Implement FreshCap Handling for Classes and Objects #24136
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
The prefix is mapped as a normal type. It determines whether the hidden set allows to add new elements. ThisType and NoPrefix prefixes allow it, other prefixes forbid it. The pathRoot and ccOwner of a FreshCap now depend on the prefix.
Also: add another test that demonstrates some behavior relating to default parameters. Also: Rename a test so that we don't accidentally get the test file. In general, it's a bad idea to use a source file name in the compiler code for a test since we often get the wrong file when loading it into the editor.
Was cap.rd before, which was a left-over of the old model.
Refine the isField test in SymUtils to exclude non-members and phantom symbols
odersky force-pushed the add-fresh-prefix branch from 47fc1f9 to dfcca31 Compare A fresh in the capture set of a class field now causes a fresh to be added to the capture set of every instance of that class.
odersky force-pushed the add-fresh-prefix branch from dfcca31 to d2abdfe Compare - Drop the assertion and handle the case of empty overlap - At the same time, avoid the empty overlap by taking complete instead of direct footprints. Also: Improve printing of FreshCap prefixes
and a fresh for one of its fields.
Needed for use checking, since if we just take the leftmost prefix we sometimes end up with a `this`.
odersky force-pushed the add-fresh-prefix branch from 7e813cd to a0f1613 Compare odersky changed the title Private fields in publicly accessible classes that contribute fresh caps to their class still need an explicitly declared type, so that separate compilation can work.
odersky added area:experimental:cc odersky force-pushed the add-fresh-prefix branch from 9848c89 to 6689785 Compare 
| Similar to #24137 (comment) I could patch the stdlib in the same way to completely avoid untrackedCaptures: natsukagami@f638614 I made a PR: dotty-staging#73 |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
This is a complete implementation of https://github.com/lampepfl/papers/pull/155. It consists of the following major parts:
Let fresh caps in field types contribute to class captures
A fresh in the capture set of a class field now causes a fresh
to be added to the capture set of every instance of that class.
This caused 18 failures in stdlib of which 2 were significant (the rest was deprecated stuff). So far, we make
this compile with the
@caps.unsafe.untrackedCapturesannotation. #24137 explains what would be neededto fix this in a safer way.
Add prefixes to fresh caps and relate class and field fresh caps
Fresh created in classes now carry a prefix referring to the
thisof the class. The prefix gets mappedby TypeMaps, including
asSeenFrom. We establish a "covers" relation between a fresh for an object and a fresh for one of its fields. We use the same relation for subsumption.Example:
Here the type of
a.fisA^{a.cap2}. Furthermorecap1, the cap captured by the type ofa, both covers and subsumesa.cap2.Make sure that private fields with inferred types don't capture a fresh cap
Fields contribute fresh caps to the class. The problem is what to do with capsets in inferred types of fields. These pose problems of separate compilation. We already demand explicit declarations of capsets of non-private fields. We need to also demand an explicit declaration when a private field has an inferred capset that is not otherwise accounted for in the capset of the enclosing class. That condition can be checked post-cc, when all capsets are known. I.e. it is checked at the same time as when we check non-private fields.