perf: replace LinkedList and Optionals with ArrayDeque and EnumMap

[Photon-GitHub]

This pull request mainly does three things:

1. It swaps out the LinkedList implementation with an ArrayDeque. In most cases (i.e. adding and querying) this will improve performance as we only write or read an array. Moreover this allows to directly init with the default capacity which makes sure the array does not need to be grown. This is faster than the LinkedList which wraps every new entry.
   Only the remove and move calls might be slower, but even in the worst case, we only have 2 arraycopy calls on a size of default_capacity, which should be negligient. Also, I think that these operations are less often used compared to inserting elements and querying, but that is dependent on the use case.

2. Replace the Optional list of children with an EnumMap. There were a lot of list.get(c.ordinal()).get() calls in the code, which goes against the reason Optional is used (as the isPresent check is missing). Furthermore, the list init is quite hard to understand right now, because the Child enum is not linked to list size (4 is a magic value appearing there) and we always use c.ordinal() as the list index. For this, an EnumMap is much clearer. First of all, it directly reflects the Enum, makes the connection clear and does not need the verbose init of the current list. Removing the optional part was easy, and due to the already often missing isPresent checks I only needed to add one single null (in setChild()) in the entire codebase, and even that would be easy to remove.

3. Thanks to the removal of the children absent optionals some methods have become much smaller and could be removed, like hasChildren() which is now equal to !children.isEmpty() or getChild(...) which is now children.get()

[codecov]

Codecov Report

Merging #166 (15054cf) into master (5007123) will decrease coverage by 0.91%.
The diff coverage is 76.47%.

@@             Coverage Diff              @@
##             master     #166      +/-   ##
============================================
- Coverage     76.03%   75.11%   -0.92%     
+ Complexity       79       73       -6     
============================================
  Files             1        1              
  Lines           217      209       -8     
  Branches         38       37       -1     
============================================
- Hits            165      157       -8     
  Misses           25       25              
  Partials         27       27              

Impacted Files	Coverage Δ
.../java/org/danilopianini/util/FlexibleQuadTree.java	75.11% <76.47%> (-0.92%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5007123...15054cf. Read the comment docs.

[DanySK]

Thank you for contributing.

Looks good. I'm going to squash this in as soon as the tests pass. Since the main motivation for change seems to be performance, I modified the PR type. Also, this should cause the generation of a new release.

[github-actions]

🎉 This PR is included in version 0.1.8 🎉

The release is available on:

• 0.1.8
• GitHub release

Your semantic-release bot 📦🚀