About Text Object Model

The Text Object Model (TOM) defines a set of text manipulation interfaces that are supported in varying degrees by several Microsoft text solutions, including the rich edit control. This topic provides a high-level overview of the TOM. It discusses the following topics.

• TOM Version 2 Objects
• TOM Interface Conventions
• The tomBool Type
• Math BuildUp and Build Down
• TOM RTF
• Finding Rich Text
• TOM Accessibility
  • Interface from Running Object Table
  • Interface from Window Messages
  • Accessibility Oriented Methods
• Character Match Sets

TOM Version 2 Objects

TOM version 2 (TOM 2) extends the original text object model; the new interfaces are derived from the old ones. The updated TOM API includes support for new character and paragraph format properties, a table model, multiple selection, and inline object support for math and ruby.

The top-level TOM 2 object is defined by the ITextDocument2 interface, which has methods for creating and retrieving objects lower in the object hierarchy. For simple plain-text processing, you can obtain an ITextRange2 object from an ITextDocument2 object and do most everything with that. If you need to add rich-text formatting, you can obtain ITextFont2 and ITextPara2 objects from an ITextRange2 object. ITextFont2 provides the programming equivalent of the Microsoft Word format-font dialog, and ITextPara2 provides the equivalent of the Word format-paragraph dialog.

In addition to these three lower-level objects, TOM 2 has a selection object (ITextSelection2), which is an ITextRange2 object with selection highlighting and some UI-oriented methods.

The range and selection objects include screen-oriented methods that enable programs to examine text on screen or text that could be scrolled onto the screen. These capabilities help make text accessible to people with impaired vision, for example.

Each interface that has the 2 suffix inherits from the corresponding interface without the 2 suffix. For example, ITextDocument2 inherits from ITextDocument.

The TOM 2 objects have the following hierarchy.

ITextDocument2         Top-level editing object
    ITextRange2        Primary text interface: a range of text
        ITextFont2     Character-attribute interface
        ITextPara2     Paragraph-attribute interface
        ITextRow       Table interface
    ITextSelection2    Screen highlighted text range
        ITextRange2    Selection inherits all range methods
    ITextDisplays      Displays collection (not yet defined)
    ITextStrings       Rich-text strings collection
    ITextStoryRanges2  Enumerator for stories in document

An ITextDocument2 object describes one or more contiguous ranges of text called stories. Stories represent various parts of a document, such as the main text of the document, headers and footers, footnotes, annotations, and rich-text scratch pads. A scratch pad story is used when translating between linearly formatted math expressions and a built-up form. A scratch pad story is also used when saving the contents of a range that is the current copy source when the contents are about to be changed.

An ITextRange2 object is defined by its start and end character-position offsets and a story object. It does not exist independently of its parent story object, although its text can be copied to the clipboard or to other targets. A text range object is different from spreadsheet and other range objects, which are defined by other kinds of offsets; for example, row/column or graphics position (x, y). A text range object can modify itself in various ways, can return a duplicate of itself, and it can be commanded to copy its start and end character positions and its story pointer to the current selection.

An explicit story object is not needed, since an ITextRange object can always be created to represent any given story. In particular, the ITextDocument object can create an ITextStoryRanges object to enumerate the stories in the document in terms of ranges with start and end character position values that describe complete stories (such as, 0 and tomForward).

With an ITextStoryRanges2 object, an explicit story object is not needed, since the each story is described by an ITextRange2 object. In particular, the ITextDocument2 object can create an ITextStoryRanges2 object to enumerate the stories in the document in terms of ranges with start and end character position values that describe complete stories (such as, 0 and tomForward).

The ITextRow interface together with the ITextRange::Move and ITextRange::Expand methods give the capability to insert, query, and change tables.

TOM Interface Conventions

All TOM methods return HRESULT values. In general, the TOM methods return the following standard values.

