Log

• Merge pull request #18 from zeebo/cdf-fix 💬 by Caio 8 years ago

  fix cdf for values near the last centroid
  
  Fixes #17 

• fix cdf for values near the last centroid by Jeff Wendling 8 years ago
• Properly compute extreme CDFs 💬 by Caio 8 years ago

  When we reach the last centroid in the summary, what we want to
  do is assume that the last two centroids are of equal width
  then estimate the CDF via a simple interpolation. Before this
  patch we would wrongly bound (and compute) this estimation with
  the last item instead of the one before the last.
  
  Fixes #17
  

• Update more examples using the old New() api by Caio 8 years ago
• Make the example actually compile :flushed: by Caio 8 years ago
• Fix misspell by Caio 8 years ago
• Merge pull request #15 from nathanleclaire/nathanleclaire-fix-readme-typo-1 💬 by Caio 8 years ago

  Fix typo

• Fix typo by Nathan LeClaire 8 years ago
• Test go's tip, but don't wait for it 💬 by Caio 8 years ago

  On 1804f1cf2b I stopped testing `tip` because it was too slow,
  but it seems like I can test it, but ignore its results and
  check for test suite completion without waiting for it.
  

• Report allocations on benchmarks by Caio 8 years ago
• Use a fenwick tree to optimize prefix sums 💬 by Caio 8 years ago

  This patch makes summary use the excellent fenwick package from
  https://github.com/yourbasic/fenwick to dramatically speed
  up Add() calls while adding a bounded cost to memory.
  
      benchmark             old ns/op     new ns/op     delta
      BenchmarkAdd1-4       187           206           +10.16%
      BenchmarkAdd10-4      332           274           -17.47%
      BenchmarkAdd100-4     1092          325           -70.24%
  
  The tree creation could be faster, but profiling has shown that
  it doesn't seem to be that much of a problem (that's because
  `summary.setAt()` happens a *lot* more than `summary.Add()`)
  

• Refactor TDigest.AddWeighted 💬 by Caio 8 years ago

  This patch simply extracts some functionality into a couple
  a new private methods: findNeighbors and chooseMergeCandidate
  

• Tidy things up [gometalinter] by Caio 8 years ago
• Add a basic .gitignore by Caio 8 years ago
• More README tweaks 💬 by Caio 8 years ago

  Now we frame gopkg.in secondly since the expected preferred
  installation method is soon gonna be via `get+dep`
  

• Make travis use `dep` when testing by Caio 8 years ago
• Control dependencies via dep 💬 by Caio 8 years ago

  This patch freezes the go_rng dependency to a particular
  commit since the project doesn't tag its stable releases.
  

• Add tests for a heavily skewed gamma distribution 💬 by Caio 8 years ago

  The code is mostly borrowed from TDigestTest.java and
  could easily be modified to allow testing with multiple
  distributions/ranges. Closes #13
  
  Note that this patch adds a new test-only dependency but
  we don't use any form of dependency management - this
  will come in a subsequent patch.
  

• Add new CDF(float64) public method by Caio 8 years ago
• Add a simple CONTRIBUTING.md guide by Caio 8 years ago
• Beef the README up a bit by Caio 8 years ago
• Remove TDigest.Len() from the public interface 💬 by Caio 8 years ago

  I can't think of a scenario where one would really care about
  how many distinct centroids are in the digest, so away this
  goes on a major release. Adding it back in case it's needed
  won't require a major release.
  

• Tweak the docs a bit 💬 by Caio 8 years ago

  Paragraphs are good!
  

• Introduce TDigest.Count() 💬 by Caio 8 years ago

  Expose the count of samples publicly so that users can more
  easily deicde what to do when the digest has too many samples.
  

• Change TDigest.count to uint64 (was uint32) 💬 by Caio 8 years ago

  So now we can register more than 4B samples and still reply with valid
  quantile estimations, but notice that centroid counts are still uint32
  and that `t.count` is used for floating point divisions across the
  algorithms so a really big count is not really desirable.
  
  Individual centroid counts wrapping around uint32 is a pathological
  scenario that can totally be simulated and treated for, but that would
  mean a lot of extra comparisons added in the critical path, so I prefer
  to leave that to the user's decision.
  

• Update README with configuration notes by Caio 8 years ago
• Get rid of the centroid abstraction 💬 by Caio 8 years ago

  This was only being used to pack {float64,uint32}, all the
  other functionality was skipped or became unused over
  time for performance reasons. Away it goes.
  

• Expose options to change the RNG being used 💬 by Caio 8 years ago

  This patch creates a new public interface `TDigestRNG`
  and exposes two new options:
  
  - tdigest.RandomNumberGenerator(TDigestRNG)
  - tdigest.LocalRandomNumberGenerator(seed)
  

• Abstract math/rand usage into the TDigestRNG interface by Caio 8 years ago
• Make Add take only one parameter, introduce AddWeighted 💬 by Caio 8 years ago

  This patch renames the previous Add(float64,uint32) to AddWeighted
  and introduces a method Add(float64) which is simply an alias
  to AddWeighted(float64,1).