[:@var{ARG}]} @end example @noindent where @var{ARG} can have the same meanings, except for being a configuration file. In addition, package @code{WebServer} implements an older web server engine which is now superseded by Swazoo. It is based on the @sc{gpl}'ed WikiWorks project. Apart from porting to @gst{}, a number of changes were made to the code, including refactoring of classes, better aesthetics, authentication support, virtual hosting, and @sc{http} 1.1 compliance. @node SUnit @section The SUnit testing package @code{SUnit} is a framework to write and perform test cases in Smalltalk, originarily written by the father of Extreme Programming@footnote{Extreme Programming is a software engineering technique that focuses on team work (to the point that a programmer looks in real-time at what another one is typing), frequent testing of the program, and incremental design.}, Kent Beck. @code{SUnit} allows one to write the tests and check results in Smalltalk; while this approach has the disadvantage that testers need to be able to write simple Smalltalk programs, the resulting tests are very stable. What follows is a description of the philosophy of @code{SUnit} and a description of its usage, excerpted from Kent Beck's paper in which he describes @code{SUnit}. @subsection Where should you start? Testing is one of those impossible tasks. You'd like to be absolutely complete, so you can be sure the software will work. On the other hand, the number of possible states of your program is so large that you can't possibly test all combinations. If you start with a vague idea of what you'll be testing, you'll never get started. Far better to @emph{start with a single configuration whose behavior is predictable}. As you get more experience with your software, you will be able to add to the list of configurations. Such a configuration is called a @dfn{fixture}. Two example fixtures for testing Floats can be @code{1.0} and @code{2.0}; two fixtures for testing Arrays can be @code{#()} and @code{#(1 2 3)}. By choosing a fixture you are saying what you will and won't test for. A complete set of tests for a community of objects will have many fixtures, each of which will be tested many ways. To design a test fixture you have to @itemize @bulletize{Subclass TestCase} @bulletize{Add an instance variable for each known object in the fixture} @bulletize{Override setUp to initialize the variables} @end itemize @subsection How do you represent a single unit of testing? You can predict the results of sending a message to a fixture. You need to represent such a predictable situation somehow. The simplest way to represent this is interactively. You open an Inspector on your fixture and you start sending it messages. There are two drawbacks to this method. First, you keep sending messages to the same fixture. If a test happens to mess that object up, all subsequent tests will fail, even though the code may be correct. More importantly, though, you can't easily communicate interactive tests to others. If you give someone else your objects, the only way they have of testing them is to have you come and inspect them. By representing each predictable situation as an object, each with its own fixture, no two tests will ever interfere. Also, you can easily give tests to others to run. @emph{Represent a predictable reaction of a fixture as a method.} Add a method to TestCase subclass, and stimulate the fixture in the method. @subsection How do you test for expected results? If you're testing interactively, you check for expected results directly, by printing and inspecting your objects. Since tests are in their own objects, you need a way to programmatically look for problems. One way to accomplish this is to use the standard error handling mechanism (@code{#error:}) with testing logic to signal errors: @example 2 + 3 = 5 ifFalse: [self error: 'Wrong answer'] @end example When you're testing, you'd like to distinguish between errors you are checking for, like getting six as the sum of two and three, and errors you didn't anticipate, like subscripts being out of bounds or messages not being understood. There's not a lot you can do about unanticipated errors (if you did something about them, they wouldn't be unanticipated any more, would they?) When a catastrophic error occurs, the framework stops running the test case, records the error, and runs the next test case. Since each test case has its own fixture, the error in the previous case will not affect the next. The testing framework makes checking for expected values simple by providing a method, @code{#should:}, that takes a Block as an argument. If the Block evaluates to true, everything is fine. Otherwise, the test case stops running, the failure is recorded, and the next test case runs. So, you have to @emph{turn checks into a Block evaluating to a Boolean, and send the Block as the parameter to @code{#should:}}. In the example, after stimulating the fixture by adding an object to an empty Set, we want to check and make sure it's in there: @example SetTestCase>>#testAdd empty add: 5. self should: [empty includes: 5] @end example There is a variant on @code{TestCase>>#should:}. @code{TestCase>>#shouldnt:} causes the test case to fail if the Block argument evaluates to true. It is there so you don't have to use @code{(...) not}. Once you have a test case this far, you can run it. Create an instance of your TestCase subclass, giving it the selector of the testing method. Send @code{run} to the resulting object: @example (SetTestCase selector: #testAdd) run @end example If it runs to completion, the test worked. If you get a walkback, something went wrong. @subsection How do you collect and run many different test cases? As soon as you have two test cases running, you'll want to run them both one after the other without having to execute two do it's. You could just string together a bunch of expressions to create and run test cases. However, when you then wanted to run this bunch of cases and that bunch of cases'' you'd be stuck. The testing framework provides an object to represent @dfn{a bunch of tests}, @code{TestSuite}. A @code{TestSuite} runs a collection of test cases and reports their results all at once. Taking advantage of polymorphism, @code{TestSuites} can also contain other @code{TestSuites}, so you can put Joe's tests and Tammy's tests together by creating a higher level suite. @emph{Combine test cases into a test suite.} @example (TestSuite named: 'Money') add: (MoneyTestCase selector: #testAdd); add: (MoneyTestCase selector: #testSubtract); run @end example The result of sending @code{#run} to a @code{TestSuite} is a @code{TestResult} object. It records all the test cases that caused failures or errors, and the time at which the suite was run. All of these objects are suitable for being stored in the image and retrieved. You can easily store a suite, then bring it in and run it, comparing results with previous runs. @subsection Running testsuites from the command line @gst{} includes a Smalltalk script to simplify running SUnit test suites. It is called @command{gst-sunit}. The command-line to @command{gst-sunit} specifies the packages, files and classes to test: @table @option @item -I @itemx --image-file Run tests inside the given image. @item -q @itemx --quiet Hide the program's output. The results are still communicated with the program's exit code. @item -v @itemx --verbose Be more verbose, in particular this will cause @command{gst-sunit} to write which test is currently being executed. @item -f @var{FILE} @itemx --file=@var{FILE} Load @var{FILE} before running the required test cases. @item -p @var{PACKAGE} @item --package=@var{PACKAGE} Load @var{PACKAGE} and its dependencies, and add @var{PACKAGE}'s tests to the set of test cases to run. @item @var{CLASS} @itemx @var{CLASS}* Add @var{CLASS} to the set of test cases to run. An asterisk after the class name adds all the classes in @var{CLASS}'s hierarchy. In particular, each selector whose name starts with @code{test} constitutes a separate test case. @item @var{VAR}=@var{VALUE} Associate variable @var{VAR} with a value. Variables