ReSharper navigation consists of two main parts:
Context navigation ('Navigate From Here...')
Global navigation ('Go To ...')
The navigation framework is fairly pluggable, so in most cases all you need to do is to simply provide your own components. You can implement the following features:
Context navigation consists of two parts: the searching engine and the user-interaction provider. These two parts are divided into different entities in ReSharper Framework.
Feature providers represent features themselves. That is, there is just one provider per navigation feature regardless of the number of languages and environments where this feature is supposed to work.
Each context navigation provider implements either the
IContextSearchProvider or the
INavigateFromHereProvider interface. The
IContextSearchProvider interface is defined as follows:
The action that the
GetSearchesExecution() method returns is actually the execution of the navigation feature. When this method returns
null, it implies that this feature is not available for this data context. The
INavigationExecutionHost is a host object which can be used for performing certain UI activities, such as showing a drop-down box or the advanced search dialog.
INavigateFromHere interface is fairly similar to
IContextSearchProvider - it has just one method that returns navigation execution and additional data for presenting in the 'Navigate From Here' drop-down menu.
If you want your navigation items to appear in the 'Navigate From Here' menu, you have to implement the
Binding to an action
You don't need to write any code in your action handler's methods, as all feature logic needs to be written inside navigation providers. All you need to bind your action to a navigation provider is to derive your custom action handler class from
ContextNavigationActionBase<TNavigateFromHereProvider> (if you want your action to appear in the 'Navigate From Here' menu) or
Additionally, your provider should be decorated with the
If your navigation is not complicated, that's all you need to know about the context navigation framework.
Here is an example of simple navigation to the corresponding folder using Windows explorer.
If your navigation is complicated and works differently in different languages then you can use context searches to provide different search results for a specific language.
In this case, your provider needs to be derived from
ContextSearchesCollector<TContextSearch> class and all you need is to implement the
IContextSearch is an interface for providers decorated with the
FeaturePart attribute and providing specific navigation execution. There are just two methods in this interface, indicating applicability and availability of said context search.
The applicability method is more global and implies that this context search can work for this data context and will override other components it is derived from.
Availability means that for a specific data context this context search is enabled and will be fired.
You can alter the behavior of existing ReSharper navigation features by providing your own context search components. Here is an example of a context search that extends the 'Go To Implementation' search for XAML:
Global Navigation ('Go To ...')
Global nagivation relates to functionality used in various global searches such as Go To Type or Go To Member. Typically, these searches all function as a text-entry popup box that, when typed into, presents a list that can be further refined and then, upon selection, navigated to the declared element. In addition, lists can ‘morph’ so that a Go To Type search can suddenly become a Go To Symbol search with the press of one shortcut.
Occurrence Navigation Provider
In order to provide items for a particular global navigation mechanic, we need to implement an occurrence navigation provider - a class that implements the
IOccurrenceNavigationProvider interface. The following image illustrates the hierarchy of this interface:
As you can see, the four interface members are self-descriptive, and are further implemented by concrete classes based on language. For example,
ClrGotoTypeMemberProvider, which illustrates a kind of inheritance mechanic - this makes sense because Go To Symbol actually includes information from Go To Type.
The interface is fairly simple, and has only three members:
The first method,
FindMatchingInfos(), find matched items and returns a corresponding list of
MatchingInfo objects. A
MatchingInfo is simply a data class that describes the necessary info used to store a matching item - its identifier, a set of
IdentifierMatch objects (corresponding to the matches, as there can be many) and some other service information.
The second method,
GetOccurrencesByMatchingInfo, returns a set of occurrences given the matching information. An occurrence is essentially a pointer to the location where something was found. Since the definition of ‘something’ is quite vague, an occurrence can be practically anything - a range in a text editor, a project reference, a while project or a file. For example, a project file is represented by a
ProjectItemOccurrence, which in turn references an
IProjectItem it points to.
In addition, you are required to implement an interface member inherited from
The above method lets us determine if a particular provider should work in a given context. Providers can theoretically inherit from one another, and if you are overriding an existing provider, instead of working every single time (like a
SolutionComponent would), it instead checks the
The API now provides you an interface called
IGotoEverythingProvider, which happens to inherit these two interfaces and in addition specifies a sorting function declared as
This interface lets you calculate a priority of the displayed element, such that the lower the result gets returned, the higher is the priority of the element.
