abu@software-lab.de

PicoLisp Application Development

(c) Software Lab. Alexander Burger

This document presents an introduction to writing browser-based applications in PicoLisp.

It concentrates on the XHTML/CSS GUI-Framework (as opposed to the previous Java-AWT, Java-Swing and Plain-HTML frameworks), which is easier to use, more flexible in layout design, and does not depend on plug-ins, JavaScript, cookies or CSS.

A plain HTTP/HTML GUI has various advantages: It runs on any browser, and can be fully driven by scripts ("@lib/scrape.l").

To be precise: CSS can be used to enhance the layout. And browsers with JavaScript will respond faster and smoother. But this framework works just fine in browsers which do not know anything about CSS or JavaScript. All examples were also tested using the w3m text browser.

For basic informations about the PicoLisp system please look at the PicoLisp Reference and the PicoLisp Tutorial. Knowledge of HTML, and a bit of CSS and HTTP is assumed.

The examples assume that PicoLisp was started from a global installation (see Installation).


Static Pages

You can use PicoLisp to generate static HTML pages. This does not make much sense in itself, because you could directly write HTML code as well, but it forms the base for interactive applications, and allows us to introduce the application server and other fundamental concepts.


Hello World

To begin with a minimal application, please enter the following two lines into a generic source file named "project.l" in the PicoLisp installation directory.


########################################################################
(html 0 "Hello" "@lib.css" NIL
   "Hello World!" )
########################################################################

(We will modify and use this file in all following examples and experiments. Whenever you find such a program snippet between hash ('#') lines, just copy and paste it into your "project.l" file, and press the "reload" button of your browser to view the effects)

Start the application server

Open a second terminal window, and start a PicoLisp application server


$ pil @lib/http.l @lib/xhtml.l @lib/form.l  --server 8080 project.l  +

No prompt appears. The server just sits, and waits for connections. You can stop it later by hitting Ctrl-C in that terminal, or by executing 'killall pil' in some other window.

(In the following, we assume that this HTTP server is up and running)

Now open the URL 'http://localhost:8080' with your browser. You should see an empty page with a single line of text.

How does it work?

The above line loads the debugger (via the '+' switch), the HTTP server code ("@lib/http.l"), the XHTML functions ("@lib/xhtml.l") and the input form framework ("@lib/form.l", it will be needed later for interactive forms).

Then the server function is called with a port number and a default URL. It will listen on that port for incoming HTTP requests in an endless loop. Whenever a GET request arrives on port 8080, the file "project.l" will be (load)ed, causing the evaluation (= execution) of all its Lisp expressions.

During that execution, all data written to the current output channel is sent directly to to the browser. The code in "project.l" is responsible to produce HTML (or anything else the browser can understand).


URL Syntax

The PicoLisp application server uses a slightly specialized syntax when communicating URLs to and from a client. The "path" part of an URL - which remains when

are stripped off - is interpreted according so some rules. The most prominent ones are:

An application is free to extend or modify the *Mimes table with the mime function. For example


(mime "doc" "application/msword" 60)

defines a new mime type with a max-age of one minute.

Argument values in URLs, following the path and the question mark, are encoded in such a way that Lisp data types are preserved:

In that way, high-level data types can be directly passed to functions encoded in the URL, or assigned to global variables before a file is loaded.


Security

It is, of course, a huge security hole that - directly from the URL - any Lisp source file can be loaded, and any Lisp function can be called. For that reason, applications must take care to declare exactly which files and functions are to be allowed in URLs. The server checks a global variable *Allow, and - when its value is non-NIL - denies access to anything that does not match its contents.

Normally, *Allow is not manipulated directly, but set with the allowed and allow functions


(allowed ("app/")
   "!start" "!stop" "@lib.css" "!psh" )

This is usually called at the beginning of an application, and allows access to the directory "app/", to the functions 'start', 'stop' and 'psh', and to the file "@lib.css".

Later in the program, *Allow may be dynamically extended with allow


(allow "!foo")
(allow "newdir/" T)

This adds the function 'foo', and the directory "newdir/", to the set of allowed items.

The ".pw" File

For a variety of security checks (most notably for using the psh function, as in some later examples) it is necessary to create a file named ".pw" in the PicoLisp installation directory. This file should contain a single line of arbitrary data, to be used as a password for identifying local resources.

The recommeded way to create this file is to call the pw function, defined in "@lib/http.l"


$ pil @lib/http.l -'pw 12' -bye

Please execute this command.


The html Function

Now back to our "Hello World" example. In principle, you could write "project.l" as a sequence of print statements


########################################################################
(prinl "HTTP/1.0 200 OK^M")
(prinl "Content-Type: text/html; charset=utf-8")
(prinl "^M")
(prinl "<html>")
(prinl "Hello World!")
(prinl "</html>")
########################################################################

but using the html function is much more convenient.

Moreover, html is nothing more than a printing function. You can see this easily if you connect a PicoLisp Shell (psh) to the server process (you must have generated a ".pw" file for this), and enter the html statement


$ /usr/lib/picolisp/bin/psh 8080
: (html 0 "Hello" "@lib.css" NIL "Hello World!")
HTTP/1.0 200 OK
Server: PicoLisp
Date: Fri, 29 Dec 2006 07:28:58 GMT
Cache-Control: max-age=0
Cache-Control: no-cache
Content-Type: text/html; charset=utf-8

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head>
<title>Hello</title>
<base href="http://localhost:8080/"/>
<link rel="stylesheet" type="text/css" href="http://localhost:8080/@lib.css"/>
</head>
<body>Hello World!</body>
</html>
-> </html>
:  # (type Ctrl-D here to terminate PicoLisp)

These are the arguments to html:

  1. 0: A max-age value for cache-control (in seconds, zero means "no-cache"). You might pass a higher value for pages that change seldom, or NIL for no cache-control at all.
  2. "Hello": The page title.
  3. "@lib.css": A CSS-File name. Pass NIL if you do not want to use any CSS-File, or a list of file names if you want to give more than one CSS-File.
  4. NIL: A CSS style attribute specification (see the description of CSS Attributes below). It will be passed to the body tag.

After these four arguments, an arbitrary number of expressions may follow. They form the body of the resulting page, and are evaluated according to a special rule. This rule is slightly different from the evaluation of normal Lisp expressions:

Therefore, our source file might as well be written as:


########################################################################
(html 0 "Hello" "@lib.css" NIL
   (prinl "Hello World!") )
########################################################################

The most typical print statements will be some HTML-tags:


########################################################################
(html 0 "Hello" "@lib.css" NIL
   (<h1> NIL "Hello World!")
   (<br> "This is some text.")
   (ht:Prin "And this is a number: " (+ 1 2 3)) )
########################################################################

<h1> and <br> are tag functions. <h1> takes a CSS attribute as its first argument.

Note the use of ht:Prin instead of prin. ht:Prin should be used for all direct printing in HTML pages, because it takes care to escape special characters.


CSS Attributes

The html function above, and many of the HTML tag functions, accept a CSS attribute specification. This may be either an atom, a cons pair, or a list of cons pairs. We demonstrate the effects with the <h1> tag function.

An atom (usually a symbol or a string) is taken as a CSS class name


: (<h1> 'foo "Title")
<h1 class="foo">Title</h1>

For a cons pair, the CAR is taken as an attribute name, and the CDR as the attribute's value


: (<h1> '(id . bar) "Title")
<h1 id="bar">Title</h1>

Consequently, a list of cons pairs gives a set of attribute-value pairs


: (<h1> '((id . "abc") (lang . "de")) "Title")
<h1 id="abc" lang="de">Title</h1>


Tag Functions

All pre-defined XHTML tag functions can be found in "@lib/xhtml.l". We recommend to look at their sources, and to experiment a bit, by executing them at a PicoLisp prompt, or by pressing the browser's "Reload" button after editing the "project.l" file.

