Add Doxygen documentation for ResolveReferences #613
Conversation
| @@ -252,9 +251,8 @@ void ResolveReferences::end_apply(const IR::Node* node) { | |||
| refMap->updateMap(node); | |||
| } | |||
|
|
|||
| /////////////////// visitor methods //////////////////////// | |||
| /****** Visitor methods ******/ | |||
There was a problem hiding this comment.
this should not be a doxygen comment.
| class ReferenceMap final : public ProgramMap, public NameGenerator { | ||
| bool isv1; // if true this is a map for a P4 v1.0 program (P4-14) | ||
| // Maps each path in the program to the corresponding declaration | ||
| /// If `isv1` is true, then map is for a P4_14 program |
There was a problem hiding this comment.
Maps for v1 programs are different because v1 programs allow declarations to come after uses.
This difference is carried even in P4-16 program that are generated from P4-14 programs.
| std::set<const IR::IDeclaration*> used; | ||
|
|
||
| /// Map from `This` to declarations | ||
| /// @todo: briefly explain why different than `pathToDeclaration` |
There was a problem hiding this comment.
Because Path is a different class than This.
This is an experimental feature.
There was a problem hiding this comment.
Thanks. Added a note.
| @@ -440,8 +438,10 @@ void ResolveReferences::postorder(const IR::Declaration_Instance *decl) { | |||
| } | |||
|
|
|||
| bool ResolveReferences::preorder(const IR::Property *prop) { | |||
| /// @todo: why is `dynamic_cast` used here instead of `to`? | |||
There was a problem hiding this comment.
@ChrisDodd wrote this line, there are really no Attributes in P4-16.
There was a problem hiding this comment.
Let's leave the TODO for now.
| enum class ResolutionType { | ||
| Any, | ||
| Type, | ||
| TypeVariable | ||
| }; | ||
|
|
||
| /** | ||
| * Data structure used while resolving references that comprises: | ||
| * - A stack of namespaces |
There was a problem hiding this comment.
I think that the description of the fields belongs to the fields, not to the class.
This class is logically a stack of nested namespaces where the resolution of a symbol is attempted, inside-out.
| bool anyOrder; | ||
|
|
||
| /// Indicates if it should be an error for one declaration to shadow another. |
There was a problem hiding this comment.
Actually it indicates whether warnings should be emitted for shadowing.
It is never an error.
| bool checkShadow; | ||
|
|
||
| private: | ||
| /// Add @p ns to `context` |
There was a problem hiding this comment.
These next three comments are not very useful.
There was a problem hiding this comment.
True, but worth having for uniformity IMO.
| void addToGlobals(const IR::INamespace* ns); | ||
|
|
||
| /// Resolve @p path; @p isType indicates if a type is expected. |
There was a problem hiding this comment.
If 'isType' is true then the resolution will only return IR::Type nodes.
Also, refactor so that all booleans used to control forward resolution are oriented in the same direction.
Also, refactor so that all booleans used to control forward resolution
are oriented in the same direction.