Subversion Repositories or1k

Tcl/Tk has proven to be an excellent language for building small

programs that require a Graphical User Interface (GUI). However, it is

often inadequate for use in large commercial applications for a number

of reasons:

  *  Execution speed is usually too slow for serious computation.

  *  Complex data structures are difficult to construct.

  *  The lack of structure and typing in the Tcl language complicates

     the development of large codes.

  *  Tcl/Tk source code is easily read by the end user, making it hard

     for developers to protect proprietary algorithms.

  *  Large Tcl/Tk programs typically consist of many separate files

     that must be correctly positioned within the target computer's

     file system.  This can make the programs difficult to install,

     maintain and administer.

To circumvent these problems, we have constructed a system that makes

it easy for a C or C++ program to invoke and interact with Tcl/Tk.

This allows data structures and compute-intensive algorithms to be

coded in C or C++ while the GUI is coded in Tcl/Tk. It also allows the

entire application to be compiled into a single stand-alone

executable, which is not easily readable by the end user and which can

be run on computers that do not have Tcl/Tk installed.  We call our

system "ET" for "Embedded Tk".

1.  A Simple Example: ``Hello, World!''

The following is an ET implementation of the classic "Hello, World!"

program:

    void main(int argc, char **argv){

      Et_Init(&argc,argv);

      ET( button .b -text {Hello, World!} -command exit; pack .b );

      Et_MainLoop();

    }

This example is short, but is serves to illustrate the basic structure

of any ET application.  The first line of the main() procedure is a

call to the function Et_Init().  This function initializes the ET

system by creating a Tcl/Tk interpreter, connecting to the X11 server,

and creating a main window.  The last line of main() implements the

event loop.  Everything in between constitutes setup code.  In this

example, the setup code is a short Tcl/Tk script contained within the

special macro ET(). The et2c macro preprocessor will replace this ET()

macro with C code that causes the enclosed Tcl/Tk script to be

executed.

Of course, there is nothing in this example that could not also be

done by calling Tcl/Tk library routines directly, without the

intervening ET abstraction.  The advantage of using ET is that it

makes the interface between C and Tcl/Tk considerably less cumbersome

and error-prone, allowing the programmer to focus more mental energy

on the algorithm and less on the syntax of the programming language.

1.1 Compiling ``Hello, World!''

To compile the hello world example, we must first process the source

file using the et2c macro preprocessor, then link the results with the

et.o library.  Suppose the example code is contained in the file

hello.c. Then to compile the example (on most systems) requires the

following steps:

    et2c hello.c >hello_.c

    cc -o hello hello_.c et.o -ltk -ltcl -lX11 -lm

Assuming it is statically linked, the resulting executable file hello

contains everything needed to run the program: the Tcl/Tk interpreter,

the startup scripts and the application code.  The program can be

moved to other binary-compatible computers and executed there even if

the other computers do not have Tcl/Tk installed.

Additional information is provided below.

1.2 How to obtain sources and documentation

Complete sources to the et2c macro preprocessor and et.o library

comprise less than 2000 lines of code, including comments.  These

sources, together with source code to all example programs discussed

below, are available for anonymous FTP from ftp.vnet.net in the

directory /pub/users/drh.

A copy of this documentation is also available from the same FTP site.

The documentation is available in either PostScript, HTML, or an ASCII

text file.

2.  A Summary Of Services Provided By ET

The overall goal of ET is to simplify the interface between C and an

embedded Tcl/Tk-based GUI. To this end, the ET system provides a

number of services that aid in initializing the Tcl/Tk interpreter and

in transferring data and control between Tcl/Tk and C. The services

provided by ET are summarized here and described in more detail in

subsequent sections.

2.1 Routines to initialization the Tcl/Tk interpreter

The et.o library includes routines Et_Init() and Et_MainLoop() that

initialize the ET package and implement the X11 event loop.  A third

routine Et_ReadStdin() allows standard input to be read and

interpreted by the Tcl/Tk interpreter at run-time.

2.2 Macros to invoking Tcl/Tk from within C

The ET() macro looks and works just like a function in C, except that

its argument is a Tcl/Tk script instead of C code.  ET() returns

either ET_OK or ET_ERROR depending upon the success or failure of the

script.  Similar routines ET_STR(), ET_INT() and ET_DBL() also take a

Tcl/Tk script as their argument, but return a string, an integer, or a

double-precision floating point number instead of the status code.