For a suitable PicoLisp prompt, either execute (in a separate terminal window) the PicoLisp Shell (psh) command (works only if the application server is running, and you did generate a ".pw" file)


$ /usr/lib/picolisp/bin/psh 8080
:

or start the interpreter stand-alone, with "@lib/xhtml.l" loaded


$ pil @lib/http.l @lib/xhtml.l +
:

Note that for all these tag functions the above tag body evaluation rule applies.

Simple Tags

Most tag functions are simple and straightforward. Some of them just print their arguments


: (<br> "Hello world")
Hello world<br/>

: (<em> "Hello world")
<em>Hello world</em>

while most of them take a CSS attribute specification as their first argument (like the <h1> tag above)


: (<div> 'main "Hello world")
<div class="main">Hello world</div>

: (<p> NIL "Hello world")
<p>Hello world</p>

: (<p> 'info "Hello world")
<p class="info">Hello world</p>

All of these functions take an arbitrary number of arguments, and may nest to an arbitrary depth (as long as the resulting HTML is legal)


: (<div> 'main
   (<h1> NIL "Head")
   (<p> NIL
      (<br> "Line 1")
      "Line"
      (<nbsp>)
      (+ 1 1) ) )
<div class="main"><h1>Head</h1>
<p>Line 1<br/>
Line 2</p>
</div>

(Un)ordered Lists

HTML-lists, implemented by the <ol> and <ul> tags, let you define hierarchical structures. You might want to paste the following code into your copy of "project.l":


########################################################################
(html 0 "Unordered List" "@lib.css" NIL
   (<ul> NIL
      (<li> NIL "Item 1")
      (<li> NIL
         "Sublist 1"
         (<ul> NIL
            (<li> NIL "Item 1-1")
            (<li> NIL "Item 1-2") ) )
      (<li> NIL "Item 2")
      (<li> NIL
         "Sublist 2"
         (<ul> NIL
            (<li> NIL "Item 2-1")
            (<li> NIL "Item 2-2") ) )
      (<li> NIL "Item 3") ) )
########################################################################

Here, too, you can put arbitrary code into each node of that tree, including other tag functions.

Tables

Like the hierarchical structures with the list functions, you can generate two-dimensional tables with the <table> and <row> functions.

The following example prints a table of numbers and their squares:


########################################################################
(html 0 "Table" "@lib.css" NIL
   (<table> NIL NIL NIL
      (for N 10                                    # A table with 10 rows
         (<row> NIL N (prin (* N N))) ) ) )     # and 2 columns
########################################################################

The first argument to <table> is the usual CSS attribute, the second an optional title ("caption"), and the third an optional list specifying the column headers. In that list, you may supply a list for a each column, with a CSS attribute in its CAR, and a tag body in its CDR for the contents of the column header.

The body of <table> contains calls to the <row> function. This function is special in that each expression in its body will go to a separate column of the table. If both for the column header and the row function an CSS attribute is given, they will be combined by a space and passed to the HTML <td> tag. This permits distinct CSS specifications for each column and row.

As an extension of the above table example, let's pass some attributes for the table itself (not recommended - better define such styles in a CSS file and then just pass the class name to <table>), right-align both columns, and print each row in an alternating red and blue color


########################################################################
(html 0 "Table" "@lib.css" NIL
   (<table>
      '((width . "200px") (style . "border: dotted 1px;"))    # table style
      "Square Numbers"                                        # caption
      '((align "Number") (align "Square"))                    # 2 headers
      (for N 10                                                  # 10 rows
         (<row> (xchg '(red) '(blue))                         # red or blue
            N                                                 # 2 columns
            (prin (* N N) ) ) ) ) )
########################################################################

If you wish to concatenate two or more cells in a table, so that a single cell spans several columns, you can pass the symbol '-' for the additional cell data to <row>. This will cause the data given to the left of the '-' symbols to expand to the right.

You can also directly specify table structures with the simple <th>, <tr> and <td> tag functions.

If you just need a two-dimensional arrangement of components, the even simpler <grid> function might be convenient:


########################################################################
(html 0 "Grid" "@lib.css" NIL
   (<grid> 3
      "A" "B" "C"
      123 456 789 ) )
########################################################################

It just takes a specification for the number of columns (here: 3) as its first argument, and then a single expression for each cell. Instead of a number, you can also pass a list of CSS attributes. Then the length of that list will determine the number of columns. You can change the second line in the above example to


   (<grid> '(NIL NIL right)

Then the third column will be right aligned.

Menus and Tabs

The two most powerful tag functions are <menu> and <tab>. Used separately or in combination, they form a navigation framework with

The following example is not very useful, because the URLs of all items link to the same "project.l" page, but it should suffice to demonstrate the functionality:


########################################################################
(html 0 "Menu+Tab" "@lib.css" NIL
   (<div> '(id . menu)
      (<menu>
         ("Item" "project.l")                      # Top level item
         (NIL (<hr>))                              # Plain HTML
         (T "Submenu 1"                            # Submenu
            ("Subitem 1.1" "project.l")
            (T "Submenu 1.2"
               ("Subitem 1.2.1" "project.l")
               ("Subitem 1.2.2" "project.l")
               ("Subitem 1.2.3" "project.l") )
            ("Subitem 1.3" "project.l") )
         (T "Submenu 2"
            ("Subitem 2.1" "project.l")
            ("Subitem 2.2" "project.l") ) ) )
   (<div> '(id . main)
      (<h1> NIL "Menu+Tab")
      (<tab>
         ("Tab1"
            (<h3> NIL "This is Tab 1") )
         ("Tab2"
            (<h3> NIL "This is Tab 2") )
         ("Tab3"
            (<h3> NIL "This is Tab 3") ) ) ) )
########################################################################

<menu> takes a sequence of menu items. Each menu item is a list, with its CAR either

<tab> takes a list of subpages. Each page is simply a tab name, followed by arbitrary code (typically HTML tags).

Note that only a single menu and a single tab may be active at the same time.


Interactive Forms

In HTML, the only possibility for user input is via <form> and <input> elements, using the HTTP POST method to communicate with the server.

"@lib/xhtml.l" defines a function called <post>, and a collection of input tag functions, which allow direct programming of HTML forms. We will supply only one simple example:


########################################################################
(html 0 "Simple Form" "@lib.css" NIL
   (<post> NIL "project.l"
      (<field> 10 '*Text)
      (<submit> "Save") ) )
########################################################################

This associates a text input field with a global variable *Text. The field displays the current value of *Text, and pressing the submit button causes a reload of "project.l" with *Text set to any string entered by the user.

An application program could then use that variable to do something useful, for example store its value in a database.

The problem with such a straightforward use of forms is that

  1. they require the application programmer to take care of maintaining lots of global variables. Each input field on the page needs an associated variable for the round trip between server and client.
  2. they do not preserve an application's internal state. Each POST request spawns an individual process on the server, which sets the global variables to their new values, generates the HTML page, and terminates thereafter. The application state has to be passed along explicitly, e.g. using <hidden> tags.
  3. they are not very interactive. There is typically only a single submit button. The user fills out a possibly large number of input fields, but changes will take effect only when the submit button is pressed.

Though we wrote a few applications in that style, we recommend the GUI framework provided by "@lib/form.l". It does not need any variables for the client/server communication, but implements a class hierarchy of GUI components for the abstraction of application logic, button actions and data linkage.


Sessions

First of all, we need to establish a persistent environment on the server, to handle each individual session (for each connected client).

Technically, this is just a child process of the server we started above, which does not terminate immediately after it sent its page to the browser. It is achieved by calling the app function somewhere in the application's startup code.


