Deprecation fixes and other updates #54
Conversation
| @@ -118,8 +118,6 @@ function sortedcomplement(of::SourceRange, set) | |||
| end | |||
|
|
|||
| ⨳(sym::Symbol) = Expr(sym) | |||
| ⨳(sym::Symbol, args::LineNumberNode...) = Expr(sym::Symbol, args...) | |||
There was a problem hiding this comment.
@KristofferC these may not be exercised in the tests but they may be used in the ASTInterpreter.jl source code, have you checked this?
There was a problem hiding this comment.
It feels impossible to guard against breaking other packages that rely on unexported and untested methods but I can just add these back, no problems.
There was a problem hiding this comment.
Ok, they are put back now.
|
Nice! Glad to see that even with tracking the source information performance is still competitive. |
|
Do you think the type instability of Token has a major performance impact? One way of defining a Token could be with just a string field and a "Kind" field that says what type of Token it is. Kind would typically just be an Enum listing all the different Token types that exist. |
|
Hmm, I wouldn't want to speculate on the performance diff before trying it first but seems worth trying and shouldn't be too hard. This LGTM. |
Before:
Parser is ~64.83% slower than baseAfter:
Parser is ~13.38% slower than baseParses correctly the same amount of files in Base as before and passes the same amount of tests.
Fixes #52