// FetchSimilar is an extended version of Fetch
// It is uses a DirTree of crawled *links*, not actual files.
// As it moves up the DOM, it crawls every document for additional links.
// It first moves up to find similar URLs on the same depth
// /\
// /\ / \
// /\ / \ / \
// It then moves up the ladder again - to accept higher URLs
// /\
// /\
// /\
func FetchSimilar(w http.ResponseWriter, r *http.Request, m map[string]interface{}) {
lg, b := loghttp.BuffLoggerUniversal(w, r)
closureOverBuf := func(bUnused *bytes.Buffer) {
loghttp.Pf(w, r, b.String())
}
defer closureOverBuf(b) // the argument is ignored,
r.Header.Set("X-Custom-Header-Counter", "nocounter")
start := time.Now()
wpf(b, tplx.ExecTplHelper(tplx.Head, map[string]interface{}{"HtmlTitle": "Find similar HTML URLs"}))
defer wpf(b, tplx.Foot)
wpf(b, "<pre>")
defer wpf(b, "</pre>")
fs1 := GetFS(appengine.NewContext(r))
err := r.ParseForm()
lg(err)
countSimilar := 3
sCountSimilar := r.FormValue("cnt")
if sCountSimilar != "" {
i, err := strconv.Atoi(strings.TrimSpace(sCountSimilar))
if err == nil {
countSimilar = i
}
}
surl := r.FormValue(routes.URLParamKey)
ourl, err := fetch.URLFromString(surl)
lg(err)
if err != nil {
return
}
if ourl.Host == "" {
lg("host is empty (%v)", surl)
return
}
knownProtocol := ""
if r.FormValue("prot") != "" {
knownProtocol = r.FormValue("prot")
}
numWorkers := 0
sNumWorkers := r.FormValue("numworkers")
if sNumWorkers != "" {
i, err := strconv.Atoi(strings.TrimSpace(sNumWorkers))
if err == nil {
numWorkers = i
}
}
srcDepth := strings.Count(ourl.Path, "/")
cmd := FetchCommand{}
cmd.Host = ourl.Host
cmd.SearchPrefix = ourl.Path
cmd = addDefaults(cmd)
dirTree := &DirTree{Name: "/", Dirs: map[string]DirTree{}, EndPoint: true}
fnDigest := path.Join(docRoot, cmd.Host, "digest2.json")
loadDigest(w, r, lg, fs1, fnDigest, dirTree) // previous
lg("dirtree 400 chars is %v end of dirtree\t\t", stringspb.ToLen(dirTree.String(), 400))
m1 := new(MyWorker)
m1.r = r
m1.lg = lg
m1.fs1 = fs1
m1.SURL = path.Join(cmd.Host, ourl.Path)
m1.Protocol = knownProtocol
btsSrc, modSrc, usedExisting, err := fetchSave(m1)
if !usedExisting {
addAnchors(lg, cmd.Host, btsSrc, dirTree)
}
lg(err)
if err != nil {
return
}
lg("\t\t%4.2v secs so far 1", time.Now().Sub(start).Seconds())
var treePath string
treePath = "/blogs/freeexchange"
treePath = "/news/europe"
treePath = path.Dir(ourl.Path)
opt := LevelWiseDeeperOptions{}
opt.Rump = treePath
opt.ExcludeDir = "/news/americas"
opt.ExcludeDir = "/blogs/buttonwood"
opt.ExcludeDir = "/something-impossible"
opt.MinDepthDiff = 1
opt.MaxDepthDiff = 1
opt.CondenseTrailingDirs = cmd.CondenseTrailingDirs
opt.MaxNumber = cmd.DesiredNumber + 1 // one more for "self"
opt.MaxNumber = cmd.DesiredNumber + 40 // collect more, 'cause we filter out those too old later
var subtree *DirTree
links := []FullArticle{}
alreadyCrawled := map[string]struct{}{}
MarkOuter:
for j := 0; j < srcDepth; j++ {
treePath = path.Dir(ourl.Path)
MarkInner:
// for i := 1; i < srcDepth; i++ {
for i := 1; i < (srcDepth + 5); i++ {
subtree, treePath = DiveToDeepestMatch(dirTree, treePath)
lg("Looking from height %v to level %v - %v", srcDepth-i, srcDepth-j, treePath)
if _, ok := alreadyCrawled[treePath]; ok {
// lg("\t already digested %v", treePath)
continue
}
m2 := new(MyWorker)
m2.r = r
m2.lg = lg
m2.fs1 = fs1
m2.SURL = path.Join(cmd.Host, treePath)
m2.Protocol = knownProtocol
btsPar, _, usedExisting, err := fetchSave(m2)
lg(err)
if err != nil {
return
}
alreadyCrawled[treePath] = struct{}{}
if !usedExisting {
addAnchors(lg, cmd.Host, btsPar, dirTree)
}
if subtree == nil {
lg("\n#%v treePath %q ; subtree is nil", i, treePath)
} else {
// lg("\n#%v treePath %q ; subtree exists", i, treePath)
opt.Rump = treePath
opt.MinDepthDiff = i - j
opt.MaxDepthDiff = i - j
lvlLinks := LevelWiseDeeper(nil, nil, subtree, opt)
links = append(links, lvlLinks...)