allow customization of the testing environment. For example, the username with which to access a database can be specified with variables. From within a test, variables are accessible with code like this: @example TestSuitesScripter variableAt: 'mysqluser' ifAbsent: [ 'root' ] @end example Note that a @code{#variableAt:} variant does @emph{not} exist, because the testsuite should pick default values in case the variables are not specified by the user. @end table @node Network support @section Sockets, WebServer, NetClients @gst{} includes an almost complete abstraction of the @sc{tcp}, @sc{udp} and @sc{ip} protocols. Although based on the standard @sc{bsd} sockets, this library provides facilities such as buffering and preemptive I/O which a C programmer usually has to implement manually. The distribution includes a few tests (mostly loopback tests that demonstrate both client and server connection), which are class methods in @code{Socket}. This code should guide you in the process of creating and using both server and client sockets; after creation, sockets behave practically the same as standard Smalltalk streams, so you should not have particular problems. For more information, refer to @ref{Sockets, , Network programming with Sockets, gst-libs, the @gst{} Library Reference}. The library is also used by many other packages, including Swazoo and the MySQL driver. There is also code implementing the most popular Internet protocols: @sc{ftp}, @sc{http}, @sc{nntp}, @sc{smtp}, @sc{pop3} and @sc{imap}. These classes, loaded by the @code{NetClients} package, are derived from multiple public domain and free software packages available for other Smalltalk dialects and ported to @gst{}. Future version of @gst{} will include documentation for these as well. @node XML @section An XML parser and object model for @gst{} The @sc{xml} parser library for Smalltalk, loaded as package @code{XML} includes a validating @sc{xml} parser and Document Object Model. This library is rapidly becoming a standard in the Smalltalk world and a @sc{xslr} interpreter based on it is bundled with @gst{} as well (see packages @code{XPath} and @code{XSL}). Parts of the basic XML package can be loaded independently using packages @code{XML-DOM}, @code{XML-SAXParser}, @code{XML-XMLParser}, @code{XML-SAXDriver}, @code{XML-XMLNodeBuilder}. @node Other packages @section Other packages Various other minor'' packages are provided, typically as examples of writing modules for @gst{} (@pxref{External modules, , Linking your libraries to the virtual machine}). These include: @table @i @item Complex which adds transparent operations with complex numbers @item @sc{gdbm} which is an interface to the @gnu{} database manager @item Digest which provides two easy to use classes to quickly compute cryptographically strong hash values using the MD5 and SHA1 algorithms. @item NCurses which provides bindings to @i{ncurses} @item Continuations which provides more examples and tests for continuations (an advanced feature to support complex control flow). @item DebugTools which provides a way to attach to another Smalltalk process and execute it a bytecode or a method at a time. @end table @node Emacs @chapter Smalltalk interface for @gnu{} Emacs @gst{} comes with its own Emacs mode for hacking Smalltalk code. It also provides tools for interacting with a running Smalltalk system in an Emacs subwindow. Emacs will automatically go into Smalltalk mode when you edit a Smalltalk file (one with the extension @file{.st}). @menu * Editing:: Autoindent and more for @gst{}. * Interactor:: Smalltalk interactor mode. @end menu @node Editing @section Smalltalk editing mode The @gst{} editing mode is there to assist you in editing your Smalltalk code. It tries to be smart about indentation and provides a few cooked templates to save you keystrokes. Since Smalltalk syntax is highly context sensitive, the Smalltalk editing mode will occasionally get confused when you are editing expressions instead of method definitions. In particular, using local variables, thus: @example | foo | foo := 3. ^foo squared ! @end example @noindent will confuse the Smalltalk editing mode, as this might also be a definition the binary operator @code{|}, with second argument called @samp{foo}. If you find yourself confused when editing this type of expression, put a dummy method name before the start of the expression, and take it out when you're done editing, thus: @example x | foo | foo := 3. ^foo squared ! @end example @node Interactor @section Smalltalk interactor mode An interesting feature of Emacs Smalltalk is the Smalltalk interactor, which basically allows you run in @gnu{} Emacs with Smalltalk files in one window, and Smalltalk in the other. You can, with a single command, edit and change method definitions in the live Smalltalk system, evaluate expressions, make image snapshots of the system so you can pick up where you left off, file in an entire Smalltalk file, etc. It makes a tremendous difference in the productivity and enjoyment that you'll have when using @gst{}. To start up the Smalltalk interactor, you must be running @gnu{} Emacs and in a buffer that's in Smalltalk mode. Then, if you type @kbd{C-c m}. A second window will appear with @gst{} running in it. This window is in most respects like a Shell mode window. You can type Smalltalk expressions to it directly and re-execute previous things in the window by moving the cursor back to the line that contains the expression that you wish to re-execute and typing return. Notice the status in the mode line (e.g. @samp{starting-up}, @samp{idle}, etc). This status will change when you issue various commands from Smalltalk mode. When you first fire up the Smalltalk interactor, it puts you in the window in which Smalltalk is running. You'll want to switch back to the window with your file in it to explore the rest of the interactor mode, so do it now. To execute a range of code, mark the region around and type @kbd{C-c e}. The expression in the region is sent to Smalltalk and evaluated. The status will change to indicate that the expression is executing. This will work for any region that you create. If the region does not end with an exclamation point (which is syntactically required by Smalltalk), one will be added for you. There is also a shortcut, @kbd{C-c d} (also invokeable as @kbd{M-x smalltalk-doit}), which uses a simple heuristic to figure out the start and end of the expression: it searches forward for a line that begins with an exclamation point, and backward for a line that does not begin with space, tab, or the comment character, and sends all the text in between to Smalltalk. If you provide a prefix argument (by typing @kbd{C-u C-c d} for instance), it will bypass the heuristic and use the region instead (just like @kbd{C-c e} does). @kbd{C-c c} will compile a method; it uses a similar heuristic to determine the bounds of the method definition. Typically, you'll change a method definition, type @kbd{C-c c} and move on to whatever's next. If you want to compile a whole bunch of method definitions, you'll have to mark the entire set of method definitions (from the @code{methodsFor:} line to the @code{! !