migrate from govendor to dep

2018-01-19 09:48:15 +01:00
parent 5bb22b836d
commit e023f79c5a
7349 changed files with 3153302 additions and 64302 deletions
@@ -0,0 +1,2 @@
+*.test
+*.prof
@@ -0,0 +1,31 @@
+# Perks for Go (golang.org)
+
+Perks contains the Go package quantile that computes approximate quantiles over
+an unbounded data stream within low memory and CPU bounds.
+
+For more information and examples, see:
+http://godoc.org/github.com/bmizerany/perks
+
+A very special thank you and shout out to Graham Cormode (Rutgers University),
+Flip Korn (AT&T Labs–Research), S. Muthukrishnan (Rutgers University), and
+Divesh Srivastava (AT&T Labs–Research) for their research and publication of
+[Effective Computation of Biased Quantiles over Data Streams](http://www.cs.rutgers.edu/~muthu/bquant.pdf)
+
+Thank you, also:
+* Armon Dadgar (@armon)
+* Andrew Gerrand (@nf)
+* Brad Fitzpatrick (@bradfitz)
+* Keith Rarick (@kr)
+
+FAQ:
+
+Q: Why not move the quantile package into the project root?
+A: I want to add more packages to perks later.
+
+Copyright (C) 2013 Blake Mizerany
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
@@ -0,0 +1,108 @@
+// Package histogram provides a Go implementation of BigML's histogram package
+// for Clojure/Java. It is currently experimental.
+package histogram
+
+import (
+	"container/heap"
+	"math"
+	"sort"
+)
+
+type Bin struct {
+	Count int
+	Sum   float64
+}
+
+func (b *Bin) Update(x *Bin) {
+	b.Count += x.Count
+	b.Sum += x.Sum
+}
+
+func (b *Bin) Mean() float64 {
+	return b.Sum / float64(b.Count)
+}
+
+type Bins []*Bin
+
+func (bs Bins) Len() int           { return len(bs) }
+func (bs Bins) Less(i, j int) bool { return bs[i].Mean() < bs[j].Mean() }
+func (bs Bins) Swap(i, j int)      { bs[i], bs[j] = bs[j], bs[i] }
+
+func (bs *Bins) Push(x interface{}) {
+	*bs = append(*bs, x.(*Bin))
+}
+
+func (bs *Bins) Pop() interface{} {
+	return bs.remove(len(*bs) - 1)
+}
+
+func (bs *Bins) remove(n int) *Bin {
+	if n < 0 || len(*bs) < n {
+		return nil
+	}
+	x := (*bs)[n]
+	*bs = append((*bs)[:n], (*bs)[n+1:]...)
+	return x
+}
+
+type Histogram struct {
+	res *reservoir
+}
+
+func New(maxBins int) *Histogram {
+	return &Histogram{res: newReservoir(maxBins)}
+}
+
+func (h *Histogram) Insert(f float64) {
+	h.res.insert(&Bin{1, f})
+	h.res.compress()
+}
+
+func (h *Histogram) Bins() Bins {
+	return h.res.bins
+}
+
+type reservoir struct {
+	n       int
+	maxBins int
+	bins    Bins
+}
+
+func newReservoir(maxBins int) *reservoir {
+	return &reservoir{maxBins: maxBins}
+}
+
+func (r *reservoir) insert(bin *Bin) {
+	r.n += bin.Count
+	i := sort.Search(len(r.bins), func(i int) bool {
+		return r.bins[i].Mean() >= bin.Mean()
+	})
+	if i < 0 || i == r.bins.Len() {
+		// TODO(blake): Maybe use an .insert(i, bin) instead of
+		// performing the extra work of a heap.Push.
+		heap.Push(&r.bins, bin)
+		return
+	}
+	r.bins[i].Update(bin)
+}
+
+func (r *reservoir) compress() {
+	for r.bins.Len() > r.maxBins {
+		minGapIndex := -1
+		minGap := math.MaxFloat64
+		for i := 0; i < r.bins.Len()-1; i++ {
+			gap := gapWeight(r.bins[i], r.bins[i+1])
+			if minGap > gap {
+				minGap = gap
+				minGapIndex = i
+			}
+		}
+		prev := r.bins[minGapIndex]
+		next := r.bins.remove(minGapIndex + 1)
+		prev.Update(next)
+	}
+}
+
+func gapWeight(prev, next *Bin) float64 {
+	return next.Mean() - prev.Mean()
+}
@@ -0,0 +1,90 @@
+package topk
+
+import (
+	"sort"
+)
+
+// http://www.cs.ucsb.edu/research/tech_reports/reports/2005-23.pdf
+
+type Element struct {
+	Value string
+	Count int
+}
+
+type Samples []*Element
+
+func (sm Samples) Len() int {
+	return len(sm)
+}
+
+func (sm Samples) Less(i, j int) bool {
+	return sm[i].Count < sm[j].Count
+}
+
+func (sm Samples) Swap(i, j int) {
+	sm[i], sm[j] = sm[j], sm[i]
+}
+
+type Stream struct {
+	k   int
+	mon map[string]*Element
+
+	// the minimum Element
+	min *Element
+}
+
+func New(k int) *Stream {
+	s := new(Stream)
+	s.k = k
+	s.mon = make(map[string]*Element)
+	s.min = &Element{}
+
+	// Track k+1 so that less frequenet items contended for that spot,
+	// resulting in k being more accurate.
+	return s
+}
+
+func (s *Stream) Insert(x string) {
+	s.insert(&Element{x, 1})
+}
+
+func (s *Stream) Merge(sm Samples) {
+	for _, e := range sm {
+		s.insert(e)
+	}
+}
+
+func (s *Stream) insert(in *Element) {
+	e := s.mon[in.Value]
+	if e != nil {
+		e.Count++
+	} else {
+		if len(s.mon) < s.k+1 {
+			e = &Element{in.Value, in.Count}
+			s.mon[in.Value] = e
+		} else {
+			e = s.min
+			delete(s.mon, e.Value)
+			e.Value = in.Value
+			e.Count += in.Count
+			s.min = e
+		}
+	}
+	if e.Count < s.min.Count {
+		s.min = e
+	}
+}
+
+func (s *Stream) Query() Samples {
+	var sm Samples
+	for _, e := range s.mon {
+		sm = append(sm, e)
+	}
+	sort.Sort(sort.Reverse(sm))
+
+	if len(sm) < s.k {
+		return sm
+	}
+
+	return sm[:s.k]
+}