2.3 A method to pass variable contents from C to Tcl/Tk

Wherever the string %d(x) occurs inside an ET() macro, the integer C

expression x is converted to ASCII and substituted in place of the

%d(x).  Similarly, %s(x) can be used to substitute a character string,

and %f(x) will substitute a floating point value.  The string %q(x)

works like %s(x) except that a backslash is inserted before each

character that has special meaning to Tcl/Tk.

2.4 Macros for creating new Tcl/Tk commands in C

The macro "ET_PROC( newcmd ){ ... }" defines a C function that is

invoked whenever the newcmd command is executed by the Tcl/Tk

interpreter.  Parameters argc and argv describe the arguments to the

command.  If a file named xyzzy.c contains one or more ET_PROC macros,

then the commands associated with those macros are registered with the

Tcl/Tk interpreter by invoking "ET_INSTALL_COMMANDS( xyzzy.c )" after

the Et_Init() in the main procedure.

2.5 Macros for linking external Tcl/Tk scripts into a C program

The macro "ET_INCLUDE( script.tcl )" causes the Tcl/Tk script in the

file script.tcl to be made a part of the C program and executed at the

point in the C program where the ET_INCLUDE macro is found.  The

external Tcl/Tk script is normally read into the C program at

compile-time and thus becomes part of the executable.  However, if the

-dynamic option is given to the et2c macro preprocessor, loading of

the external Tcl/Tk script is deferred to run-time.

2.6 Tcl/Tk return status macros

The macros ET_OK and ET_ERROR are set equal to TCL_OK and TCL_ERROR.

This often eliminates the need to put "#include " at the beginning of

files that use ET.

2.7 Convenience variables

ET defines three global C variables as a convenience to the

programmer.  Et_Interp is a pointer to the Tcl/Tk interpreter used by

ET. Et_MainWindow is the main window of the ET application.

Et_Display is the Display pointer required as the first argument to

many XLib routines.  ET also provides two global Tcl variables,

cmd_name and cmd_dir.  These contain the name of the executable and

the directory where the executable is found.

3.  Example 2: A Decimal Clock

The preceding "Hello, World!" example program demonstrated the basic

structure of an ET application including the use of the Et_Init()

function to initialize the Tcl/Tk interpreter and the Et_MainLoop()

function for implementing the X11 event loop.  The following program

will demonstrate additional aspects of the the ET system.

3.1 Source code for the decimal clock example

  /* This file implements a clock that shows the hour as

  ** fixed-point number X, such that

  **

  **      0.000 <= X < 24.000

  **

  ** X represents a fractional hour, not hours and minutes.

  ** Thus the time "8.500" means half past 8 o'clock, not

  ** ten minutes till 9.

  */

  #include 

  void main(int argc, char **argv){

    Et_Init(&argc,argv);

    ET_INSTALL_COMMANDS;

    ET(

      label .x -width 6 -text 00.000 -relief raised -bd 2

      pack .x

      UpdateTime

    );

    Et_MainLoop();

  }

  /* Update the time displayed in the text widget named ".x".

  ** Reschedule this routine to be called again after 3.6

  ** seconds.

  */

  ET_PROC( UpdateTime ){

    struct tm *pTime; /* The time of day, decoded */

    time_t t;         /* Number of seconds since the epoch */

    char buf[40];     /* The time value is written here */

    t = time(0);

    pTime = localtime(&t);

    sprintf(buf,"%2d.%03d",pTime->tm_hour,

       (pTime->tm_sec + 60*pTime->tm_min)*10/36);

    ET( .x config -text %s(buf); after 3600 UpdateTime );

    return ET_OK;

  }

                              **Image**

             3.1 Typical appearance of the decimal clock

3.2 Discussion of the decimal clock example

This example implements a clock program that displays the time in

thousandths of the hour, rather than the more usual hours, minutes and

seconds.  (Such a display might be useful, for instance, to a

consultant who bills time in tenth hour increments.) The code for this

example is contained in the file named dclock.c.

