Wapp - A Web-Application Framework for TCL
==========================================
1.0 Introduction
----------------
Wapp is a lightweight framework that simplifies the
construction of web application written in TCL. The same Wapp-based
application can be launched in multiple ways:
1. From the command-line, with automatic web-browser startup
2. As a stand-alone web server
3. As a CGI script
4. As an SCGI program
All four methods use the same application code and present the same
interface to the application user. An application can be developed on
the desktop using stand-alone mode (1), then deployed as a stand-alone
server (2), or a CGI script (3), or as an SCGI program (4).
2.0 Hello World!
----------------
Wapp is designed to be easy to use. A hello-world program is as follows:
>
package require wapp ;# OR just "source wapp.tcl"
proc wapp-default {req} {
wapp-subst {<h1>Hello, World!</h1>\n}
}
wapp-start $::argv
The application defines one or more procedures that accept HTTP requests
and generate appropriate replies.
For an HTTP request where the initial portion of the URI is "abcde", the
procedure named "wapp-page-abcde" will be invoked to construct the reply.
If no such procedure exists, "wapp-default" is invoked instead. The latter
technique is used for the hello-world example above.
The procedure generates a reply using one or more calls to the "wapp-subst"
command. Each "wapp-subst" command appends new text to the reply, applying
various substitutions as it goes. The only substitution in this example is
the \\n at the end of the line.
The "wapp-start" command starts up the application.
To run this application, copy the code above into a file named "main.tcl"
and then enter the following command:
>
tclsh main.tcl
That command will start up a web-server bound to the loopback
IP address, then launch a web-browser pointing at that web-server.
The result is that the "Hello, World!" page will automatically
appear in your web browser.
To run this same program as a traditional web-server on TCP port 8080, enter:
>
tclsh main.tcl --server 8080
Here the built-in web-server listens on all IP addresses and so the
web page is available on other machines. But the web-broswer is not
automatically started in this case, so you will have to manually enter
"http://localhost:8080/" into your web-browser in order to see the page.
To run this program as CGI, put the main.tcl script in your web-servers
file hierarchy, in the appropriate place for CGI scripts, and make any
other web-server specific configuration changes so that the web-server
understands that the main.tcl file is a CGI script. Then point your
web-browser at that script.
Run the hello-world program as SCGI like this:
>
tclsh main.tcl --scgi 9000
Then configure your web-server to send SCGI requests to TCL port 9000
for some specific URI, and point your web-browser at that URI.
3.0 A Slightly Longer Example
-----------------------------
Information about each HTTP request is encoded in the global ::wapp
dict variable. The following sample program shows the information
available in ::wapp.
>
package require wapp
proc wapp-default {} {
set B [wapp-param BASE_URL]
wapp-trim {
<h1>Hello, World!</h1>
<p>See the <a href='%html($B)/env'>Wapp
Environment</a></p>
}
}
proc wapp-page-env {} {
global wapp
wapp-subst {<h1>Wapp Environment</h1>\n<pre>\n}
foreach var [lsort [dict keys $wapp]] {
if {[string index $var 0]=="."} continue
wapp-subst {%html($var) = %html([list [dict get $wapp $var]])\n}
}
wapp-subst {</pre>\n}
}
wapp-start $::argv
In this application, the default "Hello, World!" page has been extended
with a hyperlink to the /env page. The "wapp-subst" command has been
replaced by "wapp-trim", which works the same way with the addition that
it removes surplus whitespace from the left margin, so that the generated
HTML text does not come out indented. The "wapp-index" command uses
a "%html(...)" substitution. The "..." argument is expanded using the
usual TCL rules, but then the result is escaped so that it is safe to include
in an HTML document. Other supported substitutions are "%url(...)" for
URLs on the href= and src= attributes of HTML entities, "%qp(...)" for
query parameters, "%string(...)" for string literals within javascript,
and "%unsafe(...)" for direct literal substitution. As its name implies,
the %unsafe() substitution should be avoid whenever possible.
The /env page is implemented by the "wapp-page-env" proc. This proc
generates HTML that describes the content of the ::wapp dict.
Keys that begin with "." are for internal use by Wapp and are skipped
for this display. Notice the use of "wapp-subst" to safely escape text
for inclusion in an HTML document.
4.0 The ::wapp Global Dict
--------------------------
To better understand how the ::wapp dict works, try running the previous
sample program, but extend the /env URL with extra path elements and query
parameters. For example:
<http://localhost:8080/env/longer/path?q1=5&title=hello+world%21>
Notice how the query parameters in the input URL are decoded and become
elements of the ::wapp dict. The same thing occurs with POST parameters
and cookies - they are all converted into entries in the ::wapp dict
variable so that the parameters are easily accessible to page generation
procedures.
