.TH "tkhtml" "n" "Sat Feb 25 06:44:47 PM ICT 2006"


.SH "NAME"
tkhtml - Widget to render html documents.

.SH "SYNOPSIS"
html pathName ?options?

.SH "STANDARD OPTIONS"
.RS

.nf
-height
-width
-xscrollcommand   
-xscrollincrement
-yscrollcommand   
-yscrollincrement
.fi
.RE

.SH "WIDGET-SPECIFIC OPTIONS"
Command-Line Name:-defaultstyle
.br
Database Name:  defaultstyle
.br
Database Class: Defaultstyle
.br
.RS
.PP
This option is used to set the default style-sheet for the
widget. The option value should be the entire text of the
default style-sheet, or an empty string, in which case a
built-in default stylesheet (for HTML) is used. 

At the moment, this option is not very useful. The default
stylesheet defines things that are "built-in" to the 
document - for example the behaviour of <p> or <img> tags
in html. The idea behind making it flexible is to allow
Tkhtml to display anything that looks roughly like an XML
document. But it will not work at the moment because of
other assumptions the implementation makes about the set
of valid tags.
.RE
Command-Line Name:-imagecmd
.br
Database Name:  imagecmd
.br
Database Class: Imagecmd
.br
.RS
.PP
As well as for replacing entire document nodes (i.e. <img>),
images are used in several other contexts in CSS formatted
documents, for example as list markers or backgrounds. If the
-imagecmd option is not set to an empty string (the default),
then each time an image URI is encountered in the document, it
is appended to the -imagecmd script and the resulting list
evaluated.

The command should return either an empty string, the name of a
Tk image, or a list of exactly two elements, the name of a Tk
image and a script. If the result is an empty string, then no
image can be displayed. If the result is a Tk image name, then
the image is displayed in the widget. When the image is no
longer required, it is deleted. If the result of the command is
a list containing a Tk image name and a script, then instead of
deleting the image when it is no longer required, the script is
evaluated.

If the size or content of the image are modified while it is in
use the widget display is updated automatically.
.RE
Command-Line Name:-logcmd
.br
Database Name:  logcmd
.br
Database Class: Logcmd
.br
.RS
.PP
This option is used for internally debugging the widget.
.RE

.SH "DESCRIPTION"

The [html] command creates a new window (given by the pathName
argument) and makes it into an html widget. The html command
returns its pathName argument. At the time this command is invoked,
there must not exist a window named pathName, but pathName's parent
must exist.

.SH "WIDGET COMMAND"
The [html] command creates a new Tcl command whose name is
pathName. This command may be used to invoke various operations on
the widget as follows:

pathName bbox nodeHandle
.br
.RS
If node nodeHandle generates content, this command returns a
list of four integers that define the bounding-box of the
generated content, relative to the top-left hand corner of the
rendered document. The first two integers are the x and y
coordinates of the top-left corner of the bounding-box, the
later two are the x and y coordinates of the bottom-right
corner of the same box. If the node does not generate content,
then an empty string is returned.
.RE

pathName cget option
.br
.RS
Returns the current value of the configuration option given
by option. Option may have any of the values accepted by
the [html] command.
.RE

pathName configure ?option? ?value?
.br
.RS
Query or modify the configuration options of the widget. If
no option is specified, returns a list describing all of
the available options for pathName (see Tk_ConfigureInfo
for information on the format of this list). If option is
specified with no value, then the command returns a list
describing the one named option (this list will be
identical to the corresponding sublist of the value
returned if no option is specified). If one or more
option-value pairs are specified, then the command modifies
the given widget option(s) to have the given value(s); in
this case the command returns an empty string. Option may
have any of the values accepted by the [html] command.
.RE

pathName handler type tag script
.br
.RS
This command is used to define "handler" scripts - Tcl
callback scripts that are invoked by the widget when
document elements of specified types are encountered. The
widget supports two types of handler scripts: "node" and
"script". The type parameter to this command must
take one of these two values.

For a "node" handler script, whenever a document element
having the specified tag type (e.g. "p" or "link") is
encountered during parsing, then the node handle for the
node is appended to script and the resulting list
evaluated as a Tcl command. See the section "NODE COMMAND"
for details of how a node handle may be used to query and
manipulate a document node.

If the handler script is a "script" handler, whenever a
document node of type tag is parsed, then the text
that appears between the start and end tags of the node is
appended to script and the resulting list evaluated
as a Tcl command.

Handler callbacks are always made from within 
[pathName parse] commands. The callback for a given node
is made as soon as the node is completely parsed.  This can
happen because an implicit or explicit closing tag is
parsed, or because there is no more document data and the
-final switch was passed to the [pathName parse]
command.

TODO: Return values of handler scripts? If an exception
occurs in a handler script?
.RE

pathName image
.br
.RS
This command returns the name of a new Tk image containing 
the rendered document. Where Tk widgets would be mapped in a 
live display, the image contains blank space.