for _, art := range lvlLinks {
_ = art
// lg("#%v fnd %v", i, stringspb.ToLen(art.Url, 100))
}
if len(links) >= opt.MaxNumber {
lg("found enough links")
break MarkOuter
}
pathPrev := treePath
treePath = path.Dir(treePath)
// lg("#%v bef %v - aft %v", i, pathPrev, treePath)
if pathPrev == "." && treePath == "." ||
pathPrev == "/" && treePath == "/" ||
pathPrev == "" && treePath == "." {
lg("break to innner")
break MarkInner
}
}
}
}
//
//
//
//
lg("%v links after %4.2v secs", len(links), time.Now().Sub(start).Seconds())
lg("============================")
lg("Now reading/fetching actual similar files - not just the links")
//
tried := 0
selecteds := []FullArticle{}
nonExisting := []FullArticle{}
nonExistFetched := []FullArticle{}
for _, art := range links {
if art.Url == ourl.Path {
lg("skipping self\t%v", art.Url)
continue
}
tried++
useExisting := false
semanticUri := condenseTrailingDir(art.Url, cmd.CondenseTrailingDirs)
p := path.Join(docRoot, cmd.Host, semanticUri)
f, err := fs1.Open(p)
// lg(err) // its no error if file does not exist
if err != nil {
// lg("!nstore %q", semanticUri)
} else {
// lg("reading %q", semanticUri)
// lets put this into a func, so that f.close it called at the end of this func
// otherwise defer f.close() spans the entire func and prevents
// overwrites chmods further down
f := func() {
defer f.Close()
fi, err := f.Stat()
lg(err)
if err != nil {
} else {
age := time.Now().Sub(fi.ModTime())
if age.Hours() < 10 {
lg("\t\tusing existing file with age %4.2v hrs", age.Hours())
art.Mod = fi.ModTime()
bts, err := ioutil.ReadAll(f)
lg(err)
art.Body = bts
if len(bts) < 200 {
if bytes.Contains(bts, []byte(fetch.MsgNoRdirects)) {
return
}
}
selecteds = append(selecteds, art)
useExisting = true
}
}
}
f()
}
if !useExisting {
nonExisting = append(nonExisting, art)
}
if len(selecteds) >= countSimilar {
break
}
}
lg("============================")
lg("tried %v links - yielding %v existing similars; not existing in datastore: %v, %v were requested.",
tried, len(selecteds), len(nonExisting), countSimilar)
if len(selecteds) < countSimilar {
jobs := make([]distrib.Worker, 0, len(nonExisting))
for _, art := range nonExisting {
surl := path.Join(cmd.Host, art.Url)
wrkr := MyWorker{SURL: surl}
wrkr.Protocol = knownProtocol
wrkr.r = r
wrkr.lg = lg
wrkr.fs1 = fs1
job := distrib.Worker(&wrkr)
jobs = append(jