RFC: Completely remove the option of a zero-sized block cache

[adamreichold]

This was discussed in #1755 (comment) and would simplify the cache implementation at the cost of some scattered changes around the code base and the requirement to at least use cache size one everywhere.

[codecov-commenter]

Codecov Report

Merging #1770 (1d0fbf4) into main (7a8fce0) will decrease coverage by 0.02%.
The diff coverage is 85.71%.

@@            Coverage Diff             @@
##             main    #1770      +/-   ##
==========================================
- Coverage   94.12%   94.09%   -0.03%     
==========================================
  Files         282      282              
  Lines       52908    52910       +2     
==========================================
- Hits        49798    49788      -10     
- Misses       3110     3122      +12     

Impacted Files	Coverage Δ
src/functional_test.rs	0.00% <0.00%> (ø)
src/reader/mod.rs	91.25% <33.33%> (-1.75%)	⬇️
src/core/searcher.rs	78.37% <100.00%> (ø)
src/core/segment_reader.rs	90.80% <100.00%> (+0.03%)	⬆️
src/indexer/merger.rs	98.84% <100.00%> (ø)
src/indexer/segment_writer.rs	97.71% <100.00%> (+0.01%)	⬆️
src/store/mod.rs	99.23% <100.00%> (+<0.01%)	⬆️
src/store/reader.rs	79.91% <100.00%> (-0.71%)	⬇️
sstable/src/block_reader.rs	73.77% <0.00%> (-1.64%)	⬇️
src/indexer/segment_updater.rs	94.40% <0.00%> (-1.05%)	⬇️
... and 3 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[PSeitz]

I think the NonZero API is worse than before. Disabling the cache is a valid use case (e.g. in very resource constrained environments or for tests). I created a ticket here: jeromefroe/lru-rs#165

[adamreichold]

I think the NonZero API is worse than before. Disabling the cache is a valid use case (e.g. in very resource constrained environments or for tests). I created a ticket here: jeromefroe/lru-rs#165

Yes, I agree. Let's drop this and remove the Option layer when lru is bumped to something supporting zero-capacity caches again.