breaking changes: improved customization support

Author: n3integration

VERSION

@@ -1 +1 @@
-version=0.3.1
+version=0.4.0

func.go

@@ -1,34 +1,52 @@
 package classifier
 
+const defaultBufferSize = 50
+
 // Predicate provides a predicate function
 type Predicate func(string) bool
 
 // Mapper provides a map function
 type Mapper func(string) string
 
-// Map applies f to each element of the supplied input slice
-func Map(vs chan string, f Mapper) chan string {
-	outstream := make(chan string)
+// Map applies f to each element of the supplied input channel
+func Map(vs chan string, f ...Mapper) chan string {
+	stream := make(chan string, defaultBufferSize)
+
 	go func() {
 		for v := range vs {
-			outstream <- f(v)
+			for _, fn := range f {
+				v = fn(v)
+			}
+			stream <- v
 		}
-		close(outstream)
+		close(stream)
 	}()
-	return outstream
+
+	return stream
 }
 
-// Filter removes elements from the input slice where the supplied predicate
+// Filter removes elements from the input channel where the supplied predicate
 // is satisfied
-func Filter(vs chan string, f Predicate) chan string {
-	outstream := make(chan string)
+// Filter is a Predicate aggregation
+func Filter(vs chan string, filters ...Predicate) chan string {
+	stream := make(chan string, defaultBufferSize)
+	apply := func(text string) bool {
+		for _, f := range filters {
+			if !f(text) {
+				return false
+			}
+		}
+		return true
+	}
+
 	go func() {
-		for v := range vs {
-			if f(v) {
-				outstream <- v
+		for text := range vs {
+			if apply(text) {
+				stream <- text
 			}
 		}
-		close(outstream)
+		close(stream)
 	}()
-	return outstream
-}
+
+	return stream
+}

naive/naive.go

@@ -12,27 +12,42 @@ import (
 // ErrNotClassified indicates that a document could not be classified
 var ErrNotClassified = errors.New("unable to classify document")
 
+// Option provides a functional setting for the Classifier
+type Option func(c *Classifier) error
+
 // Classifier implements a naive bayes classifier
 type Classifier struct {
 	feat2cat  map[string]map[string]int
 	catCount  map[string]int
 	tokenizer classifier.Tokenizer
-	sync.RWMutex
+	mu        sync.RWMutex
 }
 
-// New initializes a new naive Classifier
-func New() *Classifier {
-	return &Classifier{
+// New initializes a new naive Classifier using the standard tokenizer
+func New(opts ...Option) *Classifier {
+	c := &Classifier{
 		feat2cat:  make(map[string]map[string]int),
 		catCount:  make(map[string]int),
 		tokenizer: classifier.NewTokenizer(),
 	}
+	for _, opt := range opts {
+		opt(c)
+	}
+	return c
+}
+
+// Tokenizer overrides the classifier's default Tokenizer
+func Tokenizer(t classifier.Tokenizer) Option {
+	return func(c *Classifier) error {
+		c.tokenizer = t
+		return nil
+	}
 }
 
 // Train provides supervisory training to the classifier
 func (c *Classifier) Train(r io.Reader, category string) error {
-	c.Lock()
-	defer c.Unlock()
+	c.mu.Lock()
+	defer c.mu.Unlock()
 
 	for feature := range c.tokenizer.Tokenize(r) {
 		c.addFeature(feature, category)
@@ -55,8 +70,8 @@ func (c *Classifier) Classify(r io.Reader) (string, error) {
 	classification := ""
 	probabilities := make(map[string]float64)
 
-	c.RLock()
-	defer c.RUnlock()
+	c.mu.RLock()
+	defer c.mu.RUnlock()
 
 	for _, category := range c.categories() {
 		probabilities[category] = c.probability(r, category)
@@ -65,6 +80,7 @@ func (c *Classifier) Classify(r io.Reader) (string, error) {
 			classification = category
 		}
 	}
+
 	if classification == "" {
 		return "", ErrNotClassified
 	}
@@ -152,5 +168,5 @@ func (c *Classifier) docProbability(r io.Reader, category string) float64 {
 }
 
 func asReader(text string) io.Reader {
-	return bytes.NewBuffer([]byte(text))
+	return bytes.NewBufferString(text)
 }

tokens.go

@@ -2,42 +2,48 @@ package classifier
 
 import (
 	"bufio"
-	"bytes"
 	"io"
-	"regexp"
 	"strings"
-	"unsafe"
 )
 