The returned image should be deleted when the script has 
finished with it, for example:
.RS

.nf
set img [.html image]
# ... Use $img ...
image delete $img
.fi
.RE

This command is included mainly for automated testing and 
should be used with care, as large documents can result in very
large images that take a long time to create and use vast
amounts of memory.

Currently this command is not available on windows. On that
platform an empty string is always returned.
.RE

pathName node ? ?-index? x y?
.br
.RS
This command is used to retrieve one or more document node
handles from the current document. If the x and y parameters
are omitted, then the handle returned is the root-node of the
document, or an empty string if the document has no root-node
(i.e. an empty document).

If the x and y arguments are present, then a list of node 
handles is returned. The list contains one handle for each
node that generates content currently located at viewport
coordinates (x, y). Usually this is only a single node, but
floating boxes and other overlapped content can cause
this command to return more than one node.  If no content is
located at the specified coordinates or the widget window is
not mapped, then an empty string is returned. 

If the -index option is specified along with the x and y 
coordinates, then instead of a list of node handles, a list of
two elements is returned. The first element of the list is the
node-handle associated with the generated text closest to 
the specified (x, y) coordinates. The second list value is an
index into the text obtainable by [nodeHandle text] for the
character closest to coordinates (x, y). The index may be used
with the [pathName select] commands.

The document node can be queried and manipulated using the
interface described in the "NODE COMMAND" section.
.RE

pathName parse ?-final? html-text
.br
.RS
Append extra text to the end of the (possibly empty)
document currently stored by the widget. 

If the -final option is present, this indicates that the
supplied text is the last of the document. Any subsequent
call to [pathName parse] before a call to [pathName reset]
will raise an error.
.RE

pathName reset
.br
.RS
This is used to clear the internal contents of the widget
prior to parsing a new document. The widget is reset such
that the document tree is empty (as if no calls to 
[pathName parse] had ever been made) and no stylesheets
except the default stylesheet are loaded (as if no
invocations of [pathName style] had occured).
.RE

pathName select clear
.br
pathName select from ?nodeHandle ?index? ?
.br
pathName select to ?nodeHandle ?index? ?
.br
pathName select span
.br
.RS
The [pathName select] commands are used to query and
manipulate the widget selection.

TODO: Describe these commands.
.RE

pathName style ?options? stylesheet-text
.br
.RS
Add a stylesheet to the widgets internal configuration. The
stylesheet-text argument should contain the text of a
complete stylesheet.  Incremental parsing of stylesheets is
not supported, although of course multiple stylesheets may
be added to a single widget.

The following options are supported:
.RS

.nf
Option                   Default Value
--------------------------------------
-id <stylesheet-id>      "author"
-importcmd <script>      ""
-urlcmd    <script>      ""
.fi
.RE

The value of the -id option determines the priority taken
by the style-sheet when assigning property values to
document nodes (see chapter 6 of the CSS specification for
more detail on this process).  The first part of the
style-sheet id must be one of the strings "agent", "user"
or "author". Following this, a style-sheet id may contain
any text.  

When comparing two style-ids to determine which stylesheet
takes priority, the widget uses the following approach: If
the initial strings of the two style-id values are not
identical, then "user" takes precedence over "author", and
"author" takes precedence over "agent". Otherwise, the
lexographically largest style-id value takes precedence.
For more detail on why this seemingly odd approach is
taken, please refer to the "STYLESHEET LOADING" below.

The -importcmd option is used to provide a handler script
for @import directives encountered within the stylesheet
text. Each time an @import directive is encountered, if the
-importcmd option is set to other than an empty string, the
URI to be imported is appended to the option value and the
resulting list evaluated as a Tcl script. The return value
of the script is ignored. If the script raises an error,
then it is propagated up the call-chain to the 
[pathName style] caller.

The -urlcmd option is used to supply a script to translate
"url(...)" CSS attribute values. If this option is not set to
"", each time a url() value is encountered the URI is appended
to the value of -urlcmd and the resulting script evaluated. The
return value is stored as the URL in the parsed stylesheet.
.RE

pathName xview ?options?
.br
.RS
This command is used to query or adjust the horizontal
position of the viewport relative to the document layout.
It is identical to the [pathName xview] command
implemented by the canvas and text widgets.
.RE

pathName yview ?options?
.br
.RS
This command is used to query or adjust the vertical
position of the viewport relative to the document layout.
It is identical to the [pathName yview] command
implemented by the canvas and text widgets.
.RE

.SH "NODE COMMAND"
There are several interfaces by which a script can obtain a "node
handle".  Each node handle is a Tcl command that may be used to
access the document node that it represents. A node handle is valid
from the time it is obtained until the next call to 
[pathName reset]. The node handle may be used to query and
manipulate the document node via the following subcommands:

nodeHandle attr ?attribute?
.br
.RS
If the attribute argument is present, then return the value
of the named html attribute, or an empty string if the
attribute specified does not exist. If it is not present,
return a key-value list of the defined attributes of the
form that can be passed to [array set].
.RS

.nf
# Html code for node
<p class="normal" id="second" style="color : red">

# Value returned by [nodeHandle attr]
{class normal id second style {color : red}}

# Value returned by [nodeHandle attr class]
normal
.fi
.RE
.RE

nodeHandle child index
.br
.RS
Return the node handle for the index'th child of the node.
Children are numbered from zero upward.
.RE

nodeHandle nChild
.br
.RS
Return the number of children the node has.
.RE

nodeHandle parent
.br
.RS
Return the node handle for the node's parent. If the node
does not have a parent (i.e. it is the document root), then
return an empty string.
.RE

nodeHandle replace ? ?options? newValue?
.br
.RS
This command is used to set and get the name of the
replacement object for the node, if any. If the newValue
argument is present, then this command sets the nodes
replacement object name and returns the new value. If
newValue is not present, then the current value is
returned.

A nodes replacement object may be set to the name of a Tk
window or an empty string. If it is an empty string (the
default and usual case), then the node is rendered normally.
If the node replacement object is set to the name of a Tk
window, then the Tk window is mapped into the widget in place
of any other content (for example to implement form elements or
plugins).

The following options are supported:
.RS

.nf
Option                   Default Value
--------------------------------------
-deletecmd    <script>   "destroy <window>"
-configurecmd <script>   ""
.fi
.RE

When a replacement object is no longer being used by the
widget (e.g. because the node has been deleted or 
[pathName reset] is invoked), the value of the
-deletecmd option is evaluated as Tcl script. 

If it is not set to an empty string (the default) each time
the nodes CSS properties are recalculated, a serialized
array is appended to the value of the -configurecmd option
and the result evaluated as a Tcl command. The script
should update the replacement objects appearance where
appropriate to reflect the property values. The format of
the appended argument is {p1 v1 p2 v2 ... pN vN} where the
pX values are property names (i.e. "background-color") and
the vX values are property values (i.e. "#CCCCCC"). The
CSS properties that currently may be present in the array
are listed below. More may be added in the future.
.RS

.nf
background-color    color
font                selected
.fi
.RE

The value of the "font" property, if present in the
serialized array is not set to the value of the
corresponding CSS property. Instead it is set to the name
of a Tk font determined by combining the various
font-related CSS properties. Unless they are set to
"transparent", the two color values are guaranteed to parse
as Tk colors. The "selected" property is either true or
false, depending on whether or not the replaced object is
part of the selection or not. Whether or not an object is
part of the selection is governed by previous calls to the
[pathName select] command.

The -configurecmd callback is always executed at least once
between the [nodeHandle replace] command and when the
replaced object is mapped into the widget display.
.RE

nodeHandle tag
.br
.RS
Return the name of the Html tag that generated this
document node (i.e. "p" or "link"), or an empty string if
the node is a text node.
.RE

nodeHandle text
.br
.RS
If the node is a "text" node, return the string contained
by the node. If the node is not a "text" node, return an
empty string.
.RE

TODO: Add the "write" part of the DOM compatible interface to this
section.

.SH "STYLESHEET LOADING"
Apart from the default stylesheet that is always loaded (see the
description of the -defaultstyle option above), a script may
configure the widget with extra style information in the form of
CSS stylesheet documents. Complete stylesheet documents (it is not
possible to incrementally parse stylesheets as it is HTML document
files) are passed to the widget using the [pathName style]
command.

As well as any stylesheets specified by the application,
stylesheets may be included in HTML documents by document authors
in several ways: 
.RS
.IP * 2
Embedded in the document itself, using a <style> tag. To
handle this case an application script must register a
"script" type handler for <style> tags using the 
[pathName handler] command. The handler command should
call [pathName style] to configure the widget with the
stylesheet text.
.IP * 2
Linked from the document, using a <link> tag. To handle
this case the application script should register a "node"
type handler for <link> tags.
.IP * 2
Linked from another stylesheet, using the @import
directive. To handle this, an application needs to
configure the widget -importcommand option.
.RE
.RS

.nf
# Implementations of application callbacks to load
# stylesheets from the various sources enumerated above.
# ".html" is the name of the applications tkhtml widget.
# The variable $document contains an entire HTML document.
# The pseudo-code <LOAD URI CONTENTS> is used to indicate
# code to load and return the content located at $URI.

proc script_handler {tagcontents} {
    incr ::stylecount
    set id "author.[format %.4d $::stylecount]"
    set handler "import_handler $id"
    .html style -id $id.9999 -importcmd $handler $tagcontents
}

proc link_handler {node} {
    if {[node attr rel] == "stylesheet"} {
        set URI [node attr href]
        set stylesheet [<LOAD URI CONTENTS>]

        incr ::stylecount
        set id "author.[format %.4d $::stylecount]"
        set handler "import_handler $id"
        .html style -id $id.9999 -importcmd $handler $stylesheet
    }
}

proc import_handler {parentid URI} {
    set stylesheet [<LOAD URI CONTENTS>]

    incr ::stylecount
    set id "$parentid.[format %.4d $::stylecount]"
    set handler "import_handler $id"
    .html style -id $id.9999 -importcmd $handler $stylesheet
}

.html handler script style script_handler
.html handler node link link_handler

set ::stylecount 0

.html parse -final $document
.fi
.RE

The complicated part of the example code above is the generation of
stylesheet-ids, the values passed to the -id option of the 
[.html style] command. Stylesheet-ids are used to determine the
precedence of each stylesheet passed to the widget, and the role it
plays in the CSS cascade algorithm used to assign properties to
document nodes. The first part of each stylesheet-id, which must be
either "user", "author" or "agent", determines the role the
stylesheet plays in the cascade algorithm. In general, author
stylesheets take precedence over user stylesheets which take
precedence over agent stylesheets. An author stylesheet is one
supplied or linked by the author of the document. A user stylesheet
is supplied by the user of the viewing application, possibly by
configuring a preferences dialog or similar. An agent stylesheet is
supplied by the viewing application, for example the default
stylesheet configured using the -defaultstyle option.

The stylesheet id mechanism is designed so that the cascade can be
correctly implemented even when the various stylesheets are passed
to the widget asynchronously and out of order (as may be the case
if they are being downloaded from a network server or servers).
.RS

.nf
#
# Contents of HTML document
#

<html><head>
    <link rel="stylesheet" href="A.css">
    <style>
        @import uri("B.css")
        @import uri("C.css")
        ... rules ...
    </style>
    <link rel="stylesheet" href="D.css">
... remainder of document ...

#
# Contents of B.css
#

@import "E.css"
... rules ...
.fi
.RE

In the example above, the stylesheet documents A.css, B.css, C.css,
D.css, E.css and the stylesheet embedded in the <style> tag are all
author stylesheets. CSS states that the relative precedences of the
stylesheets in this case is governed by the following rules:
.RS
.IP * 2
Linked, embedded or imported stylesheets take precedence
over stylesheets linked, embedded or imported earlier in
the same document or stylesheet.
.IP * 2
Rules specified in a stylesheet take precedence over rules
specified in imported stylesheets.
.RE

Applying the above two rules to the example documents indicates
that the order of the stylesheets from least to most important is:
A.css, E.css, B.css, C.css, embedded <stylesheet>, D.css. For the
widget to implement the cascade correctly, the stylesheet-ids
passed to the six [pathName style] commands must sort
lexigraphically in the same order as the stylesheet precedence
determined by the above two rules. The example code above shows one
approach to this. Using the example code, stylesheets would
be associated with stylesheet-ids as follows: 
.RS

.nf
Stylesheet         Stylesheet-id
-------------------------------
A.css              author.0001.9999
<embedded style>   author.0002.9999
B.css              author.0002.0003.9999
E.css              author.0002.0003.0004.9999
C.css              author.0002.0005.9999
D.css              author.0006.9999
.fi
.RE

Entries are specified in the above table in the order in which the
calls to [html style] would be made. Of course, the example code
fails if 10000 or more individual stylesheet documents are loaded.
More inventive solutions that avoid this kind of limitation are
possible.

Other factors, namely rule specificity and the !IMPORTANT directive
are involved in determining the precedence of individual stylesheet
rules. These are completely encapsulated by the widget, so are not
described here. For complete details of the CSS cascade algorithm,
refer to [1].

.SH "INCREMENTAL DOCUMENT LOADING"
This section discusses the widget API in the context of loading a
document incrementally, for example from a network server. We
assume both remote stylesheets and image files are retrieved as
well as the document.

Before a new document (html file) is loaded, any previous document
should be purged from memory using the [pathName reset] command.
The portion of the new document that is read is passed to the
widget using the [pathName parse] command. As new chunks of the
document are downloaded, they should also be passed to 
[pathName parse]. When the final chunk of the document file is
passed to the [pathName parse] command the -final option should
be specified. This ensures node-handler callbacks (see the 
description of the [pathname handler] command above) are made
for tags that are closed implicitly by the end of the document.

The widget display is updated in an idle callback scheduled after
each invocation of the [pathName parse] command.

TODO: Finish this section.

.SH "HTML SPECIFIC PROCESSING"
TODO: Detail implicit opening and closing tag rules here. LATER: If 
only I could figure them out...

.SH "REFERENCES"

[1] CSS2 specification.