########################################################################
(app)  # Start a session

(html 0 "Simple Session" "@lib.css" NIL
   (<post> NIL "project.l"
      (<field> 10 '*Text)
      (<submit> "Save") ) )
########################################################################

Nothing else changed from the previous example. However, when you connect your browser and then look at the terminal window where you started the application server, you'll notice a colon, the PicoLisp prompt


$ pil @lib/http.l @lib/xhtml.l @lib/form.l  --server 8080 project.l  +
:

Tools like the Unix ps utility will tell you that now two picolisp processes are running, the first being the parent of the second.

If you enter some text, say "abcdef", into the text field in the browser window, press the submit button, and inspect the Lisp *Text variable,


: *Text
-> "abcdef"

you see that we now have a dedicated PicoLisp process, "connected" to the client.

You can terminate this process (like any interactive PicoLisp) by hitting Ctrl-D on an empty line. Otherwise, it will terminate by itself if no other browser requests arrive within a default timeout period of 5 minutes.

To start a (non-debug) production version, the server is commonly started without the '+' flag, and with -wait


$ pil @lib/http.l @lib/xhtml.l @lib/form.l  --server 8080 project.l  -wait

In that way, no command line prompt appears when a client connects.


Action Forms

Now that we have a persistent session for each client, we can set up an active GUI framework.

This is done by wrapping the call to the html function with action. Inside the body of html can be - in addition to all other kinds of tag functions - one or more calls to form


########################################################################
(app)                                              # Start session

(action                                            # Action handler
   (html 0 "Form" "@lib.css" NIL                   # HTTP/HTML protocol
      (form NIL                                    # Form
         (gui 'a '(+TextField) 10)                 # Text Field
         (gui '(+Button) "Print"                   # Button
            '(msg (val> (: home a))) ) ) ) )
########################################################################

Note that there is no longer a global variable like *Text to hold the contents of the input field. Instead, we gave a local, symbolic name 'a' to a +TextField component


         (gui 'a '(+TextField) 10)                 # Text Field

Other components can refer to it


            '(msg (val> (: home a)))

(: home) is always the form which contains this GUI component. So (: home a) evaluates to the component 'a' in the current form. As msg prints its argument to standard error, and the val> method retrieves the current contents of a component, we will see on the console the text typed into the text field when we press the button.

An action without embedded forms - or a form without a surrounding action - does not make much sense by itself. Inside html and form, however, calls to HTML functions (and any other Lisp functions, for that matter) can be freely mixed.

In general, a typical page may have the form


(action                                            # Action handler
   (html ..                                        # HTTP/HTML protocol
      (<h1> ..)                                    # HTML tags
      (form NIL                                    # Form
         (<h3> ..)
         (gui ..)                                  # GUI component(s)
         (gui ..)
         .. )
      (<h2> ..)
      (form NIL                                    # Another form
         (<h3> ..)
         (gui ..)                                  # GUI component(s)
         .. )
      (<br> ..)
      .. ) )

The gui Function

The most prominent function in a form body is gui. It is the workhorse of GUI construction.

Outside of a form body, gui is undefined. Otherwise, it takes an optional alias name, a list of classes, and additional arguments as needed by the constructors of these classes. We saw this example before


         (gui 'a '(+TextField) 10)                 # Text Field
Here, 'a' is an alias name for a component of type (+TextField). The numeric argument 10 is passed to the text field, specifying its width. See the chapter on GUI Classes for more examples.

During a GET request, gui is basically a front-end to new. It builds a component, stores it in the internal structures of the current form, and initializes it by sending the init> message to the component. Finally, it sends it the show> message, to produce HTML code and transmit it to the browser.

During a POST request, gui does not build any new components. Instead, the existing components are re-used. So gui does not have much more to do than sending the show> message to a component.

Control Flow

HTTP has only two methods to change a browser window: GET and POST. We employ these two methods in a certain defined, specialized way:

A button's action code can do almost anything: Read and modify the contents of input fields, communicate with the database, display alerts and dialogs, or even fake the POST request to a GET, with the effect of showing a completely different document (See Switching URLs).

GET builds up all GUI components on the server. These components are objects which encapsulate state and behavior of the HTML page in the browser. Whenever a button is pressed, the page is reloaded via a POST request. Then - before any output is sent to the browser - the action function takes control. It performs error checks on all components, processes possible user input on the HTML page, and stores the values in correct format (text, number, date, object etc.) in each component.

The state of a form is preserved over time. When the user returns to a previous page with the browser's BACK button, that state is reactivated, and may be POSTed again.

The following silly example displays two text fields. If you enter some text into the "Source" field, you can copy it in upper or lower case to the "Destination" field by pressing one of the buttons


########################################################################
(app)

(action
   (html 0 "Case Conversion" "@lib.css" NIL
      (form NIL
         (<grid> 2
            "Source" (gui 'src '(+TextField) 30)
            "Destination" (gui 'dst '(+Lock +TextField) 30) )
         (gui '(+JS +Button) "Upper Case"
            '(set> (: home dst)
               (uppc (val> (: home src))) ) )
         (gui '(+JS +Button) "Lower Case"
            '(set> (: home dst)
               (lowc (val> (: home src))) ) ) ) ) )
########################################################################

The +Lock prefix class in the "Destination" field makes that field read-only. The only way to get some text into that field is by using one of the buttons.

Switching URLs

Because an action code runs before html has a chance to output an HTTP header, it can abort the current page and present something different to the user. This might, of course, be another HTML page, but would not be very interesting as a normal link would suffice. Instead, it can cause the download of dynamically generated data.

The next example shows a text area and two buttons. Any text entered into the text area is exported either as a text file via the first button, or a PDF document via the second button


########################################################################
(load "@lib/ps.l")

(app)

(action
   (html 0 "Export" "@lib.css" NIL
      (form NIL
         (gui '(+TextField) 30 8)
         (gui '(+Button) "Text"
            '(let Txt (tmp "export.txt")
               (out Txt (prinl (val> (: home gui 1))))
               (url Txt) ) )
         (gui '(+Button) "PDF"
            '(psOut NIL "foo"
               (a4)
               (indent 40 40)
               (down 60)
               (hline 3)
               (font (14 . "Times-Roman")
                  (ps (val> (: home gui 1))) )
               (hline 3)
               (page) ) ) ) ) )
########################################################################

(a text area is built when you supply two numeric arguments (columns and rows) to a +TextField class)

The action code of the first button creates a temporary file (i.e. a file named "export.txt" in the current process's temporary space), prints the value of the text area (this time we did not bother to give it a name, we simply refer to it as the form's first gui list element) into that file, and then calls the url function with the file name.

The second button uses the PostScript library "@lib/ps.l" to create a temporary file "foo.pdf". Here, the temporary file creation and the call to the url function is hidden in the internal mechanisms of psOut. The effect is that the browser receives a PDF document and displays it.

Alerts and Dialogs

Alerts and dialogs are not really what they used to be ;-)

They do not "pop up". In this framework, they are just a kind of simple-to-use, pre-fabricated form. They can be invoked by a button's action code, and appear always on the current page, immediately preceding the form which created them.

Let's look at an example which uses two alerts and a dialog. In the beginning, it displays a simple form, with a locked text field, and two buttons


########################################################################
(app)

(action
   (html 0 "Alerts and Dialogs" "@lib.css" NIL
      (form NIL
         (gui '(+Init +Lock +TextField) "Initial Text" 20 "My Text")
         (gui '(+Button) "Alert"
            '(alert NIL "This is an alert " (okButton)) )
         (gui '(+Button) "Dialog"
            '(dialog NIL
               (<br> "This is a dialog.")
               (<br>
                  "You can change the text here "
                  (gui '(+Init +TextField) (val> (: top 1 gui 1)) 20) )
               (<br> "and then re-submit it to the form.")
               (gui '(+Button) "Re-Submit"
                  '(alert NIL "Are you sure? "
                     (yesButton
                        '(set> (: home top 2 gui 1)
                           (val> (: home top 1 gui 1)) ) )
                     (noButton) ) )
               (cancelButton) ) ) ) ) )
########################################################################

The +Init prefix class initializes the "My Text" field with the string "Initial Text". As the field is locked, you cannot modify this value directly.

