// Fingerprint generates the fingerprint of an HTML from the io.Reader r and a shingle factor.
// Shingle refers to the level of shuffling.
// E.g. with shingle factor =2, input "a", "b", "c" will be converted to "a b", "b c"
func Fingerprint(r io.Reader, shingle int) uint64 {
if shingle < 1 {
shingle = 1
}
// collect the features via this cf channel.
cf := make(chan string, 1000)
cs := make(chan uint64, 1000)
v := simhash.Vector{}
// Tokenize and then Generate Features. .
go func() {
defer close(cf)
z := html.NewTokenizer(r)
// TODO - export the max token count as an function argument.
count := 0
for tt := z.Next(); count < 5000 && tt != html.ErrorToken; tt = z.Next() {
t := z.Token()
count++
genFeatures(&t, cf)
}
}()
// Collect the features.
go func() {
defer close(cs)
a := make([][]byte, shingle)
for f := <-cf; f != ""; f = <-cf {
// shingle: generate the k-gram token as a single feature.
a = append(a[1:], []byte(f))
// fmt.Printf("%#v\n", a)
// fmt.Printf("%s\n", bytes.Join(a, []byte(" ")))
cs <- simhash.NewFeature(bytes.Join(a, []byte(" "))).Sum()
// cs <- simhash.NewFeature([]byte(f)).Sum()
}
}()
// from the checksum (of feature), append to vector.
for s := <-cs; s != 0; s = <-cs {
for i := uint8(0); i < 64; i++ {
bit := ((s >> i) & 1)
if bit == 1 {
v[i]++
} else {
v[i]--
}
}
}
return simhash.Fingerprint(v)
}
func Compute(r io.Reader, N, shingle int) (uint64, error) {
if shingle < 1 {
shingle = 1
}
chFeature := make(chan string, 128)
z := html.NewTokenizer(r)
go func() {
count := 1
for tt := z.Next(); count < N && tt != html.ErrorToken; tt = z.Next() {
t := z.Token()
count++
genFeature(&t, chFeature)
}
close(chFeature)
}()
ch := make(chan uint64, 128)
go func() {
// Avoid allocation
s := make([][]byte, shingle)
joined := make([][]byte, 2*shingle-1)
space := []byte(" ")
var i, n int
for f := range chFeature {
// Collect enough features
if n < shingle {
s[n] = []byte(f)
if n++; n == shingle {
goto JOIN
}
continue
}
// Shift array to produce one space
for i = 0; i < shingle-1; i++ {
s[i] = s[i+1]
}
s[i] = []byte(f)
JOIN:
for i, f := range s {
joined[2*i] = f
if i+1 != len(s) {
joined[2*i+1] = space
}
}
ch <- hash(joined...)
}
close(ch)
}()
v := simhash.Vector{}
var i uint
var bit int
for n := range ch {
for i = 0; i < 64; i++ {
bit = int((n >> i) & 1)
// bit == 1 ? 1 : -1
v[i] += (bit ^ (bit - 1))
}
}
if err := z.Err(); err != io.EOF {
return 0, err
}
return simhash.Fingerprint(v), nil
}
