Describes guidelines and conventions for implementing ITextProvider, ITextProvider2, and ITextRangeProvider, including information about properties and methods. The Text control pattern enables applications and controls to expose a simple text object model, enabling clients to retrieve textual content, text attributes, and embedded objects from text-based controls.
To support the Text control pattern, controls implement the ITextProvider and ITextProvider2 interfaces. Control types that should support the Text control pattern include the Edit and Document control types, and any other control type that enables the user to enter text or select read-only text.
The Text control pattern can be used with other Microsoft UI Automation control patterns to support several types of embedded objects in the text, including tables, hyperlinks, and command buttons.
The ITextProvider and ITextProvider2 interfaces include a number of methods for acquiring text ranges. A text range is an object that represents a contiguous span of text—or multiple, disjoint spans of text—in a text container. One ITextProvider method acquires a text range that represents the entire document, while others acquire text ranges that represent some portion of the document, such as the selected text, the visible text, or an object embedded in the text.
A text range object is represented by the TextRange control pattern, which is implemented through the ITextRangeProvider interface. The TextRange control pattern provides methods and properties used to expose information about the text in the range, move the endpoints of the range, select or deselect text, scroll the range into view, and so on.
For more information about the Text and TextRange control patterns, see UI Automation Support for Textual Content.
Starting with Windows 8.1 providers can implement the ITextRangeProvider2 interface. This enables invoking context menus that are associated with a text range. This supports scenarios such as text autocorrection or Input Method Editor (IME) candidate selection.
This topic contains the following sections.
- Implementation Guidelines and Conventions
- Required Members for ITextProvider
- Required Members for ITextRangeProvider
- Supporting Text Ranges
- Interoperability with the System Caret
- Related topics
When implementing the Text control pattern, note the following guidelines and conventions:
- Any control that enables access to text (for example, entering text or selecting read-only text) should support the Text control pattern.
- The Text control pattern can be used with any UI element that presents text, even a static label on a standard button control. However, it is not required on static text controls that cannot be selected or do not have an insertion point.
- To ensure that text is fully accessible, controls that implement ITextProvider should also support the IValueProvider interface. IValueProvider complements ITextProvider by providing a programmatic way to change the text. It also offers greater compatibility with assistive technology client applications, including those based on legacy technologies such as Microsoft Active Accessibility. When both control patterns are implemented, the TextChanged event ( UIA_Text_TextChangedEventId) and AutomationPropertyChanged event ( UIA_AutomationPropertyChangedEventId) are equivalent for the Value property ( UIA_ValueValuePropertyId). Both events must be supported.
- The Text control pattern supports only one stream of text and one viewport per control. If the application offers multiple views of document in panes, each view (control) should support ITextProvider independently.
- The ITextProvider::GetSelection method may return a single text range representing the currently selected text. If a control supports the selection of multiple, noncontiguous spans of text, the GetSelection method should return an array that contains one ITextRangeProvider interface for each selected span of text.
- The Text control pattern represents the insertion point as a degenerate (empty) text range. The ITextProvider::GetSelection method should return a degenerate text range when the insertion point exists and no text is selected. For more information, see Interoperability with the System Caret.
- The ITextProvider::GetVisibleRanges method may return a single text range if a contiguous range of text is visible in the viewport, or it may return an array of disjoint text ranges representing multiple, partially visible lines of text.
- The ITextProvider::RangeFromChild method should return a degenerate range if the child element contains no text. Because an embedded object can include text, the RangeFromChild method may not always return a degenerate text range. For more information, see How UI Automation Exposes Embedded Objects.
- The ITextProvider::RangeFromPoint method performs hit testing in the document area using the specified screen coordinates. The resulting text range should be consistent with the insertion point or selection that would result from clicking the location at the specified screen coordinates. For example, if an image resides at the specified screen coordinates, the resulting text range should be the same as the text range that the ITextProvider::RangeFromChild method would acquire for the image. Similarly, if a client application requests a text range for the location at the center of the system caret (the insertion point), the resulting text range should be the same as the system caret location.
- The ITextProvider::DocumentRange property should always provide a text range that includes all of the text supported by the corresponding ITextProvider implementation.
- The UIA_Text_TextChangedEventId event must be raised after any text change occurs, even if the change is not visible in the viewport. For example, the provider should raise the event even if the user pastes the exact same text over the selected text.
- The UIA_Text_TextSelectionChangedEventId must be raised whenever the selection of text changes, or whenever the insertion point (caret) moves among the text.
When implementing the TextRange control pattern, note the following guidelines and conventions:
- All methods of the TextRange control pattern should perform text operations regardless of the visibility state of the text. The visibility of a text range can always be determined by querying the IsHidden text attribute ( UIA_IsHiddenAttributeId).
- If possible, a provider should ensure that any text changes, such as deletions, insertions, and moves, are reflected in the associated text range objects (instances of ITextRangeProvider interface) and raise a UIA_Text_TextChangedEventId event. Clients may use the event as a hint to confirm editorial changes made to the text of a control.
- All of the text range objects used by the ITextRangeProvider::Compare, CompareEndpoints, and MoveEndpointByRange methods must be peers of the same Text control pattern implementation.
- Although not required, the pRetVal value retrieved by the ITextRangeProvider::CompareEndpoints method can indicate the distance, in characters ( TextUnit_Character), between the two endpoints. However, client applications should not depend on the accuracy of pRetVal beyond its positive or negative value.
- The ITextRangeProvider::ExpandToEnclosingUnit, Move, and MoveEndpointByUnit methods require careful consideration of the specified text unit. For more information, see Manipulating a TextRange by Text Unit.
- For implementation requirements related to the ITextRangeProvider::Select, AddToSelection, and RemoveFromSelection methods, see Selecting Text in a Text Range.
- The ITextRangeProvider::FindText and FindAttribute methods search forward or backward for a single matching text string or text attribute. They should return NULL if no match is found.
- The ITextRangeProvider::GetAttributeValue method must return the address acquired from the UiaGetReservedMixedAttributeValue or UiaGetReservedNotSupportedValue function if the associated attribute varies in the range, or if the attribute is not supported by the text control. The TextRange control pattern specification does not allow adding new text-attribute identifiers or changing how the existing attributes are defined.
- If possible, the ITextRangeProvider::GetBoundingRectangles method should return an array that contains one bounding rectangle for each fully or partially visible line of text in a text range. If this is not possible, the provider can return an array that contains the bounding rectangles of only fully visible lines; however, this limits a client application's ability to accurately describe how the text is being presented on the screen.
- The ITextRangeProvider::GetChildren method should return all child elements embedded in a text range, but it does not need to return any children of the child elements. For example, if a text range contains a table that has a number of child cells, the GetChildren method may return just the table element and not the cell elements. For performance or architectural reasons, a provider may not be able to expose all embedded objects that are hosted in a document in the automation tree. In this case, the provider should at least support the enumerating of child objects through the GetChildren method and, as an option, support the VirtualizedItem control pattern for de-virtualization support.
- The ITextRangeProvider::GetEnclosingElement method typically returns the text provider that supplies the text range. However, if the text provider supports child objects such as tables or hyperlinks, the enclosing element could be a descendant of the text provider. The element returned by GetEnclosingElement should be the element closest to the given text range. For example, if the text range is in a cell of a table, GetEnclosingElement should return the containing cell instead of the table element.
- The ITextRangeProvider::GetText method should return the plain text in the range. For more information, see Acquiring Text from a Text Range.
- Calling ITextRangeProvider::ScrollIntoView should align the text in the viewport of the text control as specified by the alignToTop parameter. Although there is no requirement in terms of horizontal alignment, the text range should be visible both horizontally and vertically. When evaluating the alignToTop parameter, a provider must take into account the orientation of the text control and the flow direction of the text. For example, if alignToTop is TRUE for a vertically oriented text control containing text that flows from right to left, the provider should align the text range with the right side of the viewport.
- When moving through a document by TextUnit_Line, if the text range enters an embedded table, each line of text in a cell should be treated as a line.
The following properties and methods are required for implementing the ITextProvider interface.
|Required members||Member type||Notes|
The following additional properties and methods are required for implementing the ITextProvider2 interface.
|Required members||Member type||Notes|
The following properties and methods are required for implementing the ITextRangeProvider interface.
|Required members||Member type||Notes|
The following additional properties and methods are required for implementing the ITextRangeProvider2 interface.
|Required members||Member type||Notes|
|ShowContextMenu||Method||See "Implementing ShowContextMenu" section|
The TextRange control pattern has no associated events.
The ITextRangeProvider interface provides several methods for manipulating and navigating text ranges in a text-based control. The ITextRangeProvider::Move, MoveEndpointByUnit, and ExpandToEnclosingUnit methods move a text range or one of its endpoints by the specified text unit, such as character, word, paragraph, and so on. For more information, see UI Automation Text Units.
Despite its name, the ITextRangeProvider::ExpandToEnclosingUnit method does not necessarily expand a text range. Instead, it "normalizes" a text range by moving the endpoints so that the range encompasses the specified text unit. The range is expanded if it is smaller than the specified unit, or shortened if it is longer than the specified unit. It is critical that the ExpandToEnclosingUnit method always normalizes text ranges in a consistent manner; otherwise, other aspects of text range manipulation by text unit would be unpredictable. The following diagram shows how ExpandToEnclosingUnit normalizes a text range by moving the endpoints of the range.
If the text range starts at the beginning of a text unit and ends at the beginning of, or before, the next text unit boundary, the ending endpoint is moved to the next text unit boundary (see 1 and 2 in the previous diagram).