The first button brings up an alert saying "This is an alert.". You can dispose it by pressing "OK".

The second button brings up a dialog with an editable text field, containing a copy of the value from the form's locked text field. You can modify this value, and send it back to the form, if you press "Re-Submit" and answer "Yes" to the "Are you sure?" alert.

A Calculator Example

Now let's forget our "project.l" test file for a moment, and move on to a more substantial and practical, stand-alone, example. Using what we have learned so far, we want to build a simple bignum calculator. ("bignum" because PicoLisp can do only bignums)

It uses a single form, a single numeric input field, and lots of buttons. It can be found in the PicoLisp distribution (e.g. under "/usr/share/picolisp/") in "misc/calc.l", together with a directly executable wrapper script "misc/calc".

To use it, change to the PicoLisp installation directory, and start it as


$ misc/calc

or call it with an absolute path, e.g.


$ /usr/share/picolisp/misc/calc

If you like to get a PicoLisp prompt for inspection, start it instead as


$ pil misc/calc.l -main -go +

Then - as before - point your browser to 'http://localhost:8080'.

The code for the calculator logic and the GUI is rather straightforward. The entry point is the single function calculator. It is called directly (as described in URL Syntax) as the server's default URL, and implicitly in all POST requests. No further file access is needed once the calculator is running.

Note that for a production application, we inserted an allow-statement (as recommended by the Security chapter)


(allowed NIL "!calculator" "@lib.css")

at the beginning of "misc/calc.l". This will restrict external access to that single function.

The calculator uses three global variables, *Init, *Accu and *Stack. *Init is a boolean flag set by the operator buttons to indicate that the next digit should initialize the accumulator to zero. *Accu is the accumulator. It is always displayed in the numeric input field, accepts user input, and it holds the results of calculations. *Stack is a push-down stack, holding postponed calculations (operators, priorities and intermediate results) with lower-priority operators, while calculations with higher-priority operators are performed.

The function digit is called by the digit buttons, and adds another digit to the accumulator.

The function calc does an actual calculation step. It pops the stack, checks for division by zero, and displays an error alert if necessary.

operand processes an operand button, accepting a function and a priority as arguments. It compares the priority with that in the top-of-stack element, and delays the calculation if it is less.

finish is used to calculate the final result.

The calculator function has one numeric input field, with a width of 60 characters


         (gui '(+Var +NumField) '*Accu 60)

The +Var prefix class associates this field with the global variable *Accu. All changes to the field will show up in that variable, and modification of that variable's value will appear in the field.

The square root operator button has an +Able prefix class


         (gui '(+Able +JS +Button) '(ge0 *Accu) (char 8730)
            '(setq *Accu (sqrt *Accu)) )

with an argument expression which checks that the current value in the accumulator is positive, and disables the button if otherwise.

The rest of the form is just an array (grid) of buttons, encapsulating all functionality of the calculator. The user can enter numbers into the input field, either by using the digit buttons, or by directly typing them in, and perform calculations with the operator buttons. Supported operations are addition, subtraction, multiplication, division, sign inversion, square root and power (all in bignum integer arithmetic). The 'C' button just clears the accumulator, while the 'A' button also clears all pending calculations.

All that in 53 lines of code!


Charts

Charts are virtual components, maintaining the internal representation of two-dimensional data.

Typically, these data are nested lists, database selections, or some kind of dynamically generated tabular information. Charts make it possible to view them in rows and columns (usually in HTML tables), scroll up and down, and associate them with their corresponding visible GUI components.

In fact, the logic to handle charts makes up a substantial part of the whole framework, with large impact on all internal mechanisms. Each GUI component must know whether it is part of a chart or not, to be able to handle its contents properly during updates and user interactions.

Let's assume we want to collect textual and numerical data. We might create a table


########################################################################
(app)

(action
   (html 0 "Table" "@lib.css" NIL
      (form NIL
         (<table> NIL NIL '((NIL "Text") (NIL "Number"))
            (do 4
               (<row> NIL
                  (gui '(+TextField) 20)
                  (gui '(+NumField) 10) ) ) )
         (<submit> "Save") ) ) )
########################################################################

with two columns "Text" and "Number", and four rows, each containing a +TextField and a +NumField.

You can enter text into the first column, and numbers into the second. Pressing the "Save" button stores these values in the components on the server (or produces an error message if a string in the second column is not a legal number).

There are two problems with this solution:

  1. Though you can get at the user input for the individual fields, e.g.
    
    : (val> (get *Top 'gui 2))  # Value in the first row, second column
    -> 123
    
    there is no direct way to get the whole data structure as a single list. Instead, you have to traverse all GUI components and collect the data.
  2. The user cannot input more than four rows of data, because there is no easy way to scroll down and make space for more.

A chart can handle these things:


########################################################################
(app)

(action
   (html 0 "Chart" "@lib.css" NIL
      (form NIL
         (gui '(+Chart) 2)                         # Inserted a +Chart
         (<table> NIL NIL '((NIL "Text") (NIL "Number"))
            (do 4
               (<row> NIL
                  (gui 1 '(+TextField) 20)         # Inserted '1'
                  (gui 2 '(+NumField) 10) ) ) )    # Inserted '2'
         (<submit> "Save") ) ) )
########################################################################

Note that we inserted a +Chart component before the GUI components which should be managed by the chart. The argument '2' tells the chart that it has to expect two columns.

Each component got an index number (here '1' and '2') as the first argument to gui, indicating the column into which this component should go within the chart.

Now - if you entered "a", "b" and "c" into the first, and 1, 2, and 3 into the second column - we can retrieve the chart's complete contents by sending it the val> message


: (val> (get *Top 'chart 1))  # Retrieve the value of the first chart
-> (("a" 1) ("b" 2) ("c" 3))

BTW, a more convenient function is chart


: (val> (chart))  # Retrieve the value of the current chart
-> (("a" 1) ("b" 2) ("c" 3))

chart can be used instead of the above construct when we want to access the "current" chart, i.e. the chart most recently processed in the current form.

Scrolling

To enable scrolling, let's also insert two buttons. We use the pre-defined classes +UpButton and +DnButton


########################################################################
(app)

(action
   (html 0 "Scrollable Chart" "@lib.css" NIL
      (form NIL
         (gui '(+Chart) 2)
         (<table> NIL NIL '((NIL "Text") (NIL "Number"))
            (do 4
               (<row> NIL
                  (gui 1 '(+TextField) 20)
                  (gui 2 '(+NumField) 10) ) ) )
         (gui '(+UpButton) 1)                   # Inserted two buttons
         (gui '(+DnButton) 1)
         (----)
         (<submit> "Save") ) ) )
########################################################################

to scroll down and up a single (argument '1') line at a time.

Now it is possible to enter a few rows of data, scroll down, and continue. It is not necessary (except in the beginning, when the scroll buttons are still disabled) to press the "Save" button, because any button in the form will send changes to the server's internal structures before any action is performed.

Put and Get Functions

As we said, a chart is a virtual component to edit two-dimensional data. Therefore, a chart's native data format is a list of lists: Each sublist represents a single row of data, and each element of a row corresponds to a single GUI component.

In the example above, we saw a row like


   ("a" 1)

being mapped to


   (gui 1 '(+TextField) 20)
   (gui 2 '(+NumField) 10)

Quite often, however, such a one-to-one relationship is not desired. The internal data structures may have to be presented in a different form to the user, and user input may need conversion to an internal representation.

For that, a chart accepts - in addition to the "number of columns" argument - two optional function arguments. The first function is invoked to 'put' the internal representation into the GUI components, and the second to 'get' data from the GUI into the internal representation.

