This module provides the facilities for matching code offsets to CFG nodes.
As it happens, this is not a straight-forward, 1-to-1 mapping, and requires some care.
Generally, AST nodes are converted into CFG nodes, and this mapping is relatively straightforward.
However, a lof of the code does not belong to any node in particular,
and some mappings need to be changed to make the matching more intuitive.
In the following explanation, we'll use "syntax" to refer to AST nodes,
and "nodes" to refer to CFG nodes. This matches the code for the most part.
To deal with that, we use a multi-level mapping system:
Syntax is mapped to nodes
Offsets are mapped to syntax.
This allows us to separate some concerns, and makes the resulting code a bit easier to read.
As the CFG is generated, we add the relevant mappings to a collection.
Once CFG creation is complete, we sort the ranges from large to small,
and construct a SimpleRange[] mapping for them.
Note that during construction, no smaller node may overlap more than a single larger node.
This matches the hierarchical nature of the AST, and makes our life easier as there are
no conflicts to resolve.
Once the ranges are created, they can be used to query the node matching a specific offset.
In some cases, there is another step (not implemented here).
If the graph is simplified, and nodes are collapsed together, the mappings will change.
To manage that, nodes contain a targets field that holds their name.
When two nodes are collaped, the new targets fiels will contain both their targets.
Generally, we'll remap the nodes using the per-node targets after simplification.
This module provides the facilities for matching code offsets to CFG nodes.
As it happens, this is not a straight-forward, 1-to-1 mapping, and requires some care. Generally, AST nodes are converted into CFG nodes, and this mapping is relatively straightforward. However, a lof of the code does not belong to any node in particular, and some mappings need to be changed to make the matching more intuitive.
In the following explanation, we'll use "syntax" to refer to AST nodes, and "nodes" to refer to CFG nodes. This matches the code for the most part.
To deal with that, we use a multi-level mapping system:
This allows us to separate some concerns, and makes the resulting code a bit easier to read.
As the CFG is generated, we add the relevant mappings to a collection. Once CFG creation is complete, we sort the ranges from large to small, and construct a
SimpleRange[]mapping for them. Note that during construction, no smaller node may overlap more than a single larger node. This matches the hierarchical nature of the AST, and makes our life easier as there are no conflicts to resolve.Once the ranges are created, they can be used to query the node matching a specific offset.
In some cases, there is another step (not implemented here). If the graph is simplified, and nodes are collapsed together, the mappings will change. To manage that, nodes contain a
targetsfield that holds their name. When two nodes are collaped, the newtargetsfiels will contain both their targets. Generally, we'll remap the nodes using the per-node targets after simplification.