#2417 Explore reimplementing HashMap as CHAMP (Compressed Hash-Array Mapped Prefix-tree)

[wrandelshofer]

This is a draft intended for discussion.

This merge requests explores how a migration of the HAMT trie to a CHAMP trie could look like:

• HashSet, HashMap: Replaces the internal HAMT data structure by a CHAMP data structure.
• LinkedHashSet, LinkedHashMap: Replaces the internal HAMT+Seq data strucures by CHAMP+Vector data structures.
• The changes are intended to affect the performance of vavr, but are not supposed to change any API.

The changes are somewhat sub-optimal and somewhat over-complicated, because I strived to not break any of the unit tests. (I believe that I should change a number of unit tests, because some rely on the iteration sequence of the HAMT implementation. However, the vavr API specifies that the iteration sequence is unspecified.)

The code is a port from the JHotDraw 8 collections module. https://github.com/wrandelshofer/jhotdraw8/tree/main/org.jhotdraw8.collection
The JHotDraw 8 collections are in turn a port of the 'capsule' collection library https://github.com/usethesource/capsule, and of vavr.

I do not expect that this draft can be merged in the current form. In JHotDraw 8 we have a mutable 'partner' collection for every immutable collection. This allows to test all immutable collections with the 'guava-testlib' library. Which provides very through tests. I did not want to change the API of vavr - so I have downgraded the mutable collections to transient collections. But I have not written new tests yet. So code coverage by unit tests is currently lower than I want it to be.

[wrandelshofer]

I made now a release that is binary compatible with vavr 0.10.5.

You can get it here: https://github.com/wrandelshofer/vavr/releases/tag/v0.10.5

[codecov]

Codecov Report

Attention: Patch coverage is 70.56169% with 587 lines in your changes missing coverage. Please review.

Project coverage is 89.24%. Comparing base (e43e44b) to head (734fc83).
Report is 73 commits behind head on master.

Files with missing lines	Patch %	Lines
src/main/java/io/vavr/collection/ChampTrie.java	50.00%	252 Missing and 63 partials ⚠️
...c/main/java/io/vavr/collection/ChampSequenced.java	49.74%	89 Missing and 8 partials ⚠️
...rc/main/java/io/vavr/collection/LinkedHashSet.java	81.38%	35 Missing and 8 partials ⚠️
src/main/java/io/vavr/collection/HashSet.java	76.78%	32 Missing and 7 partials ⚠️
...rc/main/java/io/vavr/collection/LinkedHashMap.java	89.11%	24 Missing and 13 partials ⚠️
.../main/java/io/vavr/collection/ChampTransience.java	68.11%	14 Missing and 8 partials ⚠️
src/main/java/io/vavr/collection/HashMap.java	90.59%	6 Missing and 13 partials ⚠️
...c/main/java/io/vavr/collection/ChampIteration.java	88.23%	5 Missing and 3 partials ⚠️
...rc/main/java/io/vavr/collection/BitMappedTrie.java	86.66%	4 Missing ⚠️
src/main/java/io/vavr/collection/Vector.java	95.16%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #2745      +/-   ##
============================================
- Coverage     92.74%   89.24%   -3.50%     
- Complexity     5247     5294      +47     
============================================
  Files            89       92       +3     
  Lines         12538    13844    +1306     
  Branches       1604     1857     +253     
============================================
+ Hits          11628    12355     +727     
- Misses          721     1187     +466     
- Partials        189      302     +113     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.