The ::wapp dict contains additional information about the request,
roughly corresponding to CGI environment variables. To prevent environment
information from overlapping and overwriting query parameters, all the
environment information uses upper-case names and all query parameters
are required to be lower case. If an input URL contains an upper-case
query parameter (or POST parameter or cookie), that parameter is silently
omitted from the ::wapp dict.
The ::wapp dict contains the following environment values:
+ **HTTP\_HOST**
The hostname (or IP address) and port that the client used to create
the current HTTP request. This is the first part of the request URL,
right after the "http://" or "https://". The format for this value
is "HOST:PORT". Examples: "sqlite.org:80" or "127.0.0.1:32172".
+ **HTTP\_USER\_AGENT**
The name of the web-browser or other client program that generated
the current HTTP request.
+ **HTTP\_COOKIE**
The values of all cookies in the HTTP header
+ **HTTPS**
If the HTTP request arrived of SSL (via "https://"), then this variable
has the value "on". For an unencrypted request ("http://"), this
variable does not exist.
+ **REMOTE\_ADDR**
The IP address from which the HTTP request originated.
+ **REMOTE\_PORT**
The TCP port from which teh HTTP request originated.
+ **SCRIPT_NAME**
In CGI mode, this is the name of the CGI script in the URL. In other
words, this is the initial part of the URL path that identifies the
CGI script. For other modes, this variable is an empty string.
+ **PATH\_INFO**
The part of the URL path that follows the SCRIPT\_NAME. For all modes
other than CGI, this is exactly the URL pathname, though with the
query parameters removed. PATH_INFO begins with a "/".
+ **REQUEST\_URI**
The URL for the inbound request, without the initial "http://" or
"https://" and without the HTTP\_HOST. This variable is the same as
the concatenation of $SCRIPT\_NAME and $PATH\_INFO.
+ **REQUEST\_METHOD**
"GET" or "HEAD" or "POST"
* **CONTENT\_LENGTH**
The number of bytes of POST data.
* **CONTENT\_TYPE**
The mimetype of the POST data. Usually this is
application/x-www-form-urlencoded.
All of the above are standard CGI environment values.
The following are additional values add by Wapp:
* **CONTENT**
The raw POST data text.
+ **BASE\_URL**
The text of the request URL through the SCRIPT\_NAME. This value can
be prepended to hyperlinks to ensure that the correct page is reached by
those hyperlinks.
+ **PATH\_HEAD**
The first element in the PATH\_INFO. The value of PATH\_HEAD is used to
select one of the "wapp-page-XXXXX" commands to run in order to generate
the output web page.
+ **PATH\_TAIL**
All of PATH\_INFO that follows PATH\_HEAD.
+ **SELF\_URL**
The URL for the current page, stripped of query parameter. This is
useful for filling in the action= attribute of forms.
### 4.1 URL Parsing Example
For the input URL "http://example.com/cgi-bin/script/method/extra/path?q1=5"
and for a CGI script named "script" in the /cgi-bin/ directory,
the following CGI environment values are generated:
+ **HTTP\_HOST** → "example.com:80"
+ **SCRIPT\_NAME** → "/cgi-bin/script"
+ **PATH\_INFO** → "/method/extra/path"
+ **REQUEST\_URI** → "/cgi-bin/script/method/extra/path"
+ **QUERY\_STRING** → "q1=5"
+ **BASE\_URL** → "http://example.com/cgi-bin/script"
+ **SELF\_URL** → "http://example.com/cgi-bin/script/method"
+ **PATH\_HEAD** → "method"
+ **PATH\_TAIL** → "extra/path"
The first five elements of the example above, HTTP\_HOST through
QUERY\_STRING, are standard CGI. The final four elements are Wapp
extensions.
5.0 Wapp Commands
-----------------
The following utility commands are available for use by applications built
on Wapp:
+ **wapp-start** _ARGLIST_
Start up the application. _ARGLIST_ is typically the value of $::argv,
though it might be some subset of $::argv if the containing application
has already processed some command-line parameters for itself.
+ **wapp** _TEXT_
Add _TEXT_ to the web page output currently under construction. _TEXT_
must not contain any TCL variable or command substitutions.
+ **wapp-subst** _TEXT_
The _TEXT_ argument should be enclosed in {...} to prevent substitutions.
The "wapp-subst" command itself will do all necessary backslash
substitutions. Command and variable substitutions only occur within
"%html(...)", "%url(...)", "%qp(...)", "%string(...)", and
"%unsafe(...)". The substitutions are escaped (except in the case of
"%unsafe(...)") so that the result is safe for inclusion within the
body of an HTML document, a URL, a query parameter, or a javascript
string literal, respectively.