}) as the region and use @kbd{C-c e}. After you've compiled and executed some expressions, you may want to take a snapshot of your work so that you don't have to re-do things next time you fire up Smalltalk. To do this, you use the @kbd{C-c s} command, which invokes @code{ObjectMemory snapshot}. If you invoke this command with a prefix argument, you can specify a different name for the image file, and you can have that image file loaded instead of the default one by using the @code{-I} flag on the command line when invoking Smalltalk. You can also evaluate an expression and have the result of the evaluation printed by using the @kbd{C-c p} command. Mark the region and use the command. To file in an entire file (perhaps the one that you currently have in the buffer that you are working on), type @kbd{C-c f}. You can type the name of a file to load at the prompt, or just type return and the file associated with the current buffer will be loaded into Smalltalk. When you're ready to quit using @gst{}, you can quit cleanly by using the @kbd{C-c q} command. If you want to fire up Smalltalk again, or if (heaven forbid) Smalltalk dies on you, you can use the @kbd{C-c m} command, and Smalltalk will be reincarnated. Even if it's running, but the Smalltalk window is not visible, @kbd{C-c m} will cause it to be displayed right away. You might notice that as you use this mode, the Smalltalk window will scroll to keep the bottom of the buffer in focus, even when the Smalltalk window is not the current window. This was a design choice that I made to see how it would work. On the whole, I guess I'm pretty happy with it, but I am interested in hearing your opinions on the subject. @node C and Smalltalk @chapter Interoperability between C and @gst{} @menu * External modules:: Linking your libraries to the virtual machine * C callout:: Calls from Smalltalk to C * C data types:: Manipulating C data from Smalltalk * Smalltalk types:: Manipulating Smalltalk data from C * Smalltalk callin:: Calls from C to Smalltalk * Smalltalk callbacks:: Smalltalk blocks as C function pointers * Object representation:: Manipulating your own Smalltalk objects * Incubator:: Protecting newly created objects from garbage collections * Other C functions:: Handling and creating OOPs * Using Smalltalk:: The Smalltalk environment as an extension library @end menu @node External modules @section Linking your libraries to the virtual machine A nice thing you can do with @gst{} is enhancing it with your own goodies. If they're written in Smalltalk only, no problem: getting them to work as packages (@pxref{Packages}), and to fit in with the @gst{} packaging system, is likely to be a five-minutes task. If your goodie is creating a binding to an external C library and you do not need particular glue to link it to Smalltalk (for example, there are no callbacks from C code to Smalltalk code), you can use the @code{dynamic library linking} system. When using this system, you have to link @gst{} with the library at run-time using @sc{dld}, using either @code{DLD class>>#addLibrary:} or a @code{} tag in a @file{package.xml} file (@pxref{Packages}). The following line: @example DLD addLibrary: 'libc' @end example @noindent is often used to use the standard C library functions from Smalltalk. However, if you want to provide a more intimate link between C and Smalltalk, as is the case with for example the GTK bindings, you should use the @code{dynamic module linking} system. This section explains what to do, taking the Digest library as a guide. A module is distinguished from a standard shared library because it has a function which Smalltalk calls to initialize the module; the name of this function must be @code{gst_initModule}. Here is the initialization function used by Digest: @example void gst_initModule(proxy) VMProxy *proxy; @{ vmProxy = proxy; vmProxy->defineCFunc ("MD5AllocOOP", MD5AllocOOP); vmProxy->defineCFunc ("MD5Update", md5_process_bytes); vmProxy->defineCFunc ("MD5Final", md5_finish_ctx); vmProxy->defineCFunc ("SHA1AllocOOP", SHA1AllocOOP); vmProxy->defineCFunc ("SHA1Update", sha1_process_bytes); vmProxy->defineCFunc ("SHA1Final", sha1_finish_ctx); @} @end example Note that the @code{defineCFunc} function is called through a function pointer in @code{gst_initModule}, and that the value of its parameter is saved in order to use it elsewhere in its code. This is not strictly necessary on many platforms, namely those where the module is effectively @emph{linked with the Smalltalk virtual machine} at run-time; but since some@footnote{The most notable are @sc{aix} and Windows.} cannot obtain this, for maximum portability you must always call the virtual machine through the proxy and never refer to any symbol which the virtual machine exports. For uniformity, even programs that link with @file{libgst.a} should not call these functions directly, but through a @code{VMProxy} exported by @file{libgst.a} and accessible through the @code{gst_interpreter_proxy} variable. Modules are shared libraries; the default directory in which modules are searched for is stored in a @file{gnu-smalltalk.pc} file that is installed by @gst{} so that it can be used with @command{pkg-config}. An Autoconf macro @code{AM_PATH_GST} is also installed that will put the directory in the @code{gstmoduledir} Autoconf substitution. When using @gnu{} Automake and Libtool, you can then build modules by including something like this in @file{Makefile.am}: @example gstmodule_LTLIBRARIES = libdigest.la libdigest_la_LDFLAGS = -module -no-undefined @dfn{... more flags ...} libdigest_la_SOURCES = @dfn{... your source files ...} @end example While you can use @code{DLD class>>#addModule:} to link a module into the virtual machine at run time, usually bindings that require a module are complex enough to be packaged as @file{.star} files. In this case, you will have to add the name of the module in a package file (@pxref{Packages}). In this case, the relevant entry in the file will be @example Digest digest.st md5.st sha1.st digest MD5Test SHA1Test mdtests.st @end example There is also a third case, in which the bindings are a mixture of code written specially for @gst{}, and the normal C library. In this case, you can use a combination of dynamic shared libraries and dynamic modules. To do this, you can specify both @code{} and @code{} tags in the @file{package.xml} file; alternatively, the following functions allow you to call @code{DLD class>>#addLibrary:} from within a module. @deftypefun mst_Boolean dlOpen (void *filename, int module) Open the library pointed to by with @var{filename} (which need not include an extension), and invoke gst_initModule if it is found in the library. If @var{module} is false, add the file to the list of libraries that Smalltalk searches for external symbols. Return true if the library was found. @end deftypefun @deftypefun void dlAddSearchDir (const char *dir) Add @var{dir} at the beginning of the search path of @code{dlOpen}. @end deftypefun @deftypefun void dlPushSearchPath (void) Save the current value of the search path for @code{dlOpen}. This can be used to temporarily add the search path for the libraries added by a module, without affecting subsequent libraries manually opened with the @code{DLD} class. @end deftypefun @deftypefun void dlPopSearchPath (void) Restore the last saved value of the search path. @end deftypefun @node C callout @section Using the C callout mechanism To use the C callout mechanism, you first need to inform Smalltalk about the C functions that you wish to call. You currently need to do this in two places: 1) you need to establish the mapping between your C function's address and the name that you wish to refer to it by, and 2) define that function along with how the argument objects should be mapped to C data types to the Smalltalk interpreter. As an example, let us use the pre-defined (to @gst{}) functions of @code{system} and @code{getenv}. First, the mapping between these functions and string names for the functions needs to be established in your module. If you are writing an external Smalltalk module (which can look at Smalltalk objects and manipulate them), see @ref{External modules, , Linking your libraries to the virtual machine}; if you are using function from a dynamically loaded library, see @ref{Dynamic loading}. Second, we need to define a method that will invoke these C functions and describe its arguments to the Smalltalk runtime system. Such a method is defined with a primitive-like syntax, similar to the following example (taken from @file{kernel/CFuncs.st}) @example system: aString getenv: aString @end example These methods were defined on class @code{SystemDictionary}, so that we would invoke it thus: @example Smalltalk system: 'lpr README' ! @end example However, there is no special significance to which class receives the method; it could have just as well been Float, but it might look kind of strange to see: @example 1701.0 system: 'mail help-smalltalk@@g@:nu.org' ! @end example The various keyword arguments are described below. @table @b @item @code{cCall: 'system'} This says