Eliminate newNakedSymbol methods
#11122
Conversation
odersky
force-pushed
the
refactor-symbols
branch
from
a839248
to
e1cc241
Compare
|
test performance please |
|
performance test scheduled: 29 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/11122/ to see the changes. Benchmarks is based on merging with master (ba70cc0) |
odersky
force-pushed
the
refactor-symbols
branch
from
93775dd
to
653d8b4
Compare
|
test performance please |
|
performance test scheduled: 26 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/11122/ to see the changes. Benchmarks is based on merging with master (65b17af) |
|
@liufengyun Can it be that the performance test does not come back for three days? |
|
@odersky There is something abnormal, I'm looking into it. |
|
I restarted the job for this PR yesterday, it's running since then. However, we have a significant slowdown of the benchmarks, ~5h per PR job. It might be related to the changes in the file system or related to the jmh update in #10591. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/11122/ to see the changes. Benchmarks is based on merging with master (26398f4) |
These methods produce a partially initialized symbol, which is dangerous for any code using them. It turns out that any delays in the symbol initialization we need can be provided by the symbol's completer.
Last step before we can make Symbols inherit from Denotations
Use SymbolImpl instead of Symbol in all Denotation class parameters. This is necessary to avoid cyclic references when Symbols are made subclasses of SymDenotations.
We need to go through denot when going from a Symbol to a SymDenotation. Hence, the fact that Symbol will extend SymDenotations is not allowed to be known even in Symbols.
This is to avoid polymorphic dispatch on name once Symbols are subtypes of SymDenotations
odersky
force-pushed
the
refactor-symbols
branch
from
653d8b4
to
6777965
Compare
|
test performance please |
|
performance test scheduled: 43 job(s) in queue, 1 running. |
|
Performance test finished successfully: Visit http://dotty-bench.epfl.ch/11122/ to see the changes. Benchmarks is based on merging with master (b731573) |
|
@odersky what's the status on this one? Are you planning to continue working on it or can it be closed? |
|
The first commit is now in #13641. The other refactorings complicate things and don't seem to improve performance. So that might be a dead end, or a blocker until we have better ideas. |
These methods produce a partially initialized symbol, which is dangerous for any code
using them. It turns out that any delays in the symbol initialization we need can be
provided by the symbol's completer.