TiddlyWiki5/tiddlywiki5/tiddlers/TiddlyWikiArchitecture.tid

title: TiddlyWikiArchitecture
modifier: JeremyRuston

!! Overview

The heart of TiddlyWiki can be seen as an extensible representation transformation engine. Given the text of a tiddler and its associated MIME type, the engine can produce a rendering of the tiddler in a new MIME type. Furthermore, it can efficiently selectively update the rendering to track any changes in the tiddler or its dependents.

The most important transformations are from `text/x-tiddlywiki` wikitext into `text/html` or `text/plain` but the engine is used throughout the system for other transformations, such as converting images for display in HTML, sanitising fragments of JavaScript, and processing CSS.

The key feature of wikitext is the ability to include one tiddler within another (usually referred to as //transclusion//). For example, one could have a tiddler called //Disclaimer// that contains the boilerplate of a legal disclaimer, and then include it within lots of different tiddlers with the macro call `<<tiddler Disclaimer>>`. This simple feature brings great power in terms of encapsulating and reusing content, and evolving a clean, usable implementation architecture to support it efficiently is a key objective of the TiddlyWiki5 design.

It turns out that the transclusion capability combined with the selective refreshing mechanism provides a good foundation for building TiddlyWiki's user interface itself. Consider, for example, the StoryMacro in its simplest form:
{{{
<<story story:MyStoryTiddler>>
}}}
The story macro looks for a list of tiddler titles in the tiddler `MyStoryTiddler`, and displays them in sequence. The subtle part is that subsequently, if `MyStoryTiddler` changes, the `<<story>>` macro is selectively re-rendered. So, to navigate to a new tiddler, code merely needs to add the name of the tiddler and a line break to the top of `MyStoryTiddler`:
{{{
var storyTiddler = store.getTiddler("MyStoryTiddler");
store.addTiddler(new Tiddler(storyTiddler,{text: navigateTo + "\n" + storyTiddler.text}));
}}}
The mechanisms that allow all of this to work are fairly intricate. The sections below progressively build the key architectural concepts of TiddlyWiki5 in a way that should provide a good basis for exploring the code directly.
!! Tiddlers
Tiddlers are an immutable dictionary of name:value pairs called fields.

The only field that is required is the {{{title}}} field, but useful tiddlers also have a {{{text}}} field, and some or all of the standard fields {{{modified}}}, {{{modifier}}}, {{{created}}}, {{{creator}}}, {{{tags}}} and {{{type}}}.

Hardcoded in the system is the knowledge that the `tags` field is a string array, and that the `modified` and `created` fields are JavaScript `Date` objects. All other fields are strings.

The {{{type}}} field identifies the representation of the tiddler text with a MIME type.
!! ~WikiStore
Groups of uniquely titled tiddlers are contained in WikiStore objects.

The WikiStore also manages the plugin modules used for macros, and operations like serializing, deserializing, parsing and rendering tiddlers.

Each WikiStore is connected to another shadow store that is used to provide default content. Under usual circumstances, when an attempt is made to retrieve a tiddler that doesn't exist in the store, the search continues into its shadow store (and so on, if the shadow store itself has a shadow store).
!! ~WikiStore Events
Clients can register event handlers with the WikiStore object. Event handlers can be registered to be triggered for modifications to any tiddler in the store, or with a filter to only be invoked when a particular tiddler or set of tiddlers changes.

Whenever a change is made to a tiddler, the wikistore registers a `nexttick` handler (if it hasn't already done so). The `nexttick` handler looks back at all the tiddler changes, and dispatches any matching event handlers. 
!! Parsing and Compiling
TiddlyWiki parses the content of tiddlers to build an internal tree representation that is used for several purposes:
* Rendering a tiddler to other formats (e.g. converting wikitext to HTML)
* Detecting outgoing links from a tiddler, and from them...
* ...computing incoming links to a tiddler
* Detecting tiddlers that are orphans with no incoming links
* Detecting tiddlers that are referred to but missing
The parse tree is built when needed, and then cached by the WikiStore until the tiddler changes.

TiddlyWiki5 uses multiple parsers:
* Wikitext ({{{text/x-tiddlywiki}}}) in `js/WikiTextParser.js`
* JavaScript ({{{text/javascript}}}) in `js/JavaScriptParser.js`
* Images ({{{image/png}}} and {{{image/jpg}}}) in `js/ImageParser.js`
* JSON ({{{application/json}}}) in `js/JSONParser.js`
Additional parsers are planned:
* CSS ({{{text/css}}})
* Recipe ({{{text/x-tiddlywiki-recipe}}})
One global instance of each parser is instantiated in `js/App.js` and registered with the main WikiStore object. In some cases the constructors require special parameters or options (eg, the JavaScript parser requires the [[PEG.JS]] JavaScript parser text).