3.2.1 Initialization and event loop routines.  As in the first example

, the main() function to dclock begins with a call to Et_Init() and

ends with a call to Et_MainLoop(), with setup code in between.  If you

didn't see it before, note here that the Et_Init() function takes two

arguments -- a pointer to an integer that is the number of parameters

to the program, and a pointer to an array of pointers to strings that

are the program parameters.  Note especially that the first argument

is passed by reference, not by value.  The Et_Init() function requires

these arguments so that it can detect and act upon command line

arguments related to the initialization of Tcl/Tk. Any such arguments

detected are removed from the argc and argv variables before Et_Init()

returns, so the rest of the program need not be aware of their

existence.  The arguments currently understood by Et_Init() are

-geometry, -display, -name and -sync.  The use and meaning of these

arguments is exactly the same as in the standard Tcl/Tk interpreter

program "wish".

3.2.2 The ET_PROC macro.  The main difference between dclock and the

first example is that the setup code for dclock has an "

ET_INSTALL_COMMANDS" macro and there is an "ET_PROC" function defined

after main().  Let's begin by describing the ET_PROC macro.

The ET_PROC macro is nothing more than a convenient shorthand for

creating new Tcl/Tk commands in C. To create a new Tcl/Tk command, one

writes ET_PROC followed by the name of the new command in parentheses

and the C code corresponding to the new command in curly braces.

Within a single ET source file there can be any number of ET_PROC

macros, as long as the command names defined are all unique.  The et2c

macro preprocessor translates the ET_PROC macro into a C function

definition that implements the command, so ET_PROC macros should only

be used in places where it is legal to write C function definitions.

3.2.2.1 Parameters to an ET_PROC function.  The function created by an

ET_PROC macro has four parameters, though only two are commonly used.

The two useful parameters are argc and argv, which are the number of

arguments to the Tcl/Tk command and the value of each argument.  (The

command name itself counts as an argument here.) Hence, the argc and

argv parameters work just like the first two parameters to main() in a

typical C program.  Another parameter to every ET_PROC function is the

pointer to the Tcl/Tk interpreter, interp.  This variable is exactly

equal to the global variable Et_Interp.  The last parameter is called

clientData and is defined to be a pointer to anything.  It actually

points to the structure that defines the main window of the

application, and is therefore the same as the global variable

Et_MainWindow.

                              **Image**

        3.2 Summary of the parameters to each ET_PROC command

3.2.3 The ET_INSTALL_COMMANDS macro.  The ET_PROC macro will create a

C function that can be used as a Tcl/Tk command, but that function and

the corresponding command name must still be registered with the

Tcl/Tk interpreter before the command can be used.  This is the job of

the ET_INSTALL_COMMANDS macro.  Thus, in the dclock example, we must

invoke the ET_INSTALL_COMMANDS macro to register the UpdateTime

command prior to using the the UpdateTime command in any Tcl script.

Because new Tcl/Tk commands must be registered before they are used,

the ET_INSTALL_COMMANDS macros are usually the first setup code to

follow the Et_Init() function call.

Each instance of an ET_INSTALL_COMMANDS macro registers all ET_PROC

commands defined in a single source file.  The dclock example has only

a single ET_PROC command, but even if it had had 50, a single

ET_INSTALL_COMMANDS macro within the main() function would have been

sufficient to install them all.

The name of the source file containing the ET_PROC commands that are

to be registered is given as an argument to the ET_INSTALL_COMMANDS

macro.  If no argument is given, then the name of the file containing

the ET_INSTALL_COMMANDS macro is used.  Hence, the line in the dclock

example that registers the UpdateTime command can be written in either

of the following ways:

  ET_INSTALL_COMMANDS;

  ET_INSTALL_COMMANDS( dclock.c );

Note that the ET_INSTALL_COMMANDS macro does not actually open or read

the file named in its argument.  The macro just mangles the file name

in order to generate a unique procedure name for its own internal use.

The file itself is never accessed.  For this reason, the file name

specified as an argument to the ET_INSTALL_COMMANDS macro should not

contain a path, even if the named file is in a different directory.

3.2.4 The ET() macro.  We have already considered the ET() macro once,

in connection with the setup code for the "Hello, World!" example, and

we also observe that the ET() macro reappears in the setup code for

dclock and in the UpdateTime function.  Let's look at this macro in

more detail.

An ET() macro works just like a function, except that its argument is

a Tcl/Tk script instead of a C expression.  When an ET() macro is

executed, its argument is evaluated by the Tcl/Tk interpreter and an

integer status code is returned.  The status code will be either ET_OK

if the script was successful, or ET_ERROR if the script encountered an

error.  (An ET() macro might also return TCL_RETURN, TCL_BREAK, or

TCL_CONTINUE under rare circumstances.)