Now, to create a provider, simply implement the interface and decorate the class with
Having a set of occurrences is great, but they’re useless until presented on-screen. And this is where occurrence presenters come in. An occurrence presenter is, basically, a class that knows how to present a particular occurrence as a menu item. In addition to being decorated by an
OccurrencePresenter attribute, it is embodied by the
IOccurrencePresenter interface defined below:
First, there’s the
IsApplicable() method. This method determines whether this occurrence presenter is applicable for the particular type of occurrence. These typically go hand-in-hand, for example:
Then there’s the
Present() method, which is called when the
IOccurrence needs to be presented. The goal of this method is to make assignments to the
descriptor parameter. The
IMenuItemDescriptor interface is quite large, but the only property we need to concern ourselves with here is
Text - this is a
RichText definition that will be displayed as a menu item for this occurrence.
Here's a sample implementation. Note that
descriptor.Style must be defined - otherwise, the item will be disabled.
Occurrence Kind Provider
When you open up, say, Find Usages, you’ll find that you have an ability to filter information. Some examples are read usages, write usages, attribute references, and so on. Each has its own special icon and an ability to filter it out if necessary.
The types of elements just mentioned are called occurrence kinds. Many kinds of occurrences can be found as static members of the
OccurrenceKind class, and new ones can be created. An occurrence kind is simply an enum-like class, keeping only two bits of information: its name and whether it is primary or not. If the occurrence kind is set as primary, the Find Results window will always display this kind of occurrence regardless of what filter is set. Other occurrences are non-primary, i.e. they are only displayed when there are occurrences of this occurrence kind (e.g., unit tests).
Now, we can talk about occurrence type providers. These are typically solution components (i.e., decorated with the
SolutionComponent attributes) that also implement the
IOccurrenceKindProvider interface. This interface is defined as follows:
GetAllPossibleOccurrenceKinds() method returns a list of all possible occurrences that this provider can give in theory. For example, in C# we can yield interface qualifications, base method calls and invocations, whereas in VB we also have the
GetOccurrenceKinds() method returns a collection of occurrence kinds (or you can return an
EmptyList<OccurrenceKind>.InstanceList if there aren’t any) that are applicable for a given occurrence. Naturally, this typically assumes that you would try to cast the
IOccurrence to a concrete type (e.g.,
ReferenceOccurrence) to investigate its contents and yield the occurrence kinds accordingly.
Occurrence Section Provider
When search results are presented in a dialogue window, they are typically split into sections via horizontal lines. The component that provides information about the different types of section is a
FeaturePart that implements the
IOccurrenceSectionProvider interface. The interface is defined as follows:
Let’s go through the methods. Firstly,
IsApplicable() checks whether this section provider is applicable given the type of the occurrence browser we are working with. This is typically a simple
is call, e.g.:
The second method is
GetTreeSections(), and it is this method that returns a set of
TreeSection objects specific to this section provider. We’ll discuss the
TreeSection structure in a moment, but let’s briefly discuss the third method first.
GetGroupSectionId basically lets us determine the groups (there can be more than one) in which a particular
IOccurrence needs to appear. The default implementation as defined in the
OccurrenceSectionProvider class (you should inherit from this, rather than implementing the interface by hand) is to use a special method
GetSectionId() to pick a known section given the type of the occurrence:
Now, let’s go back and talk about
TreeSection definitions or, rather, about tree model, since the
TreeSection is nothing more than a container class for a title and a
TreeModel and it is this tree model that’s of interest to us.
TreeModel, as you may have guessed, is simply a definition for the classical N-ary tree. The model itself has itself a collection of roots, i.e. a property
Roots of type
IList<TreeModelNode> as well as various mechanisms for insertion, removal and updating of
TreeModelNode in turn has a set of children, who are also
TreeModelNode s themselves. Sorting is possible at both the
TreeModelNode level. Note that the actual data storage inside the nodes is defined as an
object, i.e. is untyped.
It helps to remember that both
TreeModelNode are abstract classes, and that you are more likely to be working with their descendants, the most basic of which is
The menu items themselves are simply items which implement the
IOccurrence interface. This interface defines a rather large number of elements:
This interface is very feature-complete in terms of supporting, e.g., constructs which exist in code. In the case where you are providing ad hoc items not related to the PSI, you can provide the following defaults instead:
DumpToString()- up to you, this is a debugging method.
TextRange- just return a
new TextRange(). You might want to consider making a single instance to return for each query.
MergeKey- this yields a key which indicates whether or not several items should be merged. Thus, it makes sense to have this property yield a unique value for each item you don't intend to merge.
MergedItems- simply return
PresentationOption- define it as a
For examples of leveraging the various
IOccurrence members, take a look at some of the its built-in ReSharper implementors.