A typical example is a chart displaying customers in a database. While the internal representation is a (one-dimensional) list of customer objects, 'put' expands each object to a list with, say, the customer's first and second name, telephone number, address and so on. When the user enters a customer's name, 'get' locates the matching object in the database and stores it in the internal representation. In the following, 'put' will in turn expand it to the GUI.

For now, let's stick with a simpler example: A chart that holds just a list of numbers, but expands in the GUI to show also a textual form of each number (in German).


########################################################################
(app)

(load "@lib/zahlwort.l")

(action
   (html 0 "Numerals" "@lib.css" NIL
      (form NIL
         (gui '(+Init +Chart) (1 5 7) 2
            '((N) (list N (zahlwort N)))
            car )
         (<table> NIL NIL '((NIL "Numeral") (NIL "German"))
            (do 4
               (<row> NIL
                  (gui 1 '(+NumField) 9)
                  (gui 2 '(+Lock +TextField) 90) ) ) )
         (gui '(+UpButton) 1)
         (gui '(+DnButton) 1)
         (----)
         (<submit> "Save") ) ) )
########################################################################

"@lib/zahlwort.l" defines the utility function zahlwort, which is required later by the 'put' function. zahlwort accepts a number and returns its wording in German.

Now look at the code


         (gui '(+Init +Chart) (1 5 7) 2
            '((N) (list N (zahlwort N)))
            car )

We prefix the +Chart class with +Init, and pass it a list of numbers (1 5 7) for the initial value of the chart. Then, following the '2' (the chart has two columns), we pass a 'put' function


            '((N) (list N (zahlwort N)))

which takes a number and returns a list of that number and its wording, and a 'get' function


            car )

which in turn accepts such a list and returns a number, which happens to be the list's first element.

You can see from this example that 'get' is the inverse function of 'put'. 'get' can be omitted, however, if the chart is read-only (contains no (or only locked) input fields).

The field in the second column


                  (gui 2 '(+Lock +TextField) 90) ) ) )

is locked, because it displays the text generated by 'put', and is not supposed to accept any user input.

When you start up this form in your browser, you'll see three pre-filled lines with "1/eins", "5/fünf" and "7/sieben", according to the +Init argument (1 5 7). Typing a number somewhere into the first column, and pressing ENTER or one of the buttons, will show a suitable text in the second column.


GUI Classes

In previous chapters we saw examples of GUI classes like +TextField, +NumField or +Button, often in combination with prefix classes like +Lock, +Init or +Able. Now we take a broader look at the whole hierarchy, and try more examples.

The abstract class +gui is the base of all GUI classes. A live view of the class hierarchy can be obtained with the dep ("dependencies") function:


: (dep '+gui)
+gui
   +Img
   +field
      +Radio
      +TextField
         +UpField
         +PwField
         +BlobField
         +FileField
         +TimeField
         +DateField
         +MailField
         +AtomField
         +HttpField
         +LinesField
         +ClassField
         +TelField
         +numField
            +NumField
            +FixField
         +SymField
         +SexField
         +ListTextField
      +Checkbox
   +Button
      +BubbleButton
      +DelRowButton
      +DnButton
      +GoButton
      +UpButton
      +ChoButton
         +Choice
      +ClrButton
      +PickButton
         +DstButton
      +todoButton
         +RedoButton
         +UndoButton
      +ShowButton
   +JsField
-> +gui

We see, for example, that +DnButton is a subclass of +Button, which in turn is a subclass of +gui. Inspecting +DnButton directly


: (dep '+DnButton)
   +Tiny
   +Rid
   +JS
   +Able
      +gui
   +Button
+DnButton
-> +DnButton

shows that +DnButton inherits from +Tiny, +Rid, +Able and +Button. The actual definition of +DnButton can be found in "@lib/form.l"


(class +DnButton +Tiny +Rid +JS +Able +Button)
...

In general, "@lib/form.l" is the ultimate reference to the framework, and should be freely consulted.


Input Fields

Input fields implement the visual display of application data, and allow - when enabled - input and modification of these data.

On the HTML level, they can take the form of

Except for checkboxes, which are implemented by the Checkbox class, all these HTML representations are generated by +TextField and its content-specific subclasses like +NumField, +DateField etc. Their actual appearance (as one of the above forms) depends on their arguments:

We saw already "normal" text fields. They are created with a single numeric argument. This example creates an editable field with a width of 10 characters:


   (gui '(+TextField) 10)

If you supply a second numeric for the line count ('4' in this case), you'll get a text area:


   (gui '(+TextField) 10 4)

Supplying a list of values instead of a count yields a drop-down selection (combo box):


   (gui '(+TextField) '("Value 1" "Value 2" "Value 3"))

In addition to these arguments, you can pass a string. Then the field is created with a label:


   (gui '(+TextField) 10 "Plain")
   (gui '(+TextField) 10 4 "Text Area")
   (gui '(+TextField) '("Value 1" "Value 2" "Value 3") "Selection")

Finally, without any arguments, the field will appear as a plain HTML text:


   (gui '(+TextField))

This makes mainly sense in combination with prefix classes like +Var and +Obj, to manage the contents of these fields, and achieve special behavior as HTML links or scrollable chart values.

Numeric Input Fields

A +NumField returns a number from its val> method, and accepts a number for its set> method. It issues an error message when user input cannot be converted to a number.

Large numbers are shown with a thousands-separator, as determined by the current locale.


########################################################################
(app)

(action
   (html 0 "+NumField" "@lib.css" NIL
      (form NIL
         (gui '(+NumField) 10)
         (gui '(+JS +Button) "Print value"
            '(msg (val> (: home gui 1))) )
         (gui '(+JS +Button) "Set to 123"
            '(set> (: home gui 1) 123) ) ) ) )
########################################################################

A +FixField needs an additional scale factor argument, and accepts/returns scaled fixpoint numbers.

The decimal separator is determined by the current locale.


########################################################################
(app)

(action
   (html 0 "+FixField" "@lib.css" NIL
      (form NIL
         (gui '(+FixField) 3 10)
         (gui '(+JS +Button) "Print value"
            '(msg (format (val> (: home gui 1)) 3)) )
         (gui '(+JS +Button) "Set to 123.456"
            '(set> (: home gui 1) 123456) ) ) ) )
########################################################################

Time & Date

A +DateField accepts and returns a date value.


########################################################################
(app)

(action
   (html 0 "+DateField" "@lib.css" NIL
      (form NIL
         (gui '(+DateField) 10)
         (gui '(+JS +Button) "Print value"
            '(msg (datStr (val> (: home gui 1)))) )
         (gui '(+JS +Button) "Set to \"today\""
            '(set> (: home gui 1) (date)) ) ) ) )
########################################################################

The format displayed to - and entered by - the user depends on the current locale (see datStr and expDat). You can change it, for example to


: (locale "DE" "de")
-> NIL

If no locale is set, the format is YYYY-MM-DD. Some pre-defined locales use patterns like DD.MM.YYYY (DE), YYYY/MM/DD (JP), DD/MM/YYYY (UK), or MM/DD/YYYY (US).

An error is issued when user input does not match the current locale's date format.

Independent from the locale setting, a +DateField tries to expand abbreviated input from the user. A small number is taken as that day of the current month, larger numbers expand to day and month, or to day, month and year:

Similar is the +TimeField. It accepts and returns a time value.


########################################################################
(app)

(action
   (html 0 "+TimeField" "@lib.css" NIL
      (form NIL
         (gui '(+TimeField) 8)
         (gui '(+JS +Button) "Print value"
            '(msg (tim$ (val> (: home gui 1)))) )
         (gui '(+JS +Button) "Set to \"now\""
            '(set> (: home gui 1) (time)) ) ) ) )
########################################################################

When the field width is '8', like in this example, time is displayed in the format HH:MM:SS. Another possible value would be '5', causing +TimeField to display its value as HH:MM.

An error is issued when user input cannot be converted to a time value.