that we are defining the C function @code{system}. This name must be @strong{exactly} the same as the string passed to @code{defineCFunc}. The name of the method does not have to match the name of the C function; we could have just as easily defined the selector to be @code{'rambo: fooFoo'}; it's just good practice to define the method with a similar name and the argument names to reflect the data types that should be passed. @item @code{returning: #int} This defines the C data type that will be returned. It is converted to the corresponding Smalltalk data type. The set of valid return types is: @table @code @item char Single C character value @item string A C char *, converted to a Smalltalk string @item stringOut A C char *, converted to a Smalltalk string and then freed. @item symbol A C char *, converted to a Smalltalk symbol @item symbolOut A C char *, converted to a Smalltalk symbol and then freed. @item int A C int value @item uInt A C unsigned int value @item long A C long value @item uLong A C unsigned long value @item double A C double, converted to an instance of FloatD @item longDouble A C long double, converted to an instance of FloatQ @item void No returned value (@code{self} returned from Smalltalk) @item wchar Single C wide character (@code{wchar_t}) value @item wstring Wide C string (@code{wchar_t *}), converted to a UnicodeString @item wstringOut Wide C string (@code{wchar_t *}), converted to a UnicodeString and then freed @item cObject An anonymous C pointer; useful to pass back to some C function later @item smalltalk An anonymous (to C) Smalltalk object pointer; should have been passed to C at some point in the past or created by the program by calling other public @gst{} functions (@pxref{Smalltalk types}). @item @var{ctype} You can pass an instance of CType or one of its subclasses (@pxref{C data types}). In this case the object will be sent @code{#narrow} before being returned: an example of this feature is given in the experimental Gtk+ bindings. @end table @item @code{args: #(#string)} This is an array of symbols that describes the types of the arguments in order. For example, to specify a call to open(2), the arguments might look something like: @example args: #(#string #int #int) @end example The following argument types are supported; see above for details. @table @code @item unknown Smalltalk will make the best conversion that it can guess for this object; see the mapping table below @item boolean passed as @code{char}, which is promoted to @code{int} @item char passed as @code{char}, which is promoted to @code{int} @item wchar passed as @code{wchar_t} @item string passed as @code{char *} @item byteArrayOut passed as @code{char *}. The contents are expected to be overwritten with a new C string, and copied back to the object that was passed on return from the C function @item stringOut passed as @code{char *}, the contents are expected to be overwritten with a new C string, and the object that was passed becomes the new string on return @item wstring passed as @code{wchar_t *} @item wstringOut passed as @code{wchar_t *}, the contents are expected to be overwritten with a new C wide string, and the object that was passed becomes the new string on return @item symbol passed as @code{char *} @item byteArray passed as @code{char *}, even though may contain NUL's @item int passed as @code{int} @item uInt passed as @code{unsigned int} @item long passed as @code{long} @item uLong passed as @code{unsigned long} @item double passed as @code{double} @item longDouble passed as @code{long double} @item cObject C object value passed as @code{void *}. Any class with non-pointer indexed instance variables can be passed as a @code{#cObject}, and @gst{} will pass the address of the first indexed instance variable. This however should never be done for functions that allocate objects, call back into Smalltalk code or otherwise may cause a garbage collection: after a GC, pointers passed as @code{#cObject} may be invalidated. In this case, it is safer to pass every object as @code{#smalltalk}, or to only pass @code{CObject}s that were returned by a C function previously. In addition, @code{#cObject} can be used for function pointers. These are instances of @code{CCallable} or one of its subclasses. See @ref{Smalltalk callbacks} for more information on how to create function pointers for Smalltalk blocks. @item cObjectPtr Pointer to C object value passed as @code{void **}. The @code{CObject} is modified on output to reflect the value stored into the passed object. @item smalltalk Pass the object pointer to C. The C routine should treat the value as a pointer to anonymous storage. This pointer can be returned to Smalltalk at some later point in time. @item variadic @itemx variadicSmalltalk an Array is expected, each of the elements of the array will be converted like an @code{unknown} parameter if @code{variadic} is used, or passed as a raw object pointer for @code{variadicSmalltalk}. @item self @itemx selfSmalltalk Pass the receiver, converting it to C like an @code{unknown} parameter if @code{self} is used or passing the raw object pointer for @code{selfSmalltalk}. Parameters passed this way don't map to the message's arguments, instead they map to the message's receiver. @end table @end table Table of parameter conversions: @multitable {Declared param type} {Boolean (True, False)} {@code{int} (C promotion rule)} @item Declared param type @tab Object type @tab C parameter type used @item boolean @tab Boolean (True, False)@tab int @item byteArray @tab ByteArray @tab char * @item cObject @tab CObject @tab void * @item cObject @tab ByteArray, etc. @tab void * @item cObjectPtr @tab CObject @tab void ** @item char @tab Boolean (True, False)@tab int @item char @tab Character @tab int (C promotion rule) @item char @tab Integer @tab int @item double @tab Float @tab double (C promotion) @item longDouble @tab Float @tab long double @item int @tab Boolean (True, False)@tab int @item int @tab Integer @tab int @item uInt @tab Boolean (True, False)@tab unsigned int @item uInt @tab Integer @tab unsigned int @item long @tab Boolean (True, False)@tab long @item long @tab Integer @tab long @item uLong @tab Boolean (True, False)@tab unsigned long @item uLong @tab Integer @tab unsigned long @item smalltalk, selfSmalltalk @tab anything @tab OOP @item string @tab String @tab char * @item string @tab Symbol @tab char * @item stringOut @tab String @tab char * @item symbol @tab Symbol @tab char * @item unknown, self @tab Boolean (True, False)@tab int @item unknown, self @tab ByteArray @tab char * @item unknown, self @tab CObject @tab void * @item unknown, self @tab Character @tab int @item unknown, self @tab Float @tab double @item unknown, self @tab Integer @tab long @item unknown, self @tab String @tab char * @item unknown, self @tab Symbol @tab char * @item unknown, self @tab anything else @tab OOP @item variadic @tab Array @tab each element is passed according to "unknown" @item variadicSmalltalk @tab Array @tab each element is passed as an OOP @item wchar @tab Character @tab wchar_t @item wstring @tab UnicodeString @tab wchar_t * @item wstringOut @tab UnicodeString @tab wchar_t * @end multitable When your call-out returns @code{#void}, depending on your application you might consider using @dfn{asynchronous call-outs}. These are call-outs that do not suspend the process that initiated them, so the process might be scheduled again, executing the code that follows the call-out, during the execution of the call-out itself. This is particularly handy when writing event loops (the most common place where you call back into Smalltalk) because then @emph{you can handle events that arrive during the handling of an outer event} before the outer event's processing has ended. Depending on your application this might be correct or not, of course. In the future, asynchronous call-outs might be started into a separate thread. An asynchronous call-out is defined using an alternate primitive-like syntax, @code{asyncCCall:args:}. Note that the returned value parameter is missing because an asynchronous call-out always returns @code{nil}. @node C data