• E_OUTOFMEMORY
• E_INVALIDARG
• E_NOTIMPL
• E_FILENOTFOUND
• E_ACCESSDENIED
• E_FAIL
• CO_E_RELEASED
• NOERROR (same as S_OK)
• S_FALSE

Be aware that if the editing instance associated with a TOM object such as ITextRange is deleted, the TOM object becomes useless, and all its methods return CO_E_RELEASED.

In addition to the HRESULT return values, many methods include out parameters, which are pointers used to return values. For all interfaces, you should check all pointer parameters to ensure that they are nonzero before using them. If you pass a null value to a method that requires a valid pointer, the method returns E_INVALIDARG. Optional out pointers with null values are ignored.

Use methods with Get and Set prefixes to get and set properties. Boolean variables use tomFalse (0) for FALSE, and tomTrue (-1) for TRUE.

TOM constants are defined in the tomConstants enumeration type and begin with the prefix tom, for example tomWord.

The tomBool Type

Many TOM methods use a special type of variable called "tomBool" for rich-text attributes that have binary states. The tomBool type is different from the Boolean type because it can take four values: tomTrue, tomFalse, tomToggle, and tomUndefined. The tomTrue and tomFalse values indicate true and false. The tomToggle value is used to toggle a property. The tomUndefined value, more traditionally called NINCH, is a special no-input, no-change value that works with longs, floats, and COLORREFs. For strings, tomUndefined (or NINCH) is represented by the null string. For property setting operations, using tomUndefined does not change the target property. For property getting operations, tomUndefined means that the characters in the range have different values (it gives the grayed check box in property dialog boxes).

Math BuildUp and Build Down

You can use the ITextRange2::BuildUpMath method to convert linearly formatted math expressions into built-up versions. The ITextRange2::Linearize method does the opposite conversion, called linearization or build down, to convert built-up versions of math expressions back to linear format. The math build down capability is useful when you need to export plain text or to enable certain types of editing.

TOM RTF

In TOM, rich-text exchange can be accomplished by sets of explicit method calls or by transfers of rich text in the Rich Text Format (RTF). This section gives tables of RTF control words for paragraph properties and for character properties.

TOM RTF Paragraph Control Words

TOM RTF Character Format Control Words

Finding Rich Text

You can use TOM methods to find rich text as defined by a range of text. Finding such rich text exactly is often needed in word processing, although it has never been fulfilled in a "what you see is what you get" (WYSIWYG) word processor. There is clearly a larger domain of rich-text matching that allows for some character formatting properties to be ignored (or to include paragraph formatting and/or object content), but such generalizations are beyond the scope of this section.

One purpose for this functionality is to use a rich-text Find dialog box to define the rich text you want to locate in a document. The dialog box would be implemented using a rich edit control and TOM methods would be used to carry out the search through the document. You could either copy the desired rich text from the document into the Find dialog box, or enter and format it directly in the Find dialog box.

The following example shows how to use TOM methods to find text containing combinations of exact character formatting. The algorithm searches for the plain text in the match range, which is named pr1. If the plain text is found, it is pointed to by a trial range, which is named pr2. Then, two insertion-point ranges (prip1 and prip2) are used to walk through the trial range comparing its character formatting to that of pr1. If they match exactly, the input range (given by ppr) is updated to point at the trial range's text and the function returns the count of characters in the matched range. Two ITextFont objects, pf1 and pf2, are used in the character-formatting comparison. They are attached to the insertion-point ranges prip1 and prip2.