+ **wapp-trim** _TEXT_
Just like wapp-subst, this routine appends _TEXT_ to the web page
under construction, using the %html, %url, %qp, %string, and %unsafe
substitutions. The difference is that this routine also removes
surplus whitespace from the left margin, so that if the _TEXT_
argument is intended in the source script, it will appear at the
left margin in the generated output.
+ **wapp-param** _NAME_ _DEFAULT_
Return the value of the query parameter or environment variable _NAME_,
or return _DEFAULT_ if there is no such query parameter or environment
variable. If _DEFAULT_ is omitted, then it is an empty string.
The same information can be obtained directly from the
$::wapp dict. This command is merely a convenient shortcut.
+ **wapp-mimetype** _MIMETYPE_
Set the MIME-type for the generated web page. The default is "text/html".
+ **wapp-reply-code** _CODE_
Set the reply-code for the HTTP request. The default is "200 Ok".
+ **wapp-redirect** _TARGET-URL_
Cause an HTTP redirect to the specified URL.
+ **wapp-reset**
Reset the web page under construction back to an empty string.
+ **wapp-set-cookie** _NAME_ _VALUE_
Cause the cookie _NAME_ to be set to _VALUE_.
+ **wapp-clear-cookie** _NAME_
Erase the cookie _NAME_.
+ **wapp-safety-check**
Examine all TCL procedures in the application and report errors about
unsafe usage of "wapp".
+ **wapp-unsafe** _TEXT_
Add _TEXT_ to the web page under construction even though _TEXT_ does
contain TCL variable and command substitutions. The application developer
must ensure that the variable and command substitutions does not allow
XSS attacks. Avoid using this command. The use of "wapp-subst" is
preferred in most situations.
+ **wapp-escape-html** _TEXT_
Add _TEXT_ to the web page under construction after first escaping any
HTML markup contained with _TEXT_. This command is equivalent to
"wapp-subst {%html(_TEXT_)}".
+ **wapp-escape-url** _TEXT_
Add _TEXT_ to the web page under construction after first escaping any
characters so that the result is safe to include as the value of a
query parameter on a URL. This command is equivalent to
"wapp-subst {%url(_TEXT_)}".
The following additional interfaces are envisioned, but are not yet
implemented:
+ **wapp-send-hex** _HEX_
Cause the HTTP reply to be binary that is constructed from the
hexadecimal text in the _HEX_ argument. Whitespace in _HEX_ is ignored.
This command is useful for returning small images from a pure script
input. The "wapp-file-to-hex" command can be used at development time
to generate appropriate _HEX_ for an image file.
+ **wapp-cache-control** _CONTROL_
The _CONTROL_ argument should be one of "no-cache", "max-age=N", or
"private,max-age=N", where N is an integer number of seconds.
+ **wapp-etag** _ETAG_
Set the expiration tag for the web page.
+ **wapp-send-file** _FILENAME_
Make the content of the file _FILENAME_ be the HTTP reply.
+ **wapp-send-query** _DB_ _SQL_
Run the SQLite query _SQL_ on the _DB_ database connection and make the
HTTP reply be the value of the first column of the first row in the result.
+ **wapp-set-csp** _POLICY_
Set the Content Security Policy for the application. This command only
works for command-line and server modes. This command is a no-op for CGI
and SCGI since there is no standard way of communicating the desired
content security policy back to the server in those instances.
+ **wapp-debug-port** _PORT_
For debugging use only: open a listening TCP socket on _PORT_ and run
an interactive TCL shell on connections to that port. This allows for
interactive debugging of a running instance of the Wapp server. This
command is a no-op for short-lived CGI programs, obviously. Also, this
command should only be used during debugging, as otherwise it introduces
a severe security vulnerability into the application.
6.0 Limitations
---------------
Each Wapp process is single-threaded.
The fileevent command is used to allow accepting multiple simultaneous
HTTP requests. However, as soon as a complete request is received, and
the "wapp-page-NAME" proc runs, all other processing is suspended until
that proc completes.
Hence, a long-running "wapp-page-NAME" proc can cause the system to
become unresponsive.
If this is a problem, the Wapp application can be run in CGI mode.
In CGI mode, the overlying webserver will start a separate Wapp process
for each request, and there are no limits on the number of simultaneous
Wapp processes.
The POST data parser for Wapp currently only understands
application/x-www-form-urlencoded content. Wapp does not (currently)
have a decoder for multipart/form-data content. A multipart/form-data
decoder might be added in the future. Or, individual applications can
implement their own multipart/form-data decoder using the raw POST data
held in the CONTENT entry of the ::wapp dict.
7.0 Design Rules
----------------
All global procs and variables used by Wapp begin with the four character
prefix "wapp". Procs and variable intended for internal use begin with
the seven character prefix "wappInt".