// Simplified and altered page rank
func WeightTags(s *session.Session) {
s.IncrementState()
tags := s.GetHashtagCursor()
goodWeight := 1.0 / s.GetHashtagSize(true)
badWeight := -1.0 / s.GetHashtagSize(false)
for {
key, err := tags.Next(nil)
if err != nil {
log.Println(err)
break // No further entities match the query.
}
hashtag := s.Hashtag(key.StringID())
log.Println(hashtag.Name)
processTags(s, goodWeight, hashtag.Beneficiaries)
processTags(s, badWeight, hashtag.Victims)
}
s.SetTopTags()
// Move on
s.IncrementStep()
s.StopProcessing()
}
// Might be worth putting a defer in here incase a job breaks
// I.E to Clean up is processing
func Learn(s *session.Session) string {
if s.SetLearning() {
// New
log.Println("Set up learning")
status := http.GetStatus(s)
s.SetLimits(int(status.Follows), int(status.Followed_by))
}
if !s.SetProcessing() {
// Show we're still working
return " *"
}
switch s.GetLearningStep() {
case utils.APPRAISE:
jobs.MinePeople(s)
return StatusBar(s, "Mining Followers")
case utils.SCORN:
jobs.MinePeople(s)
return StatusBar(s, "Mining Following")
case utils.BUILD:
// Logistic Regression
// Get records and run
go jobs.LogisticRegression(s)
s.IncrementStep()
return "* Running Logistic Regression"
case utils.GOODTAGS:
go jobs.CrawlTags(s, true)
return StatusBar(s, "Finding Good Tags")
case utils.BADTAGS:
go jobs.CrawlTags(s, false)
return StatusBar(s, "Finding Bad Tags")
case utils.COMPUTETAGS:
go jobs.WeightTags(s)
return "* Ranking Tags"
case utils.SHARE:
go s.Share()
s.IncrementStep()
s.StopProcessing()
return "Sharing"
}
return "Stop"
}
func CrawlTags(s *session.Session, positive bool) {
keys := s.GetPeopleCursor(positive, utils.MAXREQUESTS*s.GetState())
s.IncrementState()
total := 0
size := 0
for {
key, err := keys.Next(nil)
total += 1
if err != nil {
if total < utils.MAXREQUESTS {
s.IncrementStep()
}
break // No further entities match the query.
}
media := http.GetMedia(s, key.StringID()).Data
captured := len(media) - 1
for i := 0; i < 3 && i < captured; i++ {
tagCount := len(media[i].Tags)
lim := 5
for j, tag := range media[i].Tags {
if tag == "" {
continue
}
if j >= lim {
break
}
h := s.Hashtag(tag)
for k := 0; k < lim && k < tagCount; k++ {
if j == k {
continue
}
if tag == media[i].Tags[k] {
lim += 1
media[i].Tags[k] = ""
continue
}
if positive {
h.Beneficiaries = append(h.Beneficiaries, media[i].Tags[k])
} else {
h.Victims = append(h.Beneficiaries, media[i].Tags[k])
}
}
s.SaveHashtag(h)
size += 1
}
}
}
s.IncrementSize(size, positive)
s.StopProcessing()
}
func MinePeople(s *session.Session) {
// Set up channel
next := make(chan *http.Users)
var batch http.Users
if s.GetNext() == "" {
if s.GetState() > 0 {
s.IncrementStep()
}
if s.GetLearningStep() == utils.APPRAISE {
batch = http.GetFollowers(s)
} else {
batch = http.GetFollowing(s)
}
} else {
batch = http.GetNext(s, s.GetNext())
}
go listen(s, next, 0, s.IncrementState())
next <- &batch
}
// Async set up multi calls
func listen(s *session.Session, next chan *http.Users, calls int, follows float64) {
for {
select {
case users := <-next:
i := len(users.Data) - 1
s.IncrementCount()
go process(s, users, i, follows)
close(next)
if calls == utils.MAXPEOPLEGRAB {
s.SetNext(users.Pagination.Next_url)
return
}
var batch http.Users
nxt := make(chan *http.Users)
if users.Pagination.Next_url != "" {
log.Println("Getting another batch")
batch = http.GetNext(s, users.Pagination.Next_url)
} else if follows == 0 { // follows == float64(s.GetLearningStep()) then have a array of functions
log.Println("Proceeding to next Step")
s.IncrementStep()
s.IncrementState()
batch = http.GetFollowing(s)
follows = float64(s.GetLearningStep())
} else {
s.SetNext("")
return
}
go listen(s, nxt, calls+1, follows)
nxt <- &batch
return
}
}
}