LONG FindRichText (
    ITextRange **ppr,             // Ptr to range to search
    ITextRange *pr1)              // Range with rich text to find
{
    BSTR        bstr;             // pr1 plain-text to search for
    LONG        cch;              // Text string count
    LONG        cch1, cch2;       // tomCharFormat run char counts
    LONG        cchMatch = 0;     // Nothing matched yet
    LONG        cp;               // Handy char position
    LONG        cpFirst1;         // pr1 cpFirst
    LONG        cpFirst2;         // pr2 cpFirst
    ITextFont  *    pf1, *pf      // Fonts corresponding to IPs prip1 and prip2
    ITextRange *pr2;              // Range duplicate to search with
    ITextRange *prip1, *prip      // Insertion points to walk pr1, pr2

    if (!ppr || !*ppr || !pr1)
        return E_INVALIDARG;

    // Initialize range and font objects used in search
    if ((*ppr)->GetDuplicate(&pr2)    != NOERROR ||
        pr1->GetDuplicate(&prip1)     != NOERROR ||
        pr2->GetDuplicate(&prip2)     != NOERROR ||
        prip1->GetFont(&pf1)          != NOERROR ||
        prip2->GetFont(&pf2)          != NOERROR ||
        pr1->GetText(&bstr)           != NOERROR )
    {
        return E_OUTOFMEMORY;
    }

    pr1->GetStart(&cpFirst1);

    // Keep searching till rich text is matched or no more plain-text hits
    while(!cchMatch && pr2->FindText(bstr, tomForward, 0, &cch) == NOERROR)
    {
        pr2->GetStart(&cpFirst2);                 // pr2 is a new trial range
        prip1->SetRange(cpFirst1, cpFirst1);      // Set up IPs to scan match
        prip2->SetRange(cpFirst2, cpFirst2);      //  and trial ranges

        while(cch > 0 &&
            pf1->IsEqual(pf2, NULL) == NOERROR)   // Walk match & trial ranges
        {                                         //  together comparing font
            prip1->GetStart(&cch1);               //  properties
            prip1->Move(tomCharFormat, 1, NULL);
            prip1->GetStart(&cp);
            cch1 = cp - cch1;                     // cch of next match font run

            prip2->GetStart(&cch2);
            prip2->Move(tomCharFormat, 1, NULL);
            prip2->GetStart(&cp);
            cch2 = cp - cch2;                      // cch of next trial font run

            if(cch1 < cch)                         // There is more to compare
            {
                if(cch1 != cch2)                   // Different run lengths:
                    break;                         //  no formatting match
                cch = cch - cch1;                  // Matched format run
            }
            else if(cch2 < cch)                    // Trial range format run too
                break;                             //  short

            else                                   // Both match and trial runs
            {                                      //  reach at least to match
                pr2->GetEnd(&cp);                  //  text end: rich-text match
                (*ppr)->SetRange(cpFirst2, cp)     // Set input range to hit
                cchMatch = cp - cpFirst2;          //  coordinates and return
                break;                             //  length of matched string
            }
        }
    }
    pr2->Release();
    prip1->Release();
    prip2->Release();
    pf1->Release();
    pf2->Release();
    SysFreeString(bstr);

    return cchMatch;
}

TOM Accessibility

TOM provides accessibility support through the ITextSelection and ITextRange interfaces. This section describes methods that are useful for accessibility as well as how a program can determine the x, y screen position of an object.

Since UI-based accessibility programs typically work with the screen and the mouse, a common concern is to find the corresponding ITextDocument interface for the current mouse location (in screen coordinates). The following sections present two ways to determine the proper interface:

• Through the running-object table
• Through the EM_GETOLEINTERFACE message, which works for windowed rich edit instances, provided the client is in the same process space (no marshaling is needed)

For more information, see the Microsoft Active Accessibility specification. After you obtain an object from a screen position, you can use for an ITextDocument interface and call the RangeFromPoint method to get an empty range object at the cp corresponding to the screen position.

Interface from Running Object Table

A running object table (ROT) tells what object instances are active. By querying this table, you can accelerate the process of connecting a client to an object when the object is already running. Before programs can access TOM interfaces through the running object table, a TOM instance with a window needs to register in the ROT using a moniker. You construct the moniker from a string containing the hexadecimal value of its HWND. The following code sample shows how to do this.

// This TOM implementation code is executed when a new windowed 
// instance starts up. 
// Variables with leading underscores are members of this class.