In the dclock example, a single ET() macro is used to initialize the

display of the decimal clock.  Three Tcl/Tk commands are contained

within the macro.  The first command creates a label widget for use as

the clock face, the second packs this label, and the third calls the

ET_PROC command named UpdateTime to cause the time on the clock face

to be updated.  (The UpdateTime command will arrange to call itself

again after a fixed interval, in order to update the time to the next

thousandth of an hour.)

The Tcl/Tk script contained in an ET() macro executes at the global

context level.  This means that the Tcl/Tk code within an ET() macro

can create and access only global Tcl/Tk variables.

3.2.4.1 The %s() phrase within an ET() macro.  Now consider the ET()

macro contained in the UpdateTime function.  The role of this macro is

to first change the label on the .x label widget to be the current

time and then reschedule the UpdateTime command to run again in 3.6

seconds.  The time value is stored in the character string buf[].

Within the argument to the ET() macro, the special phrase %s(buf)

causes the contents of the character string stored in buf[] to be

substituted in placed of the %s(buf) phrase itself.  The effect is

similar to a %s substitution in the format string of a printf

function.  In fact, the statement

    ET( .x config -text %s(buf); after 3600 UpdateTime );

is logical equivalent to

    char buf2[1000];

    sprintf(buf2," .x config -text %s; after 3600 UpdateTime ",buf);

    Tcl_GlobalEval(Et_Interp,buf2);

except that with the ET() macro there is never a danger of overflowing

the temporary buffer buf2[].

3.2.4.2 Other substitution phrases within ET() macros.  The phrase

%s(...) is replaced by the string contents of its argument within an

ET() macro.  Similarly, the phrases %d(...) and %f(...) are replaced

by ASCII representations of the integer and floating point number

given by the expression in their arguments.  The names of the

substitution phrases are taken from similar substitution tokens in the

format string of the printf function.  Note, however, that option

flags, precision and field widths are not allowed in an ET() macro

substitution phrase, as they are in printf.  The phrase %3.7f is

understood by printf but is is not understood by ET(). In an ET()

macro the only allowed form of a substitution phrase is where the

format letter immediately follows the percent symbol.

The ET() macro supports an additional substitution phrase not found in

standard printf: the %q(...). substitution.  The %q() works just like

%s() with the addition that it inserts extra backslash characters into

the substituted string in order to escape characters of the string

that would otherwise have special meaning to Tcl/Tk. Consider an

example.

  char *s = "The price is $1.45";

  ET( puts "%q(s)" );

Because the %q(...) macro was used instead of %s(...), an extra

backslash is inserted immediately before the "$".  The command string

passed to the Tcl/Tk interpreter is therefore:

  puts "The price is \$1.45"

This gives the expected result.  Without the extra backslash, Tcl/Tk

would have tried to expand "$1" as a variable, resulting in an error

message like this:

  can't read "1": no such variable

In general, it is always a good idea to use %q(...) instead of %s(...)

around strings that originate from outside the program -- you never

know when such strings may contain a character that needs to be

escaped.

                              **Image**

    3.3 Summary of substitution phrases understood by ET() macros

3.2.5 Variations on the ET() macro.  The ET() macro used in all

examples so far returns a status code indicating success or failure of

the enclosed Tcl script.  Sometimes, though, it is useful to have

access to the string returned by the Tcl script, instead of the status

code.  For these cases one can use the ET_STR() macro in place of ET()

The ET_STR() macro works just like ET() in most respects.  The sole

argument to ET_STR() is a Tcl/Tk script to which the usual %s(), %d(),

%f() and %q() substitutions are applied.  The difference between

ET_STR() and ET() is that ET_STR() returns a pointer to a

null-terminated string that is the result of the Tcl/Tk script if the

script was successful.  If the script failed, then ET_STR() returns a

NULL pointer.

It is very important to note that the string returned by ET_STR() is

ephemeral -- it will likely be deallocated, overwritten or otherwise

corrupted as soon as the next Tcl/Tk command is executed.  Therefore,

if you need to use this string for any length of time, it is a good

idea to make a copy.  In the following code fragment, the C string

variable entryText is made to point to a copy of the contents of an

entry widget named .entry.

  char *entryText = strdup( ET_STR(.entry get) );

It is not necessary to make a copy of the string returned by ET_STR()

if the string is used immediately and then discarded.  The following

two examples show uses of the ET_STR() macro where the result does not

need to be copied.  The first example shows a quick way to find the

width, height and location of the main window for an application:

  int width, height, x, y;

  sscanf(ET_STR(wm geometry .),"%dx%d+%d+%d",&width,&height,&x,&y);

The next example shows a convenient way to tell if a given widget is a

button:

  char *widget_name = ".xyz";

  if( strcmp(ET_STR(winfo class %s(widget_name)),"Button")==0 ){

    /* The widget is a button */

  }else{

    /* The widget is not a button */

  }

There also exist versions of the ET() macro that return an integer and

a floating point number: ET_INT() and ET_DBL(). These work much like

ET_STR() except that the returned string is converted to an integer or

to a double using the functions atoi() or atof().  The values 0 and

0.0 are returned if the Tcl/Tk script given in the argument fails or

if the returned string is not a valid number.

The ET_INT() and ET_DBL() macros are often used to read the values of

integer and floating point Tcl/Tk variables.  For instance, if Width

is a global Tcl/Tk variable containing an integer value, then we can

load that value into the integer C variable iWidth using the following

statement:

  iWidth = ET_INT( set Width );

The ET_INT() is also useful for recording the integer id number of an

object created on a Tcl/Tk canvas widget.  In the following example, a

line is created on the canvas widget .c and its id is recorded in the

integer C variable id.  Later, this id is used to delete the line.

  id = ET_INT( .c create line 100 100 200 200 -width 2 );

  /* ... intervening code omitted ... */

  ET( .c delete %d(id) );

The last example of the ET_INT() macro shows a convenient way to tell

if the X11 server is color or monochrome:

  if( ET_INT(winfo screendepth .)==1 ){

    /* The display is monochrome */

  }else{

    /* The display is color */

  }

                              **Image**

             3.4 Summary of variations on the ET() macro

4.  Example 3: fontchooser

As its name implies, the next example is a small utility program that

can be used to select X11 fonts.  The source code is contained in two

files, fontchooser.c and fontchooser.tcl.  We will look at the C code

first.

4.1 Source code to the font chooser

  /*

  ** This program allows the user to view the various fonts

  ** available on the X server.

  **

  ** Preprocess this file using "et2c" then link with "et.o".

  */

  #include "tk.h"     /* This automatically loads Xlib.h */

  void main(int argc, char **argv){

    Et_Init(&argc,argv);

    ET_INSTALL_COMMANDS;

    ET_INCLUDE( fontchooser.tcl );

    Et_MainLoop();

  }

  /* This function parses up font names as follows:

  **

  **         Font Family               Font size

  **    __________________________  ________________

  **   /                          \/                \

  **   -misc-fixed-medium-r-normal--10-100-75-75-c-60-iso8859-1

  **                                |   |  \___/    | \_______/

  **                                |   |    |      |     |

  **                                |   |    |      |     `-- Always as shown

  **                                |   |    |      |

  **             The point size ----'   |    |      `--- 10x average width

  **                                    |    |

  **              This field ignored----'    `--- Resolution in dots per inch

  **

  **

  ** If $name is a font name (the first 6 fields of the X11 font name)

  ** then this procedure defines the global variable $Font($name), giving

  ** it as a value a list of available font sizes in ascending order.

  ** Only fonts of a particular resolution are included.  By default, the

  ** resolution selected is 75dpi, but this can be changed by the

  ** argument to the command.

  **

  ** This command also creates global variable FontCount that holds the

  ** number of entries in the Font() array.

  */

  ET_PROC( FindFonts ){

    char **fontnames;    /* The names of all fonts in the selected resolution */

    int count;           /* Number of fonts */

    int i;               /* Loop counter */

    char pattern[400];   /* Buffer to hold a pattern used to select fonts. */

    if( argc==1 ){

      strcpy(pattern,"*-75-75-*-*-iso8859-1");

    }else if( argc==2 ){

      extern int atoi();

      int resolution = atoi(argv[1]);

      sprintf(pattern,"*-%d-%d-*-*-iso8859-1",resolution,resolution);

    }

    fontnames = XListFonts(Et_Display,pattern,1000,&count);

    ET(

      catch {unset Font}

      set FontCount 0

    );

    for(i=0; iresult = "Wrong # args";

      return ET_ERROR;

