Log

• [maven-release-plugin] prepare release 0.2.0 by Caio 7 years ago
• And make LadelData.children() a Map<String, Long> 💬 by Caio 7 years ago

  One less useless abstraction and the result API is drastically
  easier to use, just look at the tests.
  

• Change SearchResult.facets() from List to Map by Caio 7 years ago
• Remove Diet-related strictness 💬 by Caio 7 years ago

  I might regret this later, but right now it doesn't seem like I
  need to care about diet names here. The only reason I can think
  of now is to avoid indexing rubbish like a bunch of invalid
  diet names because for some reason our source data started
  emitting such.
  
  I'll take the opportunity to drop this awkward class. If I end
  up needing some strictness here, I'll do it via the client side
  as well, like it was done for taxonomy facets.
  

• Remove forgotten println() by Caio 7 years ago
• Remove reactor-core dependency 💬 by Caio 7 years ago

  Not necessary since the Loader went away
  

• Drop the Serializer 💬 by Caio 7 years ago

  Useless and poorly named
  

• Drop the Loader 💬 by Caio 7 years ago

  This is being moved to casserole since it's what controls which
  facets to be indexed/collected.
  

• Merge remote-tracking branch 'origin/category-extractor' by Caio 7 years ago
• Add the "diet" category to the test indexer 💬 by Caio 7 years ago

  This is all that was really needed to make the tests pass again (sans
  API changes from previous patches, of course). I'm happy with it!
  

• IndexField.FACET_DIET doesn't exist anymore by Caio 7 years ago
• Adjust to new Searcher.Builder exception path by Caio 7 years ago
• Make taxonomy facets configurable by the client 💬 by Caio 7 years ago

  This commit makes the Indexer now accept an optional CategoryExtractor
  that is responsible to extract categories and labels from any recipe.
  
  What this effectively mean is that we support generic facets now and,
  by default, we don't index any facet. CategoryExtractorTest has an
  example on how this is all fitting together.
  
  I'm not particularly happy about the state Searcher.Builder() is at
  right now, but that's more cosmetic...
  

• Merge branch 'facets-again' by Caio 7 years ago
• Re-add support for facets 💬 by Caio 7 years ago

  Taxonomy-based single-index static facets. Not super fast, but
  saves us from doing range collections along with taxonomy ones.
  
  I don't have performance numbers from when we would expose all
  facets, so this is just comparing what's in master (Just a
  threshold-based count for diets) with the new implementation
  that has everything that's available as filters _but_ doesn't
  compute the threshold-aware counts:
  
  Old: QueryBenchmark.facetedQuery  thrpt  5  3.584 ± 0.012  ops/s
  New: QueryBenchmark.facetedQuery  thrpt  5  4.108 ± 0.013  ops/s
  
  So, still a 50+% throughput loss when compared to `basicQuery`,
  but now with all the counts that casserole cares about. Now
  we'll be able to use the SearchPolicy to conditionally allow
  computing the facets so that filters finally have counts :-)
  
  Since this patch doesn't implement the threshold-aware diet facet,
  the tests at SearcherTest.dietThreshold() have been modified to
  not care about facets.
  

• Revert "Upgrade to Lucene 8.0.0" 💬 by Caio 7 years ago

  This reverts commit 642a2cca848815702ea74895ffbd1cc5ffed418c.
  
  This upgrade has a performance regression I'm not willing to swallow
  before understanding what's going on:
  
  7.7.1: QueryBenchmark.basicQuery  thrpt  5  9.090 ± 0.090  ops/s
  8.0.0: QueryBenchmark.basicQuery  thrpt  5  6.528 ± 0.031  ops/s
  
  (Yes, these sorry numbers are from the production box)
  

• Match Indexer.addRecipe style to SearcherImpl.toLuceneQuery 💬 by Caio 7 years ago

  It's a bit more verbose but I find it easier to "get it".
  

• Upgrade to Lucene 8.0.0 by Caio 7 years ago
• Get rid of FileSystem, rework Builders 💬 by Caio 7 years ago

  Ever since IndexConfiguration was introduced, FileSystem stated looking
  (even more) awkward. So today I finally got around getting rid of it.
  
  As this was getting quite involved I decided to clean up the Indexer and
  Searcher builders so things would be easier to grok.
  

• Add links to Lucene / Chronicle-Map by Caio 7 years ago
• Add basic LICENSE and README files 💬 by Caio 7 years ago

  Nothing special, just preparing for making the repository public.
  

• Remove maven wrapper 💬 by Caio 7 years ago

  Even since the split I switched to using the system-installed
  maven. I could set the wrapper up on all the subprojects, but
  it's easier to just drop it here instead :-)
  

• [maven-release-plugin] prepare for next development iteration by Caio 7 years ago
• [maven-release-plugin] prepare release 0.1.0 by Caio 7 years ago
• Add more metadata to the db, adjust model.Recipe 💬 by Caio 7 years ago

  This patch changes our database format to not contain the instructions
  and adds some of the metadata from model.Recipe into it (namely:
  timing and nutrition details).
  
  The end result is a database file that's less than half the original
  size:
  
  933M    new/chronicle.db
  2.0G    old/chronicle.db
  

• [maven-release-plugin] prepare for next development iteration by Caio 7 years ago
• [maven-release-plugin] prepare release 0.0.5 by Caio 7 years ago
• Make Diet.isKnown() public by Caio 7 years ago
• Adapt to new data model version 💬 by Caio 7 years ago

  Glutenfree tags were unused and Vegan is too fussy to predict reliably
  with what I'm doing at the moment, so we replace it with Vegetarian.
  

• [maven-release-plugin] prepare for next development iteration by Caio 7 years ago