types @section The C data type manipulation system @c rewrite this..... @code{CType} is a class used to represent C data types themselves (no storage, just the type). There are subclasses called things like @code{C@var{mumble}CType}. The instances can answer their size and alignment. Their @code{valueType} is the underlying type of data. It's either an integer, which is interpreted by the interpreter as the scalar type, or the underlying element type, which is another @code{CType} subclass instance. To make life easier, there are global variables which hold onto instances of @code{CScalarCType}: they are called @code{C@var{mumble}Type} (like @code{CIntType}, not like @code{CIntCType}), and can be used wherever a C datatype is used. If you had an array of strings, the elements would be CStringType's (a specific instance of CScalarCType). @code{CObject} is the base class of the instances of C data. It has a subclass called @code{CScalar}, which has subclasses called @code{C@var{mumble}}. These subclasses can answer size and alignment information. Instances of @code{CObject} can hold a raw C pointer (for example in @code{malloc}ed heap)), or can delegate their storage to a @code{ByteArray}. In the latter case, the storage is automatically garbage collected when the @code{CObject} becomes dead, and the VM checks accesses to make sure they are in bounds. On the other hand, the storage may move, and for this reason extra care must be put when using this kind of @code{CObject} with C routines that call back into Smalltalk, or that store the passed pointer somewhere. Instances of @code{CObject} can be created in many ways: @itemize @item creating an instance with @code{@var{class} new} initializes the pointer to @code{NULL}; @item doing @code{@var{type} new}, where @var{type} is a @code{CType} subclass instance, allocates a new instance with @code{malloc}. @item doing @code{@var{type} gcNew}, where @var{type} is a @code{CType} subclass instance, allocates a new instance backed by garbage-collected storage. @end itemize @code{CStruct} and @code{CUnion} subclasses are special. First, @code{new} allocates a new instance with @code{malloc} instead of initializing the pointer to @code{NULL}. Second, they support @code{gcNew} which creates a new instance backed by garbage-collected storage. @code{CObject}s created by the C callout mechanism are never backed by garbage-collected storage. @code{CObject} and its subclasses represent a pointer to a C object and as such provide the full range of operations supported by C pointers. For example, @code{+} @code{anInteger} which returns a CObject which is higher in memory by @code{anInteger} times the size of each item. There is also @code{-} which acts like @code{+} if it is given an integer as its parameter. If a CObject is given, it returns the difference between the two pointers. @code{incr}, @code{decr}, @code{incrBy:}, @code{decrBy:} adjust the string either forward or backward, by either 1 or @code{n} characters. Only the pointer to the string is changed; the actual characters in the string remain untouched. CObjects can be divided into two families, scalars and non-scalars, just like C data types. Scalars fetch a Smalltalk object when sent the @code{value} message, and change their value when sent the @code{value:} message. Non-scalars do not support these two messages. Non-scalars include instances of @code{CArray} and subclasses of @code{CStruct} and @code{CUnion} (but not @code{CPtr}). @code{CPtr}s and @code{CArray}s get their underlying element type through a @code{CType} subclass instance which is associated with the @code{CArray} or @code{CPtr} instance. @code{CPtr}'s @code{value} and @code{value:} method get or change the underlying value that's pointed to. @code{value} returns another @code{CObject} corresponding to the pointed value. That's because, for example, a @code{CPtr} to @code{long} points to a place in memory where a pointer to long is stored. It is really a @code{long **} and must be dereferenced twice with @code{cPtr value value} to get the @code{long}. @code{CString} is a subclass of @code{CPtr} that answers a Smalltalk @code{String} when sent @code{value}, and automatically allocates storage to copy and null-terminate a Smalltalk @code{String} when sent @code{value:}. @code{replaceWith:} replaces the string the instance points to with a new string or @code{ByteArray}, passed as the argument. Actually, it copies the bytes from the Smalltalk @code{String} instance aString into the same buffer already pointed to by the @code{CString}, with a null terminator. Finally, there are @code{CStruct} and @code{CUnion}, which are abstract subclasses of @code{CObject}@footnote{Actually they have a common superclass named @code{CCompound}.}. The following will refer to CStruct, but the same considerations apply to CUnion as well, with the only difference that CUnions of course implement the semantics of a C union. These classes provide direct access to C data structures including @itemize @bullet @bulletize @code{long} (unsigned too) @bulletize @code{short} (unsigned too) @bulletize @code{char} (unsigned too) & byte type @bulletize @code{double}, @code{long double}, @code{float} @bulletize @code{string} (NUL terminated char *, with special accessors) @bulletize arrays of any type @bulletize pointers to any type @bulletize other structs containing any fixed size types @end itemize Here is an example struct decl in C: @example struct audio_prinfo @{ unsigned channels; unsigned precision; unsigned encoding; unsigned gain; unsigned port; unsigned _xxx[4]; unsigned samples; unsigned eof; unsigned char pause; unsigned char error; unsigned char waiting; unsigned char _ccc[3]; unsigned char open; unsigned char active; @}; struct audio_info @{ audio_prinfo_t play; audio_prinfo_t record; unsigned monitor_gain; unsigned _yyy[4]; @}; @end example And here is a Smalltalk equivalent decision: @example CStruct subclass: AudioPrinfo [ ] CStruct subclass: AudioInfo [ ] @end example This creates two new subclasses of @code{CStruct} called @code{AudioPrinfo} and @code{AudioInfo}, with the given fields. The syntax is the same as for creating standard subclasses, with the additional metadata @code{declaration:}. You can make C functions return @code{CObject}s that are instances of these classes by passing @code{AudioPrinfo type} as the parameter to the @code{returning:} keyword. AudioPrinfo has methods defined on it like: @example #sampleRate #channels #precision #encoding @end example @noindent etc. These access the various data members. The array element accessors (xxx, ccc) just return a pointer to the array itself. For simple scalar types, just list the type name after the variable. Here's the set of scalars names, as defined in @file{kernel/CStruct.st}: @example #long CLong #uLong CULong #ulong CULong #byte CByte #char CChar #uChar CUChar #uchar CUChar #short CShort #uShort CUShort #ushort CUShort #int CInt #uInt CUInt #uint CUInt #float CFloat #double CDouble #longDouble CLongDouble #string CString #smalltalk CSmalltalk #@{...