KTextEditor::CodeCompletionModel Class

ItemSelected Whenever an item is selected, this will be requested from the underlying model. It may be used as a simple notification that the item was selected. Above that, the model may return a QString, which then should then contain html-code. A html-widget will then be displayed as a one- or two-liner under the currently selected item(it will be partially expanded)

ArgumentHintDepth Is this completion-item an argument-hint? The model should return an integral positive number if the item is an argument-hint, and QVariant() or 0 if it is not one. The returned depth-integer is important for sorting and matching.
Example: "otherFunction(function1(function2("
all functions named function2 should have ArgumentHintDepth 1, all functions found for function1 should have ArgumentHintDepth 2, and all functions named otherFunction should have ArgumentHintDepth 3
Later, a completed item may then be matched with the first argument of function2, the return-type of function2 with the first argument-type of function1, and the return-type of function1 with the argument-type of otherFunction.
If the model returns a positive value on this role for a row, the content will be treated specially:

• It will be shown in a separate argument-hint list
• It will be sorted by Argument-hint depth
• Match-qualities will be illustrated by differently highlighting the matched argument if possible The argument-hint list strings will be built from all source-model, with a little special behavior:
  • Prefix - should be all text of the function-signature up to left of the matched argument of the function
  • Name - should be the type and name of the function's matched argument. This part will be highlighted differently depending on the match-quality
  • Suffix - should be all the text of the function-signature behind the matched argument
    Example: You are matching a function with signature "void test(int param1, int param2)", and you are matching the first argument.
    The model should then return:
    Prefix "void test(" Name: "int param1" Suffix: ", int param2)"
    If you don't use the highlighting, matching, etc. you can also return the columns in the usual way.

BestMatchesCount This will be requested for each item to ask whether it should be included in computing a best-matches list. If you return a valid positive integer n here, the n best matches will be listed at top of the completion-list separately.
This is expensive because all items of the whole completion-list will be tested for their matching-quality, with each of the level 1 argument-hints.
For that reason the end-user should be able to disable this feature.

AccessibilityNext The following three enumeration-values are only used on expanded completion-list items that contain an expanding-widget.
You can use them to allow the user to interact with the widget by keyboard.
AccessibilityNext will be requested on an item if it is expanded, contains an expanding-widget, and the user triggers a special navigation short-cut to go to navigate to the next position within the expanding-widget(if applicable).
Return QVariant(true) if the input was used.

AccessibilityPrevious AccessibilityPrevious will be requested on an item if it is expanded, contains an expanding-widget, and the user triggers a special navigation short-cut to go to navigate to the previous position within the expanding-widget(if applicable).
Return QVariant(true) if the input was used.

AccessibilityAccept AccessibilityAccept will be requested on an item if it is expanded, contains an expanding-widget, and the user triggers a special shortcut to trigger the action associated with the position within the expanding-widget the user has navigated to using AccessibilityNext and AccessibilityPrevious.
This should return QVariant(true) if an action was triggered, else QVariant(false) or QVariant().

GroupRole Using this Role, it is possible to greatly optimize the time needed to process very long completion-lists.
In the completion-list, the items are usually ordered by some properties like argument-hint depth, inheritance-depth and attributes. Kate usually has to query the completion-models for these values for each item in the completion-list in order to extract the argument-hints and correctly sort the completion-list. However, with a very long completion-list, only a very small fraction of the items is actually visible.
By using a tree structure you can give the items in a grouped order to kate, so it does not need to look at each item and query data in order to initially show the completion-list.
This is how it works:

• You create a tree-structure for your items
• Every inner node of the tree defines one attribute value that all sub-nodes have in common.
  • When the inner node is queried for GroupRole, it should return the "ExtraItemDataRoles" that all sub-nodes have in common
  • When the inner node is then queried for that exact role, it should return that value.
  • No other queries will be done to inner nodes.
• Every leaf node stands for an actual item in the completion list.
• The recommended grouping order is: Argument-hint depth, inheritance depth, attributes. This role can also be used to define completely custom groups, bypassing the editors builtin grouping:
  • Return Qt::DisplayRole when GroupRole is requested
  • Return the label text of the group when Qt::DisplayRole
  • Optional: Return an integer sorting-value when InheritanceDepth is requested. This number will be used to determine the order of the groups. The order of the builtin groups is: 1=Best Matches, 100=Local Scope, 200=Public, 300=Protected, 400=Private, 500=Namespace, 600=Global
    You can pick any arbitrary number to position your group relative to these builtin groups.

UnimportantItemRole Return a nonzero value here to enforce sorting the item at the end of the list.

LastExtraItemDataRole

See also ArgumentHintDepth and ExpandingWidget.

enum CodeCompletionModel::HighlightMethod
flags CodeCompletionModel::HighlightMethods