evolution/help/white-papers/widgets/e-table.sgml

<!doctype article PUBLIC "-//Davenport//DTD DocBook V3.0//EN" [
<!entity Evolution "<application>Evolution</application>">
<!entity ETable "<classname>ETable</classname>">
<!entity ETableModel "<classname>ETableModel</classname>">
<!entity ETableSimple "<classname>ETableSimple</classname>">
<!entity ETableHeader "<classname>ETableHeader</classname>">
<!entity ETableSpecification "<classname>ETableSpecification</classname>">
<!entity ETableCol "<classname>ETableCol</classname>">
]>

<article class="whitepaper" id="e-table">

  <artheader>
    <title>The ETable Widget</title>

    <authorgroup>
      <author>
	<firstname>Chris</firstname>
	<surname>Lahey</surname>
	<affiliation>
	  <address>
	    <email>clahey@helixcode.com</email>
	  </address>
	</affiliation>
      </author>
      <author>
	<firstname>Miguel</firstname>
	<surname>de Icaza</surname>
	<affiliation>
	  <address>
	    <email>miguel@helixcode.com</email>
	  </address>
	</affiliation>
      </author>
    </authorgroup>

    <copyright>
      <year>2000</year>
      <holder>Helix Code, Inc.</holder>
    </copyright>

  </artheader>

  <sect1 id="introduction">
    <title>Introduction</title>

    <para>
      &ETable; is a table widget on steroids. It is intended to provide
      all the table functionality needed throughout &Evolution;, and
      hopefully be general purpose enough to be used in other projects.
    </para>

    <para>
      &ETable; provides a lot of interactive control over the data in the
      table.  Without any work from the programmer, &ETable; provides
      rearrangeable columns and editable data.  When finished, &ETable; will
      also provide, again with no programmer intervention, easy interactive
      sorting and grouping.  
    </para>

    <para>
      &ETable; gives you a great deal of functionality, flexibility, and
      power. Most of this power is internal to the widget, but some of
      the flexibility requires a bit of work by the programmer.
      However, once you learn it, &ETable; is not very hard at all to
      use.
    </para>

    <para>
      &ETable;'s power comes from the fact that it is fully
      model/view/controller based.  Various models are involved into
      the process of rendering the information, and various views are
      provided.  The programmer has a wide range of options: from the
      most finely hand-tuned table to a generic all-encompasing widget
      that takes over most of tasks.  It is up to the programmer: he
      can use the simple to use &ETable; widget that takes care of
      everything in a generic way, or he can use the various
      components to roll his own tabular display.
    </para>

    <para>
      &ETable; ships with a standard set of information renderers:
      strings, bitmaps, toggle-buttons, check-boxes, and multi-line
      strings.  But the programmer can write and implement his own
      renderer for his information.  This means that by default
      &ETable; provides the basic display facilities that programmers
      required, but they offer the programmer a complete freedom to
      incorporate new cell renderers.
    </para>
    
  </sect1>

  <sect1 id="model">
    <title>ETableModel</title>

    <para>
      The data back end for the &ETable; is an &ETableModel;. The
      &ETableModel is an abstract interface that acts as the
      information repository for the various &ETable components.
    </para>
    
    <para>
      To use &ETable; you have to create a subclass of the abstract
      &ETableModel; class.  However, to save you the work of defining
      a new <classname>GtkClass</classname> every time you use
      &ETable, there is a predefined subclass of &ETableModel; called
      &ETableSimple; which simply takes a list of function callbacks
      to perform the various operations.
    </para>

  </sect1>
    
  <sect1 id="columns">
    <title>Columns</title>

    <para>
      There are two different meanings to the word "column". The first
      is the model column (defined by the &ETableCol: object). A model
      column describes how it maps to the column in the &ETableModel;
      as well as containing information about its properties (name,
      resizability, resize dimensions, and a renderer for this
      specific columns). 
    </para>

    <para>
      &ETable; distinguishes between a model column index, and a view
      column index.  The former reflects the column in which the data
      is stored in the &ETableModel;  The later represents the actual
      location at which the column is being displayed in the screen.
    </para>

    <para>
      Each view column index corresponds to a specific model column,
      though a model column may have any number of view columns
      associated with it (including zero).  For example the same
      column might be rendered twice, or the data from one column
      could be used to display different bits of information
    </para>