@} @r{A given subclass of @code{CObject}} @end example The @code{#@{@dots{}@}} syntax is not in the Blue Book, but it is present in @gst{} and other Smalltalks; it returns an Association object corresponding to a global variable. To have a pointer to a type, use something like: @example (#example (#ptr #long)) @end example To have an array pointer of size @var{size}, use: @example (#example (#array #string @var{size})) @end example Note that this maps to @code{char *example[@var{size}]} in C. The objects returned by using the fields are CObjects; there is no implicit value fetching currently. For example, suppose you somehow got ahold of an instance of class AudioPrinfo as described above (the instance is a CObject subclass and points to a real C structure somewhere). Let's say you stored this object in variable @code{audioInfo}. To get the current gain value, do @example audioInfo gain value @end example @noindent to change the gain value in the structure, do @example audioInfo gain value: 255 @end example The structure member message just answers a @code{CObject} instance, so you can hang onto it to directly refer to that structure member, or you can use the @code{value} or @code{value:} methods to access or change the value of the member. Note that this is the same kind of access you get if you use the @code{addressAt:} method on CStrings or CArrays or CPtrs: they return a CObject which points to a C object of the right type and you need to use @code{value} and @code{value:} to access and modify the actual C variable. @node Smalltalk types @section Manipulating Smalltalk data from C @gst{} internally maps every object except Integers to a data structure named an @dfn{OOP} (which is short for @dfn{Ordinary Object Pointer}). An OOP is a pointer to an internal data structure; this data structure basically adds a level of indirection in the representation of objects, since it contains @itemize @bullet @item a pointer to the actual object data @item a bunch of flags, most of which interest the garbage collection process @end itemize This additional level of indirection makes garbage collection very efficient, since the collector is free to move an object in memory without updating every reference to that object in the heap, thereby keeping the heap fully compact and allowing very fast allocation of new objects. However, it makes C code that wants to deal with objects even more messy than it would be without; if you want some examples, look at the hairy code in @gst{} that deals with processes. To shield you as much as possible from the complications of doing object-oriented programming in a non-object-oriented environment like C, @gst{} provides friendly functions to map between common Smalltalk objects and C types. This way you can simply declare OOP variables and then use these functions to treat their contents like C data. These functions are passed to a module via the @code{VMProxy} struct, a pointer to which is passed to the module, as shown in @ref{External modules, , Linking your libraries to the virtual machine}. They can be divided in two groups, those that map @emph{from Smalltalk objects to C data types} and those that map @emph{from C data types to Smalltalk objects}. Here are those in the former group (Smalltalk to C); you can see that they all begin with @code{OOPTo}: @deftypefun long OOPToInt (OOP) This function assumes that the passed OOP is an Integer and returns the C @code{signed long} for that integer. @end deftypefun @deftypefun long OOPToId (OOP) This function returns an unique identifier for the given OOP, valid until the OOP is garbage-collected. @end deftypefun @deftypefun double OOPToFloat (OOP) This function assumes that the passed OOP is an Integer or Float and returns the C @code{double} for that object. @end deftypefun @deftypefun {long double} OOPToLongDouble (OOP) This function assumes that the passed OOP is an Integer or Float and returns the C @code{long double} for that object. @end deftypefun @deftypefun int OOPToBool (OOP) This function returns a C integer which is true (i.e. @code{!= 0}) if the given OOP is the @code{true} object, false (i.e. @code{== 0}) otherwise. @end deftypefun @deftypefun char OOPToChar (OOP) This function assumes that the passed OOP is a Character and returns the C @code{char} for that integer. @end deftypefun @deftypefun wchar_t OOPToWChar (OOP) This function assumes that the passed OOP is a Character or UnicodeCharacter and returns the C @code{wchar_t} for that integer. @end deftypefun @deftypefun char *OOPToString (OOP) This function assumes that the passed OOP is a String or ByteArray and returns a C null-terminated @code{char *} with the same contents. It is the caller's responsibility to free the pointer and to handle possible @samp{NUL} characters inside the Smalltalk object. @end deftypefun @deftypefun wchar_t *OOPToWString (OOP) This function assumes that the passed OOP is a UnicodeString and returns a C null-terminated @code{wchar_t *} with the same contents. It is the caller's responsibility to free the pointer and to handle possible @samp{NUL} characters inside the Smalltalk object. @end deftypefun @deftypefun char *OOPToByteArray (OOP) This function assumes that the passed OOP is a String or ByteArray and returns a C @code{char *} with the same contents, without null-terminating it. It is the caller's responsibility to free the pointer. @end deftypefun @deftypefun PTR OOPToCObject (OOP) This functions assumes that the passed OOP is a kind of CObject and returns a C @code{PTR} to the C data pointed to by the object. The caller should not free the pointer, nor assume anything about its size and contents, unless it @b{exactly} knows what it's doing. A @code{PTR} is a @code{void *} if supported, or otherwise a @code{char *}. @end deftypefun @deftypefun long OOPToC (OOP) This functions assumes that the passed OOP is a String, a ByteArray, a CObject, or a built-in object (@code{nil}, @code{true}, @code{false}, character, integer). If the OOP is @code{nil}, it answers 0; else the mapping for each object is exactly the same as for the above functions. Note that, even though the function is declared as returning a @code{long}, you might need to cast it to either a @code{char *} or @code{PTR}. @end deftypefun While special care is needed to use the functions above (you will probably want to know at least the type of the Smalltalk object you're converting), the functions below, which convert C data to Smalltalk objects, are easier to use and also put objects in the incubator so that they are not swept by a garbage collection (@pxref{Incubator}). These functions all @dfn{end} with @code{ToOOP}, except @code{cObjectToTypedOOP}: @deftypefun OOP intToOOP (long) This object returns a Smalltalk @code{Integer} which contains the same value as the passed C @code{long}. @end deftypefun @deftypefun OOP uintToOOP (unsigned long) This object returns a Smalltalk @code{Integer} which contains the same value as the passed C @code{unsigned long}. @end deftypefun @deftypefun OOP idToOOP (OOP) This function returns an OOP from a unique identifier returned by @code{OOPToId}. The OOP will be the same that was passed to @code{OOPToId} only if the original OOP has not been garbage-collected since the call to @code{OOPToId}. @end deftypefun @deftypefun OOP floatToOOP (double) This object returns a Smalltalk @code{FloatD} which contains the same value as the passed @code{double}. Unlike Integers, FloatDs have exactly the same precision as C doubles. @end deftypefun @deftypefun OOP longDoubleToOOP (long double) This object returns a Smalltalk @code{FloatQ} which contains the same value as the passed @code{long double}. Unlike Integers, FloatQs have exactly the same precision as C long doubles. @end deftypefun @deftypefun OOP boolToOOP (int) This object returns a Smalltalk @code{Boolean} which contains the same boolean value as the passed C @code{int}. That is, the returned OOP is the sole instance of either @code{False} or @code{True}, depending on where the parameter is zero or not. @end deftypefun @deftypefun OOP charToOOP (char) This object returns a Smalltalk @code{Character} which represents the same char as the passed C @code{char}. @end deftypefun @deftypefun OOP charToOOP (wchar_t) This object returns a Smalltalk @code{Character} or @code{UnicodeCharacter} which represents the same char as the passed C @code{wchar_t}. @end deftypefun @deftypefun OOP classNameToOOP (char *) This method returns the Smalltalk class (i.e. an instance of a subclass of Class) whose name is the given parameter. Namespaces are supported; the parameter must give the complete path to the class starting from the @code{Smalltalk} dictionary. @code{NULL} is returned if the class is not found. This method is slow; you can safely cache its result. @end deftypefun @deftypefun OOP stringToOOP (char *) This method returns a String which maps to the given null-terminated C string, or the builtin object @code{nil} if the parameter points to address 0 (zero). @end deftypefun @deftypefun OOP wstringToOOP (wchar_t *) This method returns a UnicodeString which maps to the given null-terminated C wide string, or the builtin object @code{nil} if the parameter points to address 0 (zero). @end