 // Tokenizer provides a common interface to tokenize documents
 type Tokenizer interface {
-	// Tokenize breaks the provided document into a token slice
-	Tokenize(r io.Reader) chan string
+	// Tokenize breaks the provided document into a channel of tokens
+	Tokenize(io.Reader) chan string
 }
 
-type regexTokenizer struct {
-	tokenizer *regexp.Regexp
-}
+// StdOption provides configuration settings for a StdTokenizer
+type StdOption func(*StdTokenizer)
 
-type stdTokenizer struct {
+// StdTokenizer provides a common document tokenizer that splits a
+// document by word boundaries
+type StdTokenizer struct {
+	transforms []Mapper
+	filters    []Predicate
+	bufferSize int
 }
 
 // NewTokenizer initializes a new standard Tokenizer instance
-func NewTokenizer() Tokenizer {
-	return &stdTokenizer{}
-}
-
-// NewRegexTokenizer initializes a new regular expression Tokenizer instance
-func NewRegexTokenizer() Tokenizer {
-	return &regexTokenizer{
-		tokenizer: regexp.MustCompile("\\W+"),
+func NewTokenizer(opts ...StdOption) *StdTokenizer {
+	tokenizer := &StdTokenizer{
+		bufferSize: 100,
+		transforms: []Mapper{
+			strings.ToLower,
+		},
+		filters: []Predicate{
+			IsNotStopWord,
+		},
+	}
+	for _, opt := range opts {
+		opt(tokenizer)
 	}
+	return tokenizer
 }
 
-func (t *stdTokenizer) Tokenize(r io.Reader) chan string {
+func (t *StdTokenizer) Tokenize(r io.Reader) chan string {
 	tokenizer := bufio.NewScanner(r)
 	tokenizer.Split(bufio.ScanWords)
-	tokens := make(chan string)
+	tokens := make(chan string, t.bufferSize)
 
 	go func() {
 		for tokenizer.Scan() {
@@ -46,27 +52,27 @@ func (t *stdTokenizer) Tokenize(r io.Reader) chan string {
 		close(tokens)
 	}()
 
-	return pipeline(tokens)
+	return t.pipeline(tokens)
 }
 
-// Tokenize extracts and normalizes all words from a text corpus
-func (t *regexTokenizer) Tokenize(r io.Reader) chan string {
-	buffer := new(bytes.Buffer)
-	buffer.ReadFrom(r)
-	b := buffer.Bytes()
-	doc := *(*string)(unsafe.Pointer(&b))
-	tokens := make(chan string)
-
-	go func() {
-		for _, token := range t.tokenizer.Split(doc, -1) {
-			tokens <- token
-		}
-		close(tokens)
-	}()
+func (t *StdTokenizer) pipeline(in chan string) chan string {
+	return Map(Filter(in, t.filters...), t.transforms...)
+}
 
-	return pipeline(tokens)
+func BufferSize(size int) StdOption {
+	return func(t *StdTokenizer) {
+		t.bufferSize = size
+	}
 }
 
-func pipeline(tokens chan string) chan string {
-	return Map(Filter(tokens, IsNotStopWord), strings.ToLower)
+func Transforms(m ...Mapper) StdOption {
+	return func(t *StdTokenizer) {
+		t.transforms = m
+	}
 }
+
+func Filters(f ...Predicate) StdOption {
+	return func(t *StdTokenizer) {
+		t.filters = f
+	}
+}

tokens_test.go

@@ -16,10 +16,6 @@ func TestTokenize(t *testing.T) {
 		tokens := NewTokenizer().Tokenize(toReader(text))
 		doTokenizeTest(t, tokens)
 	})
-	t.Run("Regexp Tokenizer", func(t *testing.T) {
-		tokens := NewRegexTokenizer().Tokenize(toReader(text))
-		doTokenizeTest(t, tokens)
-	})
 }
 
 func toReader(text string) io.Reader {