If the text range starts at the beginning of a text unit and ends at, or after, the next unit boundary, the ending endpoint stays or is moved backward to the next unit boundary after the starting endpoint (see 3 and 4 in the previous illustration). If there is more than one text unit boundary between the starting and ending endpoints, the ending endpoint is moved backward to the next unit boundary after the starting endpoint, resulting in a text range that is one text unit in length.
If the text range starts in a middle of the text unit, the starting endpoint is moved backward to the beginning of the text unit, and the ending endpoint is moved forward or backward, as necessary, to the next unit boundary after the starting endpoint (see 5 through 8 in the previous diagram).
When the ITextRangeProvider::Move method is called, the provider normalizes the text range by the specified text unit, using the same normalization logic as the ExpandToEnclosingUnit method. Then, the provider moves the range backward or forward by the specified number of text units. When moving the range, the provider should ignore the boundaries of any embedded objects in the text. (However, the unit boundary itself may be affected by the existence of an embedded object). The following diagram demonstrates how the Move method moves a text range, unit by unit, across embedded objects and text unit boundaries.
The ITextRangeProvider::MoveEndpointByUnit method moves one of the endpoints forward or backward by specified text unit, as the following illustration shows.
The ITextRangeProvider::MoveEndpointByRange method enables a client application to set one endpoint of a text range to same location as the specified endpoint of a second text range.
The ITextRangeProvider interface includes several methods for controlling the selection of text in a text-based control.
The ITextRangeProvider::Select method should select the text that corresponds to a text range, and remove the previous selection, if any, from the text control. If Select is called on a degenerate text range, the provider should move the insertion point to the location of the text range without selecting any text.
If a control supports the selection of multiple, disjoint spans of text, the ITextRangeProvider::AddToSelection and RemoveFromSelection methods add text ranges to, and remove them from, the collection of selected text ranges. If the control only supports one selected text range at a time, but the selection operation would result in the selection of multiple disjoint text ranges, the provider should either return an E_INVALIDOPERATION error, or should extend or truncate the current selection. The ITextProvider::SupportedTextSelection property should indicate whether a control supports the selection of single or multiple spans of text, or none at all.
If a text-based control supports text insertions, calling ITextRangeProvider::AddToSelection or RemoveFromSelection on a degenerate text range in the control should move the insertion point but should not select any text.
The ITextRangeProvider::GetText method should return the plain text of a text range. The plain text should include all control characters found in the source text, such as carriage returns and the Unicode left-to-right mark (LRM). The plain text should not include any markup tags such as HTML that may be present in the source text. Also, any escape codes in the source text should be converted to the plain text equivalents. For example, " " should be converted to a simple space character.
If an embedded object spans a range of text, the plain text should include the inner text of the object, but not the alternative text (the name property of the embedded object) because it might be inconsistent with the descriptive inner text. For more information, see How UI Automation Exposes Embedded Objects.
ITextRangeProvider2::ShowContextMenu should always show the context menu at the beginning end point of the range. This should be equivalent to what would happen if the user pressed the context menu key or SHIFT + F10 with the insertion point at the beginning of the range.
If showing the context menu would typically result in the insertion point moving to a given location, then it should do so for programmatically invoking ShowContextMenu for UI Automation support also.
Correctly supporting the insertion point is critical to many client applications, including those not based on UI Automation. In the Text control pattern, the insertion point is represented by a degenerate (empty) text range at the position of the system caret. When the insertion point moves, a control should raise the TextSelectionChanged event ( UIA_Text_TextSelectionChangedEventId). Some client applications depend on this event to monitor the location of the insertion point, and to keep track of the text selection.
When a control contains selected text, the current design of the Text control pattern does not provide a way to directly associate the location of the insertion point with a particular text range. The provider must keep track of the text selection and set the location of the insertion point appropriately. Because selecting text is typically done by moving the insertion point while holding down the SHIFT or CTRL key, or both, a provider can track the text selection by checking the state of these keys whenever the selection changes.
Because the accessibility framework provides built-in support for the system caret but not for a custom caret, text-based controls should use the system caret whenever possible. Controls that use a custom caret can ensure that the caret is accessible by creating a system caret that has the same dimensions as the custom caret, and positioning the system caret at the same location in the UI of the control as the custom caret—that is, at the insertion point. As an alternative, a control that uses a custom caret can implement a Microsoft Active Accessibility provider for OBJID_CARET to provide accessibility information directly for your custom caret.
For more information about the system caret, see Carets.
The screen coordinates of the center of the system caret bitmap should always match the location of the insertion point. That way, a client can use the caret screen coordinates in a call to ITextProvider::RangeFromPoint to retrieve a text range that accurately represents the location of the insertion point.
- Control Types and Their Supported Control Patterns
- Text Control Type
- TextEdit Control Pattern
- TextChild Control Pattern
- UI Automation Control Patterns Overview
- UI Automation Support for Textual Content
- UI Automation Tree Overview