deftypefun @deftypefun OOP byteArrayToOOP (char *, int) This method returns a ByteArray which maps to the bytes that the first parameters points to; the second parameter gives the size of the ByteArray. The builtin object @code{nil} is returned if the first parameter points to address 0 (zero). @end deftypefun @deftypefun OOP symbolToOOP (char *) This method returns a String which maps to the given null-terminated C string, or the builtin object @code{nil} if the parameter points to address 0 (zero). @end deftypefun @deftypefun OOP cObjectToOOP (PTR) This method returns a CObject which maps to the given C pointer, or the builtin object @code{nil} if the parameter points to address 0 (zero). The returned value has no precise CType assigned. To assign one, use @code{cObjectToTypedOOP}. @end deftypefun @deftypefun OOP cObjectToTypedOOP (PTR, OOP) This method returns a CObject which maps to the given C pointer, or the builtin object @code{nil} if the parameter points to address 0 (zero). The returned value has the second parameter as its type; to get possible types you can use @code{typeNameToOOP}. @end deftypefun @deftypefun OOP typeNameToOOP (char *) All this method actually does is evaluating its parameter as Smalltalk code; so you can, for example, use it in any of these ways: @example cIntType = typeNameToOOP("CIntType"); myOwnCStructType = typeNameToOOP("MyOwnCStruct type"); @end example This method is primarily used by @code{msgSendf} (@pxref{Smalltalk callin}), but it can be useful if you use lower level call-in methods. This method is slow too; you can safely cache its result. @end deftypefun As said above, the C to Smalltalk layer automatically puts the objects it creates in the incubator which prevents objects from being collected as garbage. A plugin, however, has limited control on the incubator, and the incubator itself is not at all useful when objects should be kept registered for a relatively long time, and whose lives in the registry typically overlap. To avoid garbage collection of such object, you can use these functions, which access a separate registry: @deftypefun OOP registerOOP (OOP) Puts the given OOP in the registry. If you register an object multiple times, you will need to unregister it the same number of times. You may want to register objects returned by Smalltalk call-ins. @end deftypefun @deftypefun void unregisterOOP (OOP) Removes an occurrence of the given OOP from the registry. @end deftypefun @deftypefun void registerOOPArray (OOP **, OOP **) Tells the garbage collector that an array of objects must be made part of the root set. The two parameters point indirectly to the base and the top of the array; that is, they are pointers to variables holding the base and the top of the array: having indirect pointers allows you to dynamically change the size of the array and even to relocate it in memory without having to unregister and re-register it every time you modify it. If you register an array multiple times, you will need to unregister it the same number of times. @end deftypefun @deftypefun void unregisterOOPArray (OOP **) Removes the array with the given base from the registry. @end deftypefun @node Smalltalk callin @section Calls from C to Smalltalk @gst{} provides seven different function calls that allow you to call Smalltalk methods in a different execution context than the current one. The priority in which the method will execute will be the same as the one of Smalltalk process which is currently active. Four of these functions are more low level and are more suited when the Smalltalk program itself gave a receiver, a selector and maybe some parameters; the others, instead, are more versatile. One of them (@code{msgSendf}) automatically handles most conversions between C data types and Smalltalk objects, while the others takes care of compiling full snippets of Smalltalk code. All these functions handle properly the case of specifying, say, 5 arguments for a 3-argument selector---see the description of the single functions for more information). In all cases except @code{msgSendf}, passing NULL as the selector will expect the receiver to be a block and evaluate it. @deftypefun OOP msgSend (OOP receiver, OOP selector, @dots{}) This function sends the given selector (should be a Symbol, otherwise @code{nilOOP} is returned) to the given receiver. The message arguments should also be OOPs (otherwise, an access violation exception is pretty likely) and are passed in a NULL-terminated list after the selector. The value returned from the method is passed back as an OOP to the C program as the result of @code{msgSend}, or @code{nilOOP} if the number of arguments is wrong. Example (same as @code{1 + 2}): @example OOP shouldBeThreeOOP = vmProxy->msgSend( intToOOP(1), symbolToOOP("+"), intToOOP(2), NULL); @end example @end deftypefun @deftypefun OOP strMsgSend (OOP receiver, char *selector, @dots{}) This function is the same as above, but the selector is passed as a C string and is automatically converted to a Smalltalk symbol. Theoretically, this function is a bit slower than @code{msgSend} if your program has some way to cache the selector and avoiding a call to @code{symbolToOOP} on every call-in. However, this is not so apparent in real'' code because the time spent in the Smalltalk interpreter will usually be much higher than the time spent converting the selector to a Symbol object. Example: @example OOP shouldBeThreeOOP = vmProxy->strMsgSend( intToOOP(1), "+", intToOOP(2), NULL); @end example @end deftypefun @deftypefun OOP vmsgSend (OOP receiver, OOP selector, OOP *args) This function is the same as msgSend, but accepts a pointer to the NULL-terminated list of arguments, instead of being a variable-arguments functions. Example: @example OOP arguments[2], shouldBeThreeOOP; arguments[0] = intToOOP(2); arguments[1] = NULL; /* @dots{} some more code here @dots{} */ shouldBeThreeOOP = vmProxy->vmsgSend( intToOOP(1), symbolToOOP("+"), arguments); @end example @end deftypefun @deftypefun OOP nvmsgSend (OOP receiver, OOP selector, OOP *args, int nargs) This function is the same as msgSend, but accepts an additional parameter containing the number of arguments to be passed to the Smalltalk method, instead of relying on the NULL-termination of args. Example: @example OOP argument, shouldBeThreeOOP; argument = intToOOP(2); /* @dots{} some more code here @dots{} */ shouldBeThreeOOP = vmProxy->nvmsgSend( intToOOP(1), symbolToOOP("+"), &argument, 1); @end example @end deftypefun @deftypefun OOP perform (OOP, OOP) Shortcut function to invoke a unary selector. The first parameter is the receiver, and the second is the selector. @end deftypefun @deftypefun OOP performWith (OOP, OOP, OOP) Shortcut function to invoke a one-argument selector. The first parameter is the receiver, the second is the selector, the third is the sole argument. @end deftypefun @deftypefun OOP invokeHook (int) Calls into Smalltalk to process a @code{ObjectMemory} hook given by the parameter. In practice, @code{changed:} is sent to @code{ObjectMemory} with a symbol derived from the parameter. The parameter can be one of: @itemize @item @code{GST_BEFORE_EVAL} @item @code{GST_AFTER_EVAL} @item @code{GST_ABOUT_TO_QUIT} @item @code{GST_RETURN_FROM_SNAPSHOT} @item @code{GST_ABOUT_TO_SNAPSHOT} @item @code{GST_FINISHED_SNAPSHOT} @end itemize All cases where the last three should be used should be covered in @gst{}'s source code. The first three, however, can actually be useful in user code. @end deftypefun The two functions that directly accept Smalltalk code are named @code{evalCode} and @code{evalExpr}, and they're basically the same. They both accept a single parameter, a pointer to the code to be submitted to the parser. The main difference is that @code{evalCode} discards the result, while @code{evalExpr} returns it to the caller as an OOP. @code{msgSendf}, instead, has a radically different syntax. Let's first look at some examples. @example /* 1 + 2 */ int shouldBeThree; vmProxy->msgSendf(&shouldBeThree, "%i %i + %i", 1, 2) /* aCollection includes: 'abc' */ OOP aCollection; int aBoolean; vmProxy->msgSendf(&aBoolean, "%b %o includes: %s", aCollection, "abc") /* 'This is a test' printNl -- in two different ways */ vmProxy->msgSendf(NULL, "%v %s printNl", "This is a test"); vmProxy->msgSendf(NULL, "%s %s printNl", "This is a test"); /* 'This is a test', ' ok?' */ char *str; vmProxy->msgSendf(&str, "%s %s , %s", "This is a test", " ok?"); @end example As you can see, the parameters to msgSendf are, in order: @itemize @bullet @item A pointer to the variable which will contain the record. If this pointer is @code{NULL}, it is discarded. @item A description of the method's interface in this format (the object types, after percent signs, will be explained later in this section) @example %result_type %receiver_type selector %param1_type %param2_type @end example @item A C variable or Smalltalk object (depending on the type specifier) for the receiver @item If needed, the C variables and/or Smalltalk object (depending on the type specifiers) for the arguments. @end itemize Note that the receiver and parameters are NOT registered in the object registry (@pxref{Smalltalk types}). @dfn{receiver_type} and @dfn{paramX_type} can be any of these characters, with these meanings: @example Specifier C data type equivalent Smalltalk class i long Integer (see intToOOP) f double Float (see floatToOOP) F long double Float (see longDoubleToOOP) b int True or False (see boolToOOP) B OOP BlockClosure c char Character (see charToOOP) C PTR CObject (see cObjToOOP) s char * String (see stringToOOP) S char * Symbol (see symbolToOOP) o OOP any t char *, PTR CObject (see below) T OOP, PTR CObject (see below) w wchar_t Character (see wcharToOOP) W wchar_t * UnicodeString (see wstringToOOP) @end example @noindent @samp{%t} and @samp{%T} are particular in the sense that you need to pass @dfn{two} additional arguments to @code{msgSendf}, not one. The first will be a description of the type of the CObject to be created, the second instead will be the CObject's address. If you specify @samp{%t}, the first of the two arguments will be converted to a Smalltalk @code{CType} via @code{typeNameToOOP} (@pxref{Smalltalk types}); instead, if you specify @samp{%T}, you will have to directly pass an OOP for the new CObject's type. For @samp{%B} you should not pass a selector, and the block will be evaluated. The type specifiers you can pass for @dfn{result_type} are a bit different: @example Result Specifier if nil C data type expected result i 0L long nil or an Integer f 0.0 double nil or a Float F 0.0 long double nil or a Float b 0 int nil or a Boolean c '\0' char nil or a Character C NULL PTR nil or a CObject s NULL char * nil, a String, or a Symbol ? 0 char *, PTR See oopToC o nilOOP OOP any (result is not converted) w '\0' wchar_t nil or a Character W NULL wchar_t * nil or a UnicodeString v / any (result is discarded) @end example Note that, if resultPtr is @code{NULL}, the @dfn{result_type} is always treated as @samp{%v}. If an error occurs, the value in the result if nil' column is returned. @node Smalltalk callbacks @section Smalltalk blocks as C function pointers The Smalltalk callin mechanism can be used effectively to construct bindings to C libraries that require callbacks into Smalltalk. However, it is a `static'' mechanism, as the callback functions passed to the libraries have to be written in C and their type signatures are fixed. If the signatures of the callbacks are not known in advance, and the only way to define callbacks is via C function pointers (as opposed to reflective mechanisms such as the ones in GTK+), then the @code{VMProxy} functions for Smalltalk callin are not enough. @gst{} provides a more dynamic way to convert Smalltalk blocks into C function pointers through the @code{CCallbackDescriptor} class. This class has a constructor method that is similar to the @code{cCall:} annotation used for callouts. The method is called @code{for:returning:withArgs:} and its parameters are: @itemize @bullet @item a block, whose number of arguments is variable @item a symbol representing the return type @item an array representing the type of the arguments. @end itemize The array passed as the third parameter represents values that are passed @emph{from C to Smalltalk} and, as such, should be filled with the same rules that are used by the @emph{return type} of a C callout. In particular, if the C callback accepts an @code{int *} it is possible (and indeed useful) to specify the type of the argument as @code{#@{CInt@}}, so that the block will receive a @code{CInt} object. Here is an example of creating a callback which is passed to @code{glutReshapeFunc}@footnote{The GLUT bindings use a different scheme for setting up callbacks.}. The desired signature in C is @code{void (*) (int, int)}. @example | glut | @r{@dots{}} glut glutReshapeFunc: (CCallbackDescriptor for: [ :x :y | self reshape: x@@y ] returning: #void withArgs: #(#int #int)) @end example It is important to note that this kind of callback does not survive across an image load (this restriction may be lifted in a future version). When the image is loaded, it has to be reset by sending it the @code{link} message before it is passed to any C function. Sending the @code{link} message to an already valid callback is harmless and cheap. @node Other C functions @section Other functions available to modules In addition to the functions described so far, the @code{VMProxy} that is available to modules contains entry-points for many functions that aid in developing @gst{} extensions in C. This node documents these functions and the macros that are defined by @file{libgst/gstpub.h}. @deftypefun void asyncCall (void (*) (OOP), OOP) This functions accepts a function pointer and an OOP (or @code{NULL}, but not an arbitrary pointer) and sets up the interpreter to call the function as soon as the next message send is executed. @emph{Caution:} This and the next two are the only functions in the @code{intepreterProxy} that are thread-safe. @end deftypefun @deftypefun void asyncSignal (OOP) This functions accepts an OOP for a @code{Semaphore} object and signals that object so that one of the processes waiting on that semaphore is waken up. Since a Smalltalk call-in is not an atomic operation, the correct way to signal a semaphore is not to send the @code{signal} method to the object but, rather, to use: @example asyncSignal(semaphoreOOP) @end example The signal request will be processed as soon as the next message send is executed. @end deftypefun @deftypefun void asyncSignalAndUnregister (OOP) This functions accepts an OOP for a @code{Semaphore} object and signals that object so that one of the processes waiting on that semaphore is waken up; the signal request will be processed as soon as the next message send is executed. The object is then removed from the registry. @end deftypefun @deftypefun void wakeUp (void) When no Smalltalk process is running, @gst{} tries to limit CPU usage by pausing until it gets a signal from the OS. @code{wakeUp} is an alternative way to wake up the main Smalltalk loop. This should rarely be necessary, since the above functions already call it automatically. @end deftypefun @deftypefun void syncSignal (OOP, mst_Boolean) This functions accepts an OOP for a @code{Semaphore} object and signals that object so that one of the processes waiting on that semaphore is waken up. If the semaphore has no process waiting in the queue and the second argument is true, an excess signal is added to the semaphore. Since a Smalltalk call-in is not an atomic operation, the correct way to signal a semaphore is not to send the @code{signal} or @code{notify} methods to the object but, rather, to use: @example syncSignal(semaphoreOOP, true) @end example The @code{sync} in the name of this function distinguishes it from @code{asyncSignal}, in that it can only be called from a procedure already scheduled with @code{asyncCall}. It cannot be called from a call-in, or from other threads than the interpreter thread. @end deftypefun @deftypefun void syncWait (OOP) This function is present for backwards-compatibility only and should not be used. @end deftypefun @deftypefun void showBacktrace (FILE *) This functions show a backtrace on the given file.