    <para>
      The view column does not necessarily depend on only one model
      column. In some cases, the view column renderer can be given a
      reference to another model column to get extra information about
      its display. For example, a mail program could display deleted
      messages with a line through them by creating a model column
      with no corresponding view column that told whether or not the
      message is deleted, and then having the text column
      strikethrough the display if the invisible column had a value
      corresponding to "deleted".
    </para>

    <para>
      The view column also specifies a few other pieces of
      information. One piece of information is the renderer. &ETable;
      provides a number of renderers to choose from, or you can write
      your own. Currently, there are renderers for text, image sets,
      and checkboxes.
    </para>

    <para>
      The view column also includes information about the header.
      There are two types of headers: text, and pixbuf. The first
      allows you to specify a string which is rendered in the header.
      The second allows you to specify an image to copy into the
      header.
    </para>
  </sect1>

  <sect1 id="header">
    <title>Header</title>

    <para>
      The &ETableHeader; represents the header information for the
      table. The &ETableHeader; is used in two different ways. The
      first is the in the <structfield>full_header</structfield>
      element of an &ETable;. This is the list of possible columns in
      the view. You add each of your columns to this &ETableHeader;
      and then pass it into the &ETable;.
    </para>

    <para>
      The second use is completely internal. &ETable; uses another
      &ETableHeader; to store the actual displayed columns. Many of
      the &ETableHeader; functions are for this purpose. The only
      functions that users of the library should need to use are
      <function>e_table_header_new</function> and
      <function>e_table_header_add_col</function>.
    </para>
  </sect1>

  <sect1 id="layout">
    <title>Layout Specification</title>

    <para>
      &ETable; uses an &ETableSpecification; to layout the columns of
      the widget. The &ETableSpecification; is specified as XML data
      passed into the &ETable; as a string.
    </para>

    <para>
      The most powerful part of the &ETableSpecification; is that when
      finished, &ETable; will allow you to get a copy of an
      &ETableSpecification; that describes the current view of the
      tree. This allows the developer to save the current view so that
      next time the user opens this table, they find it in exactly the
      state that they left it.
    </para>

    <para>
      The XML specification allows for a number of things. First, it
      allows you to pick a set of default columns to be shown. Thus,
      even if you had hundreds of pieces of data, you could choose to
      only display a few that fit on the screen by default.
    </para>

    <para>
      The second major thing that the &ETableSpecification; allows you
      to specify is the column grouping and sorting. &ETable; has a
      powerful mechanism for allowing the user to choose columns to
      group by, thus allowing multiple columns of sorting, as well as
      visual grouping of similar elements and interactive selection of
      what data to display.
    </para>

    <para>
      The grouping in &ETableSpecification; is specified as a
      hierarchy of columns to group by. Each level of the hierarchy
      lets you sort by a particular column, either ascending or
      descending. All levels except the last cause the canvas to group
      by the given column.
    </para>

    <para>
      An example &ETableSpecification; follows.
    </para>

    <programlisting>
      &lt;ETableSpecification&gt;
        &lt;columns-shown frozen_columns="2"&gt;
	  &lt;column&gt; 0 &lt;/column&gt;
	  &lt;column&gt; 1 &lt;/column&gt;
	  &lt;column&gt; 2 &lt;/column&gt;
	  &lt;column&gt; 3 &lt;/column&gt;
	  &lt;column&gt; 4 &lt;/column&gt;
	&lt;/columns-shown&gt;
	&lt;grouping&gt;
	  &lt;group column="3" ascending="1"&gt;
	    &lt;group column="4" ascending="0"&gt;
	      &lt;leaf column="2" ascending="1"/&gt;
	    &lt;/group&gt;
	  &lt;/group&gt;
	&lt;/grouping&gt;
      &lt;/ETableSpecification&gt;
    </programlisting>

    <para>
      This example has 5 columns which are initially in order. It has
      2 levels of grouping. The first is grouped by the 4th column
      (all indexes are 0 based) and sorts those groups in ascending
      order. Inside those groups, the data is grouped by the fifth
      column and sorted in descending order of the fifth column.
      Finally, the data in those groups is sorted by the third column
      in ascending order. Due to the "frozen_columns" attribute on the
      columns-shown element, the user will not be
      able to rearrange the first two columns. They will always be the
      first two.
    </para>
  </sect1>

  <sect1 id="conclusion">
    <title>Conclusion</title>

    <para>
      All in all, &ETable; is a very powerful widget. Once you learn
      to use it, you have access to a vast amount of power requiring a
      comparatively small amount of work.
    </para>
  </sect1>
</article>