The user may omit the colons. If he inputs just a small number, it should be between '0' and '23', and will be taken as a full hour. '125' expands to "12:05", '124517' to "12:45:17", and so on.

Telephone Numbers

Telephone numbers are represented internally by the country code (without a leading plus sign or zero) followed by the local phone number (ideally separated by spaces) and the phone extension (ideally separated by a hyphen). The exact format of the phone number string is not enforced by the GUI, but further processing (e.g. database searches) normally uses fold for better reproducibility.

To display a phone number, +TelField replaces the country code with a single zero if it is the country code of the current locale, or prepends it with a plus sign if it is a foreign country (see telStr).

For user input, a plus sign or a double zero is simply dropped, while a single leading zero is replaced with the current locale's country code (see expTel).


########################################################################
(app)
(locale "DE" "de")

(action
   (html 0 "+TelField" "@lib.css" NIL
      (form NIL
         (gui '(+TelField) 20)
         (gui '(+JS +Button) "Print value"
            '(msg (val> (: home gui 1))) )
         (gui '(+JS +Button) "Set to \"49 1234 5678-0\""
            '(set> (: home gui 1) "49 1234 5678-0") ) ) ) )
########################################################################

Checkboxes

A +Checkbox is straightforward. User interaction is restricted to clicking it on and off. It accepts boolean (NIL or non-NIL) values, and returns T or NIL.


########################################################################
(app)

(action
   (html 0 "+Checkbox" "@lib.css" NIL
      (form NIL
         (gui '(+Checkbox))
         (gui '(+JS +Button) "Print value"
            '(msg (val> (: home gui 1))) )
         (gui '(+JS +Button) "On"
            '(set> (: home gui 1) T) )
         (gui '(+JS +Button) "Off"
            '(set> (: home gui 1) NIL) ) ) ) )
########################################################################


Field Prefix Classes

A big part of this framework's power is owed to the combinatorial flexibility of prefix classes for GUI- and DB-objects. They allow to surgically override individual methods in the inheritance tree, and can be combined in various ways to achieve any desired behavior.

Technically, there is nothing special about prefix classes. They are just normal classes. They are called "prefix" because they are intended to be written before other classes in a class's or object's list of superclasses.

Usually they take their own arguments for their T method from the list of arguments to the gui function.

Initialization

+Init overrides the init> method for that component. The init> message is sent to a +gui component when the page is loaded for the first time (during a GET request). +Init takes an expression for the initial value of that field.


   (gui '(+Init +TextField) "This is the initial text" 30)

Other classes which automatically give a value to a field are +Var (linking the field to a variable) and +E/R (linking the field to a database entity/relation).

+Cue can be used, for example in "mandatory" fields, to give a hint to the user about what he is supposed to enter. It will display the argument value, in angular brackets, if and only if the field's value is NIL, and the val> method will return NIL despite the fact that this value is displayed.

Cause an empty field to display "<Please enter some text here>":


   (gui '(+Cue +TextField) "Please enter some text here" 30)

Disabling and Enabling

An important feature of an interactive GUI is the context-sensitive disabling and enabling of individual components, or of a whole form.

The +Able prefix class takes an argument expression, and disables the component if this expression returns NIL. We saw an example for its usage already in the square root button of the calculator example. Or, for illustration purposes, imagine a button which is supposed to be enabled only after Christmas


   (gui '(+Able +Button)
      '(>= (cdr (date (date))) (12 24))
      "Close this year"
      '(endOfYearProcessing) )

or a password field that is disabled as long as somebody is logged in


   (gui '(+Able +PwField) '(not *Login) 10 "Password")

A special case is the +Lock prefix, which permanently and unconditionally disables a component. It takes no arguments


   (gui '(+Lock +NumField) 10 "Count")

('10' and "Count" are for the +NumField), and creates a read-only field.

The whole form can be disabled by calling disable with a non-NIL argument. This affects all components in this form. Staying with the above example, we can make the form read-only until Christmas


   (form NIL
      (disable (> (12 24) (cdr (date (date)))))  # Disable whole form
      (gui ..)
      .. )

Even in a completely disabled form, however, it is often necessary to re-enable certain components, as they are needed for navigation, scrolling, or other activities which don't affect the contents of the form. This is done by prefixing these fields with +Rid (i.e. getting "rid" of the lock).


   (form NIL
      (disable (> (12 24) (cdr (date (date)))))
      (gui ..)
      ..
      (gui '(+Rid +Button) ..)  # Button is enabled despite the disabled form
      .. )

Formatting

GUI prefix classes allow a fine-grained control of how values are stored in - and retrieved from - components. As in predefined classes like +NumField or +DateField, they override the set> and/or val> methods.

+Set takes an argument function which is called whenever that field is set to some value. To convert all user input to upper case


   (gui '(+Set +TextField) uppc 30)

+Val is the complement to +Set. It takes a function which is called whenever the field's value is retrieved. To return the square of a field's value


   (gui '(+Val +NumField) '((N) (* N N)) 10)

+Fmt is just a combination of +Set and +Val, and takes two functional arguments. This example will display upper case characters, while returning lower case characters internally


   (gui '(+Fmt +TextField) uppc lowc 30)

+Map does (like +Fmt) a two-way translation. It uses a list of cons pairs for a linear lookup, where the CARs represent the displayed values which are internally mapped to the values in the CDRs. If a value is not found in this list during set> or val>, it is passed through unchanged.

Normally, +Map is used in combination with the combo box incarnation of text fields (see Input Fields). This example displays "One", "Two" and "Three" to the user, but returns a number 1, 2 or 3 internally


########################################################################
(app)

(action
   (html 0 "+Map" "@lib.css" NIL
      (form NIL
         (gui '(+Map +TextField)
            '(("One" . 1) ("Two" . 2) ("Three" . 3))
            '("One" "Two" "Three") )
         (gui '(+Button) "Print"
            '(msg (val> (field -1))) ) ) ) )
########################################################################

Side Effects

Whenever a button is pressed in the GUI, any changes caused by action in the current environment (e.g. the database or application state) need to be reflected in the corresponding GUI fields. For that, the upd> message is sent to all components. Each component then takes appropriate measures (e.g. refresh from database objects, load values from variables, or calculate a new value) to update its value.

While the upd> method is mainly used internally, it can be overridden in existing classes via the +Upd prefix class. Let's print updated values to standard error


########################################################################
(app)
(default *Number 0)

(action
   (html 0 "+Upd" "@lib.css" NIL
      (form NIL
         (gui '(+Upd +Var +NumField)
            '(prog (extra) (msg *Number))
            '*Number 8 )
         (gui '(+JS +Button) "Increment"
            '(inc '*Number) ) ) ) )
########################################################################

Validation

To allow automatic validation of user input, the chk> message is sent to all components at appropriate times. The corresponding method should return NIL if the value is all right, or a string describing the error otherwise.

Many of the built-in classes have a chk> method. The +NumField class checks for legal numeric input, or the +DateField for a valid calendar date.

An on-the-fly check can be implemented with the +Chk prefix class. The following code only accepts numbers not bigger than 9: The or expression first delegates the check to the main +NumField class, and - if it does not give an error - returns an error string when the current value is greater than 9.


########################################################################
(app)

(action
   (html 0 "+Chk" "@lib.css" NIL
      (form NIL
         (gui '(+Chk +NumField)
            '(or
               (extra)
               (and (> (val> This) 9) "Number too big") )
            12 )
         (gui '(+JS +Button) "Print"
            '(msg (val> (field -1))) ) ) ) )
########################################################################

A more direct kind of validation is built-in via the +Limit class. It controls the maxlength attribute of the generated HTML input field component. Thus, it is impossible to type to more characters than allowed into the field.


