new macro @gapattribute #647
Merged
Conversation
This file contains 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
Codecov Report
@@ Coverage Diff @@
## master #647 +/- ##
==========================================
+ Coverage 77.15% 77.28% +0.12%
==========================================
Files 47 47
Lines 4102 4129 +27
==========================================
+ Hits 3165 3191 +26
- Misses 937 938 +1
|
Now the macro works, without segmentation faults.
|
With the second attempt (after a discussion with @fingolfin), the segmentation faults are avoided. |
ThomasBreuer
changed the title
|
Note that with JuliaLang/julia#40852 at least there will be a somewhat helpful (?) error message instead of a crash when using GAP objects on the top level of package modules (where the precompilation code ends up trying to serialize them). |
fingolfin
pushed a commit
to oscar-system/Oscar.jl
that referenced
this pull request
Jul 30, 2021
Use the macros `@gapattribute` and `@gapwrap` in order to get access to GAP's "tester" and "setter" functions for its attributes. This had been suggested in issue #128. The code requires the changes from the pull request oscar-system/GAP.jl/pull/647. The branch has the same name as that of the current pull request. (There are more places where the macros can be applied. First we should agree on the setup, see below.) - The names of the tester/setter functions on the Julia side should perhaps be changed: Would `has_derived_subgroup` look better than `hasderived_subgroup`? (This is something to be changed in GAP.jl.) - Some of the functions in `Oscar/src/Groups` (`conjugacy_classes`, `conjugacy_classes_subgroups`, `conjugacy_classes_maximal_subgroups`) cannot be handled by the current `@gapattribute`, because they involve several calls to some function in `GAP.Globals`. The point is that the names of the tester and setter are "guessed" from the body of the given (getter) function. If there is just one call to a function in `GAP.Globals` then this is safe, otherwise we have to find some heuristic that is easy to document. - In a few cases (`exponent`), I have added individual `import` statements because the current `@gapattribute` does not strip module names from the beginning of the function name; we could change the macro in order to avoid these `import` statements. - In order to keep the docstrings of the functions handled by `@gapattribute`, I did not see a better way than to move the code above the docstring. - Another point for the documentation is how to document the testers and setters. We could show three-line descriptions of the kind ``` isperfect(G::Group) hasisperfect(G::Group) setisperfect(G::Group, val) ``` in all relevant places, instead of just the first of these lines. Or would this become too long? (I have extended the `export` lists by the testers and setters. Is this o.k.?) - I have changed the behaviour of the functions only in one place. Up to now, `trivial_subgroup` constructed just a group on the Julia side, without calling GAP's `GAP.Globals.TrivialSubgroup`, but now the attribute on the GAP side is called, due to `@gapattribute`. The side-effect is that now the return value of `trivial_subgroup` has an empty list of generators, whereas this list contained the identity element before. This change seems to be no problem for the tests, but perhaps it has to be documented that the result of `gens` may be empty. The conceptual question is whether we really want to change `trivial_subgroup` such that the corresponding GAP function gets called. We can take the viewpoint that the Julia side should better avoid work on the GAP side where possible (one can create a trivial subgroup in Oscar without "help from GAP"), or we can argue that the GAP side should better be kept informed about knowledge that has been computed on the Julia side. Note that in the cases of boolean values, the stored GAP attributes really serve as caches; but already for a group order stored in GAP, a conversion may be necessary before the value can be used on the Julia side. One "border case" is `ispgroup`: Its return value is a tuple not just the boolean value that is stored on the GAP side. Any access from Oscar's `ispgroup` creates a new Julia object. - The following two inconsistencies should be dealt with in other pull requests: `center` and `ischaracteristic` are known in (and imported from) Hecke, but Oscar.Groups defines `centre` and `ischaracteristic_subgroup`.
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 intended to address oscar-system/Oscar.jl/issues/128.
but it does not work as expectedApparently something is wrong with the code: When I try to call the macro inside an Oscar source file then Julia crashes with a segmentation fault.However, the following works.Here is an example how to use the new macro.
testmodule.jlwith a Julia module that calls the macro in the way we want to use it.