The Tcl Web Application Framework provides a basic framework for application developers to write web-based applications without having to deal with some of the common overhead associated with building web-based applications, such as:
It is intended to be extensible and replacable so that if the developer does wish to do these things they can do so while retaining the remaining features.
How Does it Work ?
Since Rivet (and in some cases, RivetCGI) keeps an interpreter around for longer than a single request, we utilize the persistent global namespace to avoid having to do the same work over for every request if it's already been done globally.
Thus there is a distinction between the "global" or "interpreter" state and and the "request" state time.
Because we cannot know if this is a new interpreter ahead of request-time, we must perform all per-interpreter actions if they are needed for every request followed by the per-request actions.
Developers may write scripts to be called as per-interpreter or per-request scripts based on the need. The location of the script determines whether it is called once-per-interpreter (in which case it should update the global state) or once-per-request (in which case it should only modify the request state, such as the ::request namespace or output to the HTTP stream).
local/modules/preload/*.tcl(before the page is displayed)
modules/load/*.tcl(before the page is displayed)
local/modules/load/*.tcl(before the page is displayed)
modules/unload/*.tcl(after the page is displayed)
local/modules/unload/*.tcl(after the page is displayed)
Request Processing Is Distinct From Output Processing
In the Tcl Web Application Framework, generated output is handled by displaying
Rivet fragments. These fragments are returned by various modules and the order
determines what the output will be. The "html" module, for example, supplies
an HTML header and footer (using
which are Rivet fragments displayed before and after the requested module's
A small note should be made here that all Tcl procedures are called (except for
the "unload" scripts are
[source]d) before any output is produced.
That is, request processing (taking the user request and deciding what to do with it by dispatching Rivet fragments is performed entirely before displaying any Rivet fragments.
The primary benefit to this is that any Tcl procedure called will be evaluated prior to any Rivet fragments being parsed, so when Rivet fragments are parsed they are all parsed with the same state (except for any state the Rivet fragments themselves modify, which is generally bad form).
This is used, for example, to modify CSS from a module's action which is then parsed and displayed as part of the HTML header's Rivet fragment.
Writing Your Application
All resources for your application should live in the "local" directory relative to "index.rvt". This allows for the framework to be replaced around your application without affecting your application.
As above, the "local" tree is searched along with the framework paths. In addition many of the commands from other packages will prefer a "local/" path if a replacement file can be found there. Example commands are:
The basic unit of an application is called a "module". Modules have a simple structure and are registered using the [module::register] command.
The first thing a module needs is a name. This name is also the name of the namespace the module will provide some well-known named functions from.
Generally the module directory name and the module name should be the same. In that case, if we were creating a module called "example" we would create:
local/modules/example/load/example.tcl(file, per-interpreter script)
local/modules/example/main.rvt(file, Rivet script)
The file (all paths will be relative to the module directory of "local/modules/example") "load/example.tcl" should do all of the per-interpreter initialization to make this module useful and ready to use for a request. That includes:
- Creating a namespace named after the module
- Creating a procedure called "init" in that namespace (
[::example::init]) which must return some form of "true"
- Registering the module with
- Optionally a module may be requested to be called as an "initialization
module", which means it will accept a "start", "stop", "pre-request" and
"post-request" action by calling