PR: https://github.com/aibasel/downward/pull/161

Merged, thanks.

Thanks! Indeed, looks good to merge.

I handled the comment. I think this is ready to merge then.

Thanks! For the record, I don't think an issue like this needs a grid experiment. I left a comment with the pull request for a possible further simplification.

The experiments ran and look OK to me. Overall coverage drops by 1 and there are 6 more timeouts (5 fewer memouts), but time and memory do not change by much in any of the domains. The worst losses of time score are openstacks-opt11 (-0.0167) and sokoban-opt11 (-0.0125) and the distribution of time score changes does not look systematic to me(31 domains where the scores improve vs. 24 where they decrease). Times of the new code are within 10% of the old code and larger outliers are only in tasks that take less then 0.1 seconds. 


https://ai.dmi.unibas.ch/_visualizer/?xattribute=total_time&yattribute=total_time&properties_file=https%3A%2F%2Fai.dmi.unibas.ch%2F_experiments%2Fai%2Fdownward%2Fissue1094%2Fdata%2Fissue1094-v1-eval%2Fproperties.tar.xz&entries_list=3f69b110c-astar-blind+73fa3a5e1-astar-blind+astar&y_range=%5B0.4819535182606834%2C+11.0%5D&x_range=%5B0.009000000000000001%2C+1100.0%5D&entries_selection_mode=explicit&relative=true

https://ai.dmi.unibas.ch/_experiments/ai/downward/issue1094/data/issue1094-v1-eval/

I set up a pull request for this. While removing traces of ext, I also found left-over references to tree.hh and a part of misc/style/run-clang-tidy.py that used C++ 11 instead of 20. I tried clang-tidy locally after switching to 20 and it worked.

Do we want experiments for this? How about astar(blind()) just to confirm there is no large performance impact?

We currently have one file left in src/search/ext which is a backport of  std::optional<T>. Now that we use C++ 20, it should be easily possible to replace the single remaining use of the backport by the real deal. This would then allow us to get rid of ext and simplify the CMake setup slightly.