HRESULT hr;
OLECHAR szBuf[10];            // Place to put moniker
MONIKER *pmk;

hr = StringCchPrintf(szBuff, 10, "%x", _hwnd);
if (FAILED(hr))
{
    //
    // TODO: write error handler
    //
}
CreateFileMoniker(szBuf, &pmk);
OleStdRegisterAsRunning(this, pmk, &_dwROTcookie);
....................
 
// Accessibility Client: 
//    Find hwnd for window pointed to by mouse cursor.

GetCursorPos(&pt);
hwnd = WindowFromPoint(pt);

// Look in ROT (running object table) for an object attached to hwnd

hr = StringCchPrintf(szBuff, 10, "%x", hwnd);
if (FAILED(hr))
{
    //
    // TODO: write error handler
    //
}
CreateFileMoniker(szBuf, &pmk);
CreateBindContext(0, &pbc);
pmk->BindToObject(pbc, NULL, IID_ITextDocument, &pDoc);
pbc->Release();

if( pDoc )
{
    pDoc->RangeFromPoint(pt.x, pt.y, &pRange);
    // ...now do whatever with the range pRange
}

Interface from Window Messages

The EM_GETOLEINTERFACE message provides another way to obtain an IUnknown interface for an object at a given screen position. As described in Interface from Running Object Table, you get an HWND for the screen position and then send this message to that HWND. The EM_GETOLEINTERFACE message is rich edit-specific and returns a pointer to an IRichEditOle interface in the variable addressed by lParam.

Tip If a pointer is returned (be sure to set the object to which lParam points to null before sending the message), you can call its IUnknown::QueryInterface method to obtain an ITextDocument interface. The following code sample illustrates this approach.

    HWND    hwnd;
    ITextDocument *pDoc;
    ITextRange *pRange;
    POINT    pt;
    IUnknown *pUnk = NULL;
    
    GetCursorPos(&pt);
    hwnd = WindowFromPoint(pt);
    SendMessage(hwnd, EM_GETOLEINTERFACE, 0, (LPARAM)&pUnk);
    if(pUnk && 
        pUnk->QueryInterface(IID_ITextDocument, &pDoc) == NOERROR)
    {
        pDoc->RangeFromPoint(pt.x, pt.y, &pRange);
        //  ... continue with rest of program
    }

Accessibility Oriented Methods

Some TOM methods are particularly useful for navigating around the screen, while other TOM methods enhance what you can do when you arrive at places of interest. The following table describes the most useful methods.

Character Match Sets

The variant parameter of the various Move* methods in ITextRange, such as MoveWhile and MoveUntil, can take an explicit string or a character-match set 32-bit index. The indexes are defined by either Unicode ranges or GetStringTypeEx character sets. The Unicode range starting at n and of length l (< 32768) is given by the index n + (l << 16) + 0x80000000. For example, basic Greek letters are defined by CR_Greek = 0x805f0370 and printable ASCII characters are defined by CR_ASCIIPrint = 0x805e0020. In addition, the MoveWhile and MoveUntil methods let you rapidly bypass a span of characters in any GetStringTypeEx character set, or in a span of characters that is not in any one of these character sets.

The GetStringTypeEx sets are specified by the values for Ctype1, Ctype2, and Ctype3, and are defined as follows.

Specifically, Ctype1 can be any combination of the following.

The Ctype2 types support proper layout of Unicode text. The direction attributes are assigned so that the bidirectional layout algorithm standardized by Unicode produces accurate results. These types are mutually exclusive. For more information about the use of these attributes, see The Unicode Standard: Worldwide Character Encoding, Volumes 1 and 2, Addison-Wesley Publishing Company: 1991, 1992.

The Ctype3 types are intended to be placeholders for extensions to the POSIX types required for general text processing or for the standard C library functions.

An Edit Development Kit (EDK) could include pVar index definitions for the following ranges described in the Unicode Standard.