########################################################################
(app)

(action
   (html 0 "+Limit" "@lib.css" NIL
      (form NIL
         (gui '(+Limit +TextField) 4 8)
         (gui '(+JS +Button) "Print"
            '(msg (val> (field -1))) ) ) ) )
########################################################################

Data Linkage

Although set> and val> are the official methods to get a value in and out of a GUI component, they are not very often used explicitly. Instead, components are directly linked to internal Lisp data structures, which are usually either variables or database objects.

The +Var prefix class takes a variable (described as the var data type - either a symbol or a cons pair - in the Function Reference). In the following example, we initialize a global variable with the value "abc", and let a +TextField operate on it. The "Print" button can be used to display its current value.


########################################################################
(app)

(setq *TextVariable "abc")

(action
   (html 0 "+Var" "@lib.css" NIL
      (form NIL
         (gui '(+Var +TextField) '*TextVariable 8)
         (gui '(+JS +Button) "Print"
            '(msg *TextVariable) ) ) ) )
########################################################################

+E/R takes an entity/relation specification. This is a cons pair, with a relation in its CAR (e.g. nm, for an object's name), and an expression in its CDR (typically (: home obj), the object stored in the obj property of the current form).

For an isolated, simple example, we create a temporary database, and access the nr and nm properties of an object stored in a global variable *Obj.


########################################################################
(when (app)                # On start of session
   (class +Tst +Entity)    # Define data model
   (rel nr (+Number))      # with a number
   (rel nm (+String))      # and a string
   (pool (tmp "db"))       # Create temporary DB
   (setq *Obj              # and a single object
      (new! '(+Tst) 'nr 1 'nm "New Object") ) )

(action
   (html 0 "+E/R" "@lib.css" NIL
      (form NIL
         (gui '(+E/R +NumField) '(nr . *Obj) 8)    # Linkage to 'nr'
         (gui '(+E/R +TextField) '(nm . *Obj) 20)  # Linkage to 'nm'
         (gui '(+JS +Button) "Show"                # Show the object
            '(out 2 (show *Obj)) ) ) ) )           # on standard error
########################################################################


Buttons

Buttons are, as explained in Control Flow, the only way (via POST requests) for an application to communicate with the server.

Basically, a +Button takes

Here is a minimal button, with just a label and an expression:


   (gui '(+Button) "Label" '(doSomething))

And this is a button displaying different labels, depending on the state:


   (gui '(+Button) "Enabled" "Disabled" '(doSomething))

To show an image instead of plain text, the label(s) must be preceeded by the T symbol:


   (gui '(+Button) T "img/enabled.png" "img/disabled.png" '(doSomething))

The expression will be executed during action handling (see Action Forms), when this button was pressed.

Like other components, buttons can be extended and combined with prefix classes, and a variety of predefined classes and class combinations are available.

Dialog Buttons

Buttons are essential for the handling of alerts and dialogs. Besides buttons for normal functions, like scrolling in charts or other side effects, special buttons exist which can close an alert or dialog in addition to doing their principal job.

Such buttons are usually subclasses of +Close, and most of them can be called easily with ready-made functions like closeButton, cancelButton, yesButton or noButton. We saw a few examples in Alerts and Dialogs.

Active JavaScript

When a button inherits from the +JS class (and JavaScript is enabled in the browser), that button will possibly show a much faster response in its action.

The reason is that the activation of a +JS button will - instead of doing a normal POST - first try to send only the contents of all GUI components via an XMLHttpRequest to the server, and receive the updated values in response. This avoids the flicker caused by reloading and rendering of the whole page, is much faster, and also does not jump to the beginning of the page if it is larger than the browser window. The effect is especially noticeable while scrolling in charts.

Only if this fails, for example because an error message was issued, or a dialog popped up, it will fall back, and the form will be POSTed in the normal way.

Thus it makes no sense to use the +JS prefix for buttons that cause a change of the HTML code, open a dialog, or jump to another page. In such cases, overall performance will even be worse, because the XMLHttpRequest is tried first (but in vain).

When JavaScript is disabled int the browser, the XMLHttpRequest will not be tried at all. The form will be fully usable, though, with identical functionality and behavior, just a bit slower and not so smooth.


A Minimal Complete Application

The PicoLisp release includes in the "app/" directory a minimal, yet complete reference application. This application is typical, in the sense that it implements many of the techniques described in this document, and it can be easily modified and extended. In fact, we use it as templates for our own production application development.

It is a kind of simplified ERP system, containing customers/suppliers, products (items), orders, and other data. The order input form performs live updates of customer and product selections, price, inventory and totals calculations, and generates on-the-fly PDF documents. Fine-grained access permissions are controlled via users, roles and permissions. It comes localized in six languages (English, Spanish, German, Norwegian, Russian and Japanese), with some initial data and two sample reports.


Getting Started

For a global installation (see Installation), please create a symbolic link to the place where the program files are installed. This is necessary because the application needs read/write access to the current working directory (for the database and other runtime data).


$ ln -s /usr/share/picolisp/app

As ever, you may start up the application in debugging mode


$ pil app/main.l -main -go +

or in (non-debug) production mode


$ pil app/main.l -main -go -wait

and go to 'http://localhost:8080' with your browser. You can login as user "admin", with password "admin". The demo data contain several other users, but those are more restricted in their role permissions.

Another possibility is to try the online version of this application at app.7fach.de.

Localization

Before or after you logged in, you can select another language, and click on the "Change" button. This will effect all GUI components (though not text from the database), and also the numeric, date and telephone number formats.

Navigation

The navigation menu on the left side shows two items "Home" and "logout", and three submenus "Data", "Report" and "System".

Both "Home" and "logout" bring you back to the initial login form. Use "logout" if you want to switch to another user (say, for another set of permissions), and - more important - before you close your browser, to release possible locks and process resources on the server.

The "Data" submenu gives access to application specific data entry and maintenance: Orders, product items, customers and suppliers. The "Report" submenu contains two simple inventory and sales reports. And the "System" submenu leads to role and user administration.

You can open and close each submenu individually. Keeping more than one submenu open at a time lets you switch rapidly between different parts of the application.

The currently active menu item is indicated by a highlighted list style (no matter whether you arrived at this page directly via the menu or by clicking on a link somewhere else).

Choosing Objects

Each item in the "Data" or "System" submenu opens a search dialog for that class of entities. You can specify a search pattern, press the top right "Search" button (or just ENTER), and scroll through the list of results.

While the "Role" and "User" entities present simple dialogs (searching just by name), other entities can be searched by a variety of criteria. In those cases, a "Reset" button clears the contents of the whole dialog. A new object can be created with bottom right "New" button.

In any case, the first column will contain either a "@"-link (to jump to that object) or a "@"-button (to insert a reference to that object into the current form).

By default, the search will list all database objects with an attribute value greater than or equal to the search criterion. The comparison is done arithmetically for numbers, and alphabetically (case sensitive!) for text. This means, if you type "Free" in the "City" field of the "Customer/Supplier" dialog, the value of "Freetown" will be matched. On the other hand, an entry of "free" or "town" will yield no hits.

Some search fields, however, show a different behavior depending on the application:

Using the bottom left scroll buttons, you can scroll through the result list without limit. Clicking on a link will bring up the corresponding object. Be careful here to select the right column: Some dialogs (those for "Item" and "Order") also provide links for related entities (e.g. "Supplier").

Editing

A database object is usually displayed in its own individual form, which is determined by its entity class.

The basic layout should be consistent for all classes: Below the heading (which is usually the same as the invoking menu item) is the object's identifier (name, number, etc.), and then a row with an "Edit" button on the left, and "Delete" button, a "Select" button and two navigation links on the right side.

The form is brought up initially in read-only mode. This is necessary to prevent more than one user from modifying an object at the same time (and contrary to the previous PicoLisp Java frameworks, where this was not a problem because all changes were immediately reflected in the GUIs of other users).

So if you want to modify an object, you have to gain exclusive access by clicking on the "Edit" button. The form will be enabled, and the "Edit" button changes to "Done". Should any other user already have reserved this object, you will see a message telling his name and process ID.