    }

    if( sscanf(argv[1],"%d-%*d-%*d-%*d-%*c-%d",&leftHeight,&leftWidth)!=2 ){

      interp->result = "First argument is not a font size";

      return ET_ERROR;

    }

    if( sscanf(argv[2],"%d-%*d-%*d-%*d-%*c-%d",&rightHeight,&rightWidth)!=2 ){

      interp->result = "Second argument is not a font size";

      return ET_ERROR;

    }

    result = leftHeight - rightHeight;

    if( result==0 ) result = leftWidth - rightWidth;

    sprintf(interp->result,"%d",result);

    return ET_OK;

  }

                              **Image**

          4.1 Typical appearance of the fontchooser program

4.2 Analysis of the fontchooser source code

As is the prior examples, the main() function for the fontchooser

begins and ends with calls to Et_Init() and Et_MainLoop().

Immediately following the Et_Init() call is an ET_INSTALL_COMMANDS

macro that registers the two commands FindFonts and FontSizeCompare

with the Tcl/Tk interpreter.

4.2.1 Using the argc and argv parameters to an ET_PROC function.  The

FindFonts routine is used to query the X server for the names of all

available fonts at a particular resolution specified by the argument

to the FindFonts routine.  If no resolution is specified (if the

FindFonts command is not given an argument in the Tcl/Tk script) then

the resolution defaults to 75 dots per inch.  The argc and argv

parameters are used to determine the number and value of arguments to

the FindFonts command.  The specified resolution is then used to

construct a search pattern for the fonts.

    if( argc==1 ){

      strcpy(pattern,"*-75-75-*-*-iso8859-1");

    }else if( argc==2 ){

      extern int atoi();

      int resolution = atoi(argv[1]);

      sprintf(pattern,"*-%d-%d-*-*-iso8859-1",resolution,resolution);

    }

4.2.2 Global variables defined by ET.  After creating a search

pattern, the The Xlib function XListFonts() is used find all fonts

that match that pattern.

    fontnames = XListFonts(Et_Display,pattern,1000,&count);

The first argument to XListFonts(), as in many Xlib functions, is a

pointer to a Display structure that defines the connection to the X

server.  The fontchooser program uses the convenience variable

Et_Display to fill this argument.  Et_Display is a global variable

defined in the et.o library and initialized to the active X connection

by the Et_Init() function.  The Et_Display variable is available for

use by any function that needs a Display pointer.

The ET system defines two global C variables besides Et_Display:

Et_Interp and Et_MainWindow.  The Et_Interp variable is a pointer to

the Tcl/Tk interpreter used by ET. This variable is very handy since

many routines in the Tcl/Tk library require a pointer to the

interpreter as their first argument.  The Et_MainWindow variable

defines the main window of the application, the window named "."

within Tcl/Tk scripts.  The main window is needed by a few Tcl/Tk

library routines, but is not as widely used as the other global

variables in ET.

All three global C variables in ET are initialized by the Et_Init()

routine and never change after initialization.

                              **Image**

                   4.2 Summary of global variables

4.2.3 Other actions of the FindFonts command.  After calling

XListFonts(), the FindFonts command splits each name into a "font

family" and a "font size".  For each font family, it creates an entry

in the global Tcl/Tk array variable Font with the font family name as

the index and a list of sizes for that font as the value.  A new entry

in the Font array is created, or else a new size is added to the list

of font sizes in that entry, by the following ET() macro:

      ET(

        if {![info exists {Font(%s(nameStart))}]} {

          set {Font(%s(nameStart))} {}

          incr FontCount

        }

        lappend {Font(%s(nameStart))} {%s(cp)}

      );

After all fonts returned by XListFonts have been processed, the list

of sizes on each entry in the Font array variable is sorted by the

final ET() macro in the FindFonts command:

    ET(

      foreach i [array names Font] {

        set Font($i) [lsort -command FontSizeCompare $Font($i)]

      }

    );

4.2.4 Operation of the FontSizeCompare command.  The FontSizeCompare

command is used to sort into ascending order the font sizes listed in

a single entry of the Font array.  The only place it is used is on the

lsort command contained in the final ET() macro of the FindFonts

routine.

Unlike any previously described ET_PROC command, FontSizeCompare makes

use of the interp parameter.  Recall that the interp parameter is a

pointer to the Tcl/Tk interpreter, and is therefore always equal to

the global C variable Et_Interp.  Hence, one could have used the

Et_Interp variable in place of the interp parameter throughout the

FindSizeCompare function and obtained the same result.