The parsers are all used the same way:
{{{
var parseTree = parser.parse(type,text) // Parses the text and returns a parse tree object
}}}
The parse tree object exposes the following fields:
{{{
var renderer = parseTree.compile(type); // Compiles the parse tree into a renderer for the specified MIME type
console.log(parseTree.toString(type)); // Returns a readable string representation of the parse tree (either `text/html` or `text/plain`)
var dependencies = parseTree.dependencies; // Gets the dependencies of the parse tree (see below)
}}}
The dependencies are returned as an object like this:
{{{
{
	link: {"tiddlertitle1": 2, "tiddlertitle2": 3},
	include: {"tiddlertitle3": 5},
	dependentAll: false
}
}}}
The `link` and `include` fields are hashmaps of the title of each tiddler that is linked or included in the current one. For the tiddler to be subsequently rendered correctly, the linked tiddlers must be present, at least in skinny form, and the included tiddlers must be fully loaded.

The `dependentAll` field is used to indicate that the tiddler contains a macro that scans the entire pool of tiddlers (for example the `<<list>>` macro), and is potentially dependent on any of them. The effect is that the tiddler should be rerendered whenever any other tiddler changes.
!! Rendering
The `parseTree.compile(type)` method returns a renderer object that contains a JavaScript function that generates the new representation of the original parsed text.

The renderer is invoked as follows:
{{{
var renderer = parseTree.compile("text/html");
var html = renderer.render(tiddler,store);
}}}
The `tiddler` parameter to the `render` method identifies the tiddler that is acting as the context for this rendering -- for example, it provides the fields displayed by the `<<view>>` macro. The `store` parameter is used to resolve any references to other tiddlers.
!! Rerendering
When rendering to the HTML/SVG DOM in the browser, TiddlyWiki5 also allows a previous rendering to be selectively updated in response to changes in dependent tiddlers. At the moment, only the WikiTextRenderer supports rerendering.

The rerender method on the renderer is called as follows:
{{{
var node = document.getElementById("myNode");
var renderer = parseTree.compile("text/html");
myNode.innerHTML = renderer.render(tiddler,store);
// And then, later:
renderer.rerender(node,changes,tiddler,store,renderStep);
}}}
The parameters to `rerender()` are:
|!Name |!Description |
|node |A reference to the DOM node containing the rendering to be rerendered |
|changes |A hashmap of `{title: "created|modified|deleted"}` indicating which tiddlers have changed since the original rendering |
|tiddler |The tiddler providing the rendering context |
|store |The store to use for resolving references to other tiddlers |
|renderStep |See below |
Currently, the only macro that supports rerendering is the `<<story>>` macro; all other macros are rerendered by calling the ordinary `render()` method again. The reason that the `<<story>>` macro goes to the trouble of having a `rerender()` method is so that it can be carefully selective about not disturbing tiddlers in the DOM that aren't affected by the change. If there were, for instance, a video playing in one of the open tiddlers it would be reset to the beginning if the tiddler were rerendered.

----
//The remaining text is from an earlier draft, and is in the process of being updated//

When the text of a tiddler is requested in a different format than its native type, TiddlyWiki5 compiles a JavaScript function that generates the new format from the text of the tiddler.

So, a simple tiddler in {{{application/x-tiddlywiki}}} format might read:
{{{
Hello World
}}}
The function to render it to {{{text/html}}} might look like this:
{{{
function() {
	return "<p>Hello World</p>";
}
}}}

The function can also include calls to the store to incorporate the values of other tiddlers. Consider this tiddler, called {{{HelloThere}}}:
{{{
Hello <<tiddler Who>>
}}}
And this one called {{{Who}}}:
{{{
World
}}}
The function to generate {{{HelloThere}}} in {{{text/html}}} might be:
{{{
function() {
	return ["<p>","Hello ", getTiddlerText("Who","text/html"), "</p>"].join(""); 
}
}}}

Now, the return value of this function can be cached until a tiddler in the dependency chain changes. The function itself can be cached until the tiddler itself changes, or a macro that it uses changes.

The dependency chain is calculated when a tiddler is parsed. Every tiddler that is directly referenced is accumulated (until the point at which it is concluded that it is simpler to mark the tiddler as being dependent on any other tiddler changing).

Evaluated macro parameters are parsed and can be checked for safeness, and then included in the compiled code. For example,
{{{
Hello <<echo {{2+2}}>>
}}}
Compiles to:
{{{
function() {
	return ["Hello ",(function(){
		return 2+2;
	})().toString()].join("");
}
}}}

The compilation process has several steps:
* First, the parse tree is used to generate a JavaScript tree
* The JavaScript tree is scanned to determine the tiddlers on which this one depends
* Finally, executable JavaScript text is generated by walking the JavaScript tree