TY: convert primitive types from object to class #7503
Conversation
Kobzol
added
the
internal
There was a problem hiding this comment.
LGTM.
The one thing I'd made additionally is turning name and ordinal fields in TyInteger and TyFloat classes into abstract properties with getters overridden in descendant classes.
Which will probably introduce all kinds of scary breakage and problems (especially the description of this PR #4243 scares me in relation to comparing types
Well, == just shouldn't be used for Tys. But there are already quite lot of such usages and I have no idea how to find them
I didn't change TyNever and TyUnknown, because we probably don't need aliases to work for them (?).
Most likely yes.
Kobzol
force-pushed
the
ty-reduce-singleton-usage
branch
from
a48ee79
to
cc28619
Compare
|
Good idea, I changed them. EDIT: Oh, I thought that |
Kobzol
force-pushed
the
ty-reduce-singleton-usage
branch
from
cc28619
to
08205f5
Compare
|
Now all primitive types implement these properties with getters and there should be no memory wasted on backing fields. |
|
bors r=vlad20012 |
7503: TY: convert primitive types from object to class r=vlad20012 a=Kobzol This is a naive attempt to reduce the usage of `object` in primitive types and convert them to standard classes instead. This is probably needed as a first step to solve #5080. I didn't change `TyNever` and `TyUnknown`, because we probably don't need aliases to work for them (?). The ultimate plan is to move `aliasedBy` from `TyAdt` to `Ty`. Which will probably introduce all kinds of scary breakage and problems (especially the description of this PR #4243 scares me in relation to comparing types :D ). Sending PR to see what kinds of tests fail. Co-authored-by: Jakub Beránek <berykubik@gmail.com>
|
Build failed: |
|
bors retry |
7503: TY: convert primitive types from object to class r=vlad20012 a=Kobzol This is a naive attempt to reduce the usage of `object` in primitive types and convert them to standard classes instead. This is probably needed as a first step to solve #5080. I didn't change `TyNever` and `TyUnknown`, because we probably don't need aliases to work for them (?). The ultimate plan is to move `aliasedBy` from `TyAdt` to `Ty`. Which will probably introduce all kinds of scary breakage and problems (especially the description of this PR #4243 scares me in relation to comparing types :D ). Sending PR to see what kinds of tests fail. Co-authored-by: Jakub Beránek <berykubik@gmail.com>
|
Build failed: |
|
bors retry |
7503: TY: convert primitive types from object to class r=vlad20012 a=Kobzol This is a naive attempt to reduce the usage of `object` in primitive types and convert them to standard classes instead. This is probably needed as a first step to solve #5080. I didn't change `TyNever` and `TyUnknown`, because we probably don't need aliases to work for them (?). The ultimate plan is to move `aliasedBy` from `TyAdt` to `Ty`. Which will probably introduce all kinds of scary breakage and problems (especially the description of this PR #4243 scares me in relation to comparing types :D ). Sending PR to see what kinds of tests fail. Co-authored-by: Jakub Beránek <berykubik@gmail.com>
|
Build failed: |
|
bors retry |
|
Merge conflict. |
Kobzol
force-pushed
the
ty-reduce-singleton-usage
branch
from
08205f5
to
92bc5cf
Compare
|
bors r=vlad20012 |
|
Build succeeded: |
This is a naive attempt to reduce the usage of
objectin primitive types and convert them to standard classes instead. This is probably needed as a first step to solve #5080. I didn't changeTyNeverandTyUnknown, because we probably don't need aliases to work for them (?).The ultimate plan is to move
aliasedByfromTyAdttoTy. Which will probably introduce all kinds of scary breakage and problems (especially the description of this PR #4243 scares me in relation to comparing types :D ).Sending PR to see what kinds of tests fail.