4.2.5 Constructing the GUI for the fontchooser.  The Tcl/Tk code that

defines the GUI for the fontchooser is contained in a separate file

fontchooser.tcl.  A small portion of this file follows:

  # This code accompanies the "fontchooser.c" file.  It does most of the

  # work of setting up and operating the font chooser.

  # Title the font chooser and make it resizeable.

  #

  wm title . "Font Chooser"

  wm iconname . "FontChooser"

  wm minsize . 1 1

  # Construct a panel for selecting the font family.

  #

  frame .name -bd 0 -relief raised

... 136 lines omitted ...

  # Begin by displaying the 75 dot-per-inch fonts

  #

  update

  LoadFontInfo 75

When the script in the file fontchooser.tcl executes, it constructs

the listboxes, scrollbars, menu and menu buttons of the fontchooser,

and finally calls the LoadFontInfo function.  The LoadFontInfo command

is defined by a proc statement in the part of the fontchooser.tcl file

that was omitted from the listing.  The LoadFontInfo function calls

FindFonts and then populates the listboxes accordingly.

The interesting thing about this example is how the script in

fontchooser.tcl is invoked.  In the prior examples ("Hello, World!"

and dclock) the Tcl/Tk script that setup the application was very

short and fit into an ET() macro in the main() function.  This same

approach could have been taken with the fontchooser.  We could have

put the entire text of the Tcl/Tk script into a 152 line ET() macro.

But that is inconvenient.  It is much easier to use an ET_INCLUDE

macro.

4.2.6 The ET_INCLUDE macro.  An ET_INCLUDE macro is similar to a

#include in the standard C preprocessor.  A #include reads in an

external C file as if it were part of the original C code.  ET_INCLUDE

does much the same thing for Tcl/Tk code.  It copies an external

Tcl/Tk script into the original C program, and causes that script to

be executed when control reaches the macro.

An important characteristic of the ET_INCLUDE macro is that it loads

the external Tcl/Tk script into the C program at compile time, not at

run time.  This means that a copy of the Tcl/Tk script actually

becomes part of the resulting executable.  To clarify this point,

consider the difference between the following two statements:

  ET( source fontchooser.tcl );

  ET_INCLUDE( fontchooser.tcl );

Both statements causes the file named fontchooser.tcl to be read and

executed by the Tcl/Tk interpreter.  The difference is that in the

first statement, the file is opened and read in at run-time,

immediately before the contained script is executed.  This means that

the file fontchooser.tcl must be available for reading by the program

in order for the program to work correctly.  In the second case, the

file is opened and read when the program is compiled.  The only work

left to do at run-time is to pass the contained script to the Tcl/Tk

interpreter.  In the second statement, then, the file fontchooser.tcl

does not have to be available to the program for correct operation.

4.2.7 How the ET_INCLUDE macro locates files.  The external script

file specified by an ET_INCLUDE macro need not be in the same

directory as the C program containing the ET_INCLUDE for the include

operation to work.  If the external script is in a different

directory, however, the name of that directory must be specified to

the et2c macro preprocessor using one or more "-Idirectory" command

line switches.

The algorithm used by et2c to locate a file is to first check the

working directory.  If the file is not there, then look in the

directory specified by the first -I option.  If the file is still not

found, then search the directory specified by the second -I option.

And so forth.  An error is reported only when the file mamed in the

ET_INCLUDE macro is missing from the working directory and from every

directory specified by -I options.  Note that this is essentially the

same algorithm used by the C compiler to find files named in #include

preprocessor directives.

4.2.8 The -dynamic option to et2c.  In a deliverable program, it is

usually best to load external Tcl/Tk scripts at compile time so that

the scripts will be bound into a single executable.  However, during

development it is sometimes advantageous to load external Tcl/Tk

scripts at run-time.  To do so allows these scripts to be modified

without having to recompile the C code.

The -dynamic option on the command line of the et2c preprocessor will

causes ET_INCLUDE macros to read their files at run-time instead of at

compile-time.  In effect, the -dynamic option causes macros of the

form ET_INCLUDE(X) to be converted into ET(source X). Generally

speaking, it is a good idea to use the -dynamic option on et2c

whenever the -g option (for symbolic debugging information) is being

used on the C compiler.

4.2.9 Use of ET_INCLUDE inside the et.o library.  When Tcl/Tk first

starts up, it must normally read a list of a dozen or so Tcl scripts

that contain definitions of widget bindings and related support

procedures.  In the standard interactive Tcl/Tk interpreter wish,

these files are read a run-time from a standard directory.  In an ET

application, however, these startup files are loaded into the

executable at compile time using ET_INCLUDE macros.

Startup files are loaded into the Et_Init() function that is part of

the et.o library.  The relevant source code followings:

    /*

     * Execute the start-up Tcl/Tk scripts. In the standard version of

     * wish, these are read from the library at run-time.  In this version

     * the scripts are compiled in.

     *

     * Some startup scripts contain "source" commands.  (Ex: tk.tcl in

     * Tk4.0).  This won't do for a stand-alone program.  For that reason,

     * the "source" command is disabled while the startup scripts are

     * being read.

     */

     ET( rename source __source__; proc source {args} {} );

     ET_INCLUDE( init.tcl );

     ET_INCLUDE( tk.tcl );

     ET_INCLUDE( button.tcl );

     ET_INCLUDE( dialog.tcl );

     ET_INCLUDE( entry.tcl );

     ET_INCLUDE( focus.tcl );

     ET_INCLUDE( listbox.tcl );

     ET_INCLUDE( menu.tcl );

     ET_INCLUDE( obsolete.tcl );

     ET_INCLUDE( optionMenu.tcl );

     ET_INCLUDE( palette.tcl );

     ET_INCLUDE( parray.tcl );

     ET_INCLUDE( text.tcl );

     ET_INCLUDE( scale.tcl );

     ET_INCLUDE( scrollbar.tcl );

     ET_INCLUDE( tearoff.tcl );

     ET_INCLUDE( tkerror.tcl );

     ET( rename source {}; rename __source__ source );

It is because of these 17 ET_INCLUDE macros that the et.c file must be

preprocessed by et2c before being compiled into et.o.

5.  Example 4: etwish

The short code that follows implements the interactive Tcl/Tk shell "

wish" using ET:

  main(int argc, char **argv){

    Et_Init(&argc,argv);

    Et_ReadStdin();

    Et_MainLoop();

  }

This program illustrates the use of Et_ReadStdin() routine.  The

Et_ReadStdin() routine causes ET to monitor standard input, and to

interpret all characters received as Tcl/Tk commands.  This is, of

course, the essential function of the interactive Tcl/Tk shell.

The program generated by this code example differs from the standard

wish program in two important ways.

  *  In the example here, the Tcl/Tk startup scripts are bound to the

     executable at compile-time, but in the standard wish they are

     read into the executable at run-time.

  *  This example does not support the -f command line switch that

     will cause wish to take its input from a file instead of from

     standard input.

6.  Example 5: runscript

The next example implements a version of wish that takes its input

from a file instead of from standard input.  The file that is read

must reside in the same directory as the executable and must have the

same name as the executable but with the addition of a .tcl suffix.

For instance, if the executable that results from compiling the

following program is named fuzzy, then the result of executing fuzzy

is that the Tcl/Tk script found in the same directory as fuzzy and

named fuzzy.tcl is read and executed.

  void

  main(int argc, char **argv){

    Et_Init(&argc,argv);

    ET( source $cmd_dir/$cmd_name.tcl );

    Et_MainLoop();

  }

6.1 The $cmd_dir and $cmd_name variables

The operation of the runscript program depends on the existence of two

Tcl/Tk variables computed by Et_Init() and named cmd_dir and cmd_name.

The cmd_dir variable stores the name of the directory that holds the

currently running executable.  The cmd_name variables stores the base

name of the executable.  The cmd_name and especially the cmd_dir

variables are included as a standard part of ET in order to encourage

people to write programs that do not use hard-coded absolute

pathnames.

In most modern operating systems, a file can have two kinds of names:

absolute and relative.  An absolute pathname means the name of a file

relative to the root directory of the filesystem.  A relative

pathname, on the other hand, describes a file relative to some other

reference directory, usually the working directory.  Experience has

shown that it is generally bad style to hard-code absolute pathnames

into a program.

The cmd_dir variable helps programmers to avoid hard-coded absolute

pathnames by allowing them to locate auxiliary files relative to the

executable.  For example, if a program named acctrec needs to access a

data file named acctrec.db then it can do so be look for acctrec.db in

a directory relative to the directory that contains acctrec.  The