An exception to this are objects that were just created with "New". They will automatically be reserved for you, and the "Edit" button will show up as "Done".

The "Delete" button pops up an alert, asking for confirmation. If the object is indeed deleted, this button changes to "Restore" and allows to undelete the object. Note that objects are never completely deleted from the database as long as there are any references from other objects. When a "deleted" object is shown, its identifier appears in square brackets.

The "Select" button (re-)displays the search dialog for this class of entities. The search criteria are preserved between invocations of each dialog, so that you can conveniently browse objects in this context.

The navigation links, pointing left and right, serve a similar purpose. They let you step sequentially through all objects of this class, in the order of the identifier's index.

Other buttons, depending on the entity, are usually arranged at the bottom of the form. The bottom rightmost one should always be another "Edit" / "Done" button.

As we said in the chapter on Scrolling, any button in the form will save changes to the underlying data model. As a special case, however, the "Done" button releases the object and reverts to "Edit". Besides this, the edit mode will also cease as soon as another object is displayed, be it by clicking on an object link (the pencil icon), the top right navigation links, or a link in a search dialog.

Buttons vs. Links

The only way to interact with a HTTP-based application server is to click either on a HTML link, or on a submit button (see also Control Flow). It is essential to understand the different effects of such a click on data entered or modified in the current form.

For that reason the layout design should clearly differentiate between links and buttons. Image buttons are not a good idea when in other places images are used for links. The standard button components should be preferred; they are usually rendered by the browser in a non-ambiguous three-dimensional look and feel.

Note that if JavaScript is enabled in the browser, changes will be automatically committed to the server.

The enabled or disabled state of a button is an integral part of the application logic. It must be indicated to the user with appropriate styles.


The Data Model

The data model for this mini application consists of only six entity classes (see the E/R diagram at the beginning of "app/er.l"):

The classes +Role and +User are defined in "@lib/adm.l". A +Role has a name, a list of permissions, and a list of users assigned to this role. A +User has a name, a password and a role.

In "app/er.l", the +Role class is extended to define an url> method for it. Any object whose class has such a method is able to display itself in the GUI. In this case, the file "app/role.l" will be loaded - with the global variable *ID pointing to it - whenever an HTML link to this role object is activated.

The +User class is also extended. In addition to the login name, a full name, telephone number and email address is declared. And, of course, the ubiquitous url> method.

The application logic is centered around orders. An order has a number, a date, a customer (an instance of +CuSu) and a list of positions (+Pos objects). The sum> method calculates the total amount of this order.

Each position has an +Item object, a price and a quantity. The price in the position overrides the default price from the item.

Each item has a number, a description, a supplier (also an instance of +CuSu), an inventory count (the number of these items that were counted at the last inventory taking), and a price. The cnt> method calculates the current stock of this item as the difference of the inventory and the sold item counts.

The call to dbs at the end of "app/er.l" configures the physical database storage. Each of the supplied lists has a number in its CAR which determines the block size as (64 << N) of the corresponding database file. The CDR says that the instances of this class (if the element is a class symbol) or the tree nodes (if the element is a list of a class symbol and a property name) are to be placed into that file. This allows for some optimizations in the database layout.


Usage

When you are connected to the application (see Getting Started) you might try to do some "real" work with it. Via the "Data" menu (see Navigation) you can create or modify customers, suppliers, items and orders, and produce simple overviews via the "Report" menu.

Customer/Supplier

Source in "app/cusu.l"

The Customer/Supplier search dialog (choCuSu in "app/gui.l") supports a lot of search criteria. These become necessary when the database contains a large number of customers, and can filter by zip, by phone number prefixes, and so on.

In addition to the basic layout (see Editing), the form is divided into four separate tabs. Splitting a form into several tabs helps to reduce traffic, with possibly better GUI response. In this case, four tabs are perhaps overkill, but ok for demonstration purposes, and they leave room for extensions.

Be aware that when data were modified in one of the tabs, the "Done" button has to be pressed before another tab is clicked, because tabs are implemented as HTML links (see Buttons vs. Links).

New customers or suppliers will automatically be assigned the next free number. You can enter another number, but an error will result if you try to use an existing number. The "Name" field is mandatory, you need to overwrite the "<Name>" clue.

Phone and fax numbers in the "Contact" tab must be entered in the correct format, depending on the locale (see Telephone Numbers).

The "Memo" tab contains a single text area. It is no problem to use it for large pieces of text, as it gets stored in a database blob internally.

Item

Source in "app/item.l"

Items also have a unique number, and a mandatory "Description" field.

To assign a supplier, click on the "+" button. The Customer/Supplier search dialog will appear, and you can pick the desired supplier with the "@" button in the first column. Alternatively, if you are sure to know the exact spelling of the supplier's name, you can also enter it directly into the text field.

In the search dialog you may also click on a link, for example to inspect a possible supplier, and then return to the search dialog with the browser's back button. The "Edit" mode will then be lost, however, as another object has been visited (this is described in the last part of Editing).

You can enter an inventory count, the number of items currently in stock. The following field will automatically reflect the remaining pieces after some of these items were sold (i.e. referenced in order positions). It cannot be changed manually.

The price should be entered with the decimal separator according to the current locale. It will be formatted with two places after the decimal separator.

The "Memo" is for an arbitrary info text, like in Customer/Supplier above, stored in a database blob.

Finally, a JPEG picture can be stored in a blob for this item. Choose a file with the browser's file select control, and click on the "Install" button. The picture will appear at the bottom of the page, and the "Install" button changes to "Uninstall", allowing the picture's removal.

Order

Source in "app/ord.l"

Oders are identified by number and date.

The number must be unique. It is assigned when the order is created, and cannot be changed for compliance reasons.

The date is initialized to "today" for a newly created order, but may be changed manually. The date format depends on the locale. It is YYYY-MM-DD (ISO) by default, DD.MM.YYYY in the German and YYYY/MM/DD in the Japanese locale. As described in Time & Date, this field allows input shortcuts, e.g. just enter the day to get the full date in the current month.

To assign a customer to this order, click on the "+" button. The Customer/Supplier search dialog will appear, and you can pick the desired customer with the "@" button in the first column (or enter the name directly into the text field), just as described above for Items.

Now enter order the positions: Choose an item with the "+" button. The "Price" field will be preset with the item's default price, you may change it manually. Then enter a quantity, and click a button (typically the "+" button to select the next item, or a scroll button go down in the chart). The form will be automatically recalculated to show the total prices for this position and the whole order.

Instead of the "+" or scroll buttons, as recommended above, you could of course also press the "Done" button to commit changes. This is all right, but has the disadvantage that the button must be pressed a second time (now "Edit") if you want to continue with the entry of more positions.

The "x" button at the right of each position deletes that position without further confirmation. It has to be used with care!

The "^" button is a "bubble" button. It exchanges a row with the row above it. Therefore, it can be used to rearrange all items in a chart, by "bubbling" them to their desired positions.

The "PDF-Print" button generates and displays a PDF document for this order. The browser should be configured to display downloaded PDF documents in an appropriate viewer. The source for the postscript generating method is in "app/lib.l". It produces one or several A4 sized pages, depending on the number of positions.

Reports

Sources in "app/inventory.l and "app/sales.l"

The two reports ("Inventory" and "Sales") come up with a few search fields and a "Show" button.

If no search criteria are entered, the "Show" button will produce a listing of the relevant part of the whole database. This may take a long time and cause a heavy load on the browser if the database is large.

So in the normal case, you will limit the domain by stating a range of item numbers, a description pattern, and/or a supplier for the inventory report, or a range of order dates and/or a customer for the sales report. If a value in a range specification is omitted, the range is considered open in that direction.

At the end of each report appears a "CSV" link. It downloads a file with the TAB-separated values generated by this report.