Support for locating the usages of the name at the specified module coordinates. Mandated by Find References Request.
This amounts to implementing org.eclipse.lsp4j.TextDocumentService.references.
For guidance, the places to begin digging into Swing/Anvil are:
avail.anvil.AvailEditor.handleClick
avail.anvil.AvailWorkbench.navigateToSendersOfName
avail.anvil.AvailWorkbench.allUsagesThen
avail.anvil.RenderingEngine.applyStylesAndPhrasePaths
Currently, the technique is:
- Resolve the mouse position to a document position
- Use the document position to obtain the
TokenStyle associated with the PhraseNodeAttributeKey attribute that was applied by avail.anvil.RenderingEngine.applyStylesAndPhrasePaths
- Use the
TokenStyle to obtain the name / symbol (NameInModule) and token index / message part index
- Use these to interrogate a
Repository and a NamesIndex to obtain all usages associated with the name / symbol
The message part index (usually called tokenIndexInName) is currently used only to highlight the message part that the user specified. It does not appear that LSP (or lsp4j) are capable of transmitting this additional information as opaque client data. If we want to preserve the current behavior, then we will need to reply with LocationLinks rather than mere Locations, and use the MessageSplitter (or something akin but with pared down functionality) in the IDEA plugin to identify and highlight the message part. The current behavior was nice enough to build in Swing/Anvil in the first place, but is likely to require significant additional effort to port to IDEA, so we may want to consider strict parity as a stretch goal.
The following items count as usages:
- Lexical sends appearing directly in source code
- Generated sends appearing in macro substitution phrases (produced by executing macro sends during compilation)
Support for locating the usages of the name at the specified module coordinates. Mandated by Find References Request.
This amounts to implementing
org.eclipse.lsp4j.TextDocumentService.references.For guidance, the places to begin digging into Swing/Anvil are:
avail.anvil.AvailEditor.handleClickavail.anvil.AvailWorkbench.navigateToSendersOfNameavail.anvil.AvailWorkbench.allUsagesThenavail.anvil.RenderingEngine.applyStylesAndPhrasePathsCurrently, the technique is:
TokenStyleassociated with thePhraseNodeAttributeKeyattribute that was applied byavail.anvil.RenderingEngine.applyStylesAndPhrasePathsTokenStyleto obtain the name / symbol (NameInModule) and token index / message part indexRepositoryand aNamesIndexto obtain all usages associated with the name / symbolThe message part index (usually called
tokenIndexInName) is currently used only to highlight the message part that the user specified. It does not appear that LSP (orlsp4j) are capable of transmitting this additional information as opaque client data. If we want to preserve the current behavior, then we will need to reply withLocationLinks rather than mereLocations, and use theMessageSplitter(or something akin but with pared down functionality) in the IDEA plugin to identify and highlight the message part. The current behavior was nice enough to build in Swing/Anvil in the first place, but is likely to require significant additional effort to port to IDEA, so we may want to consider strict parity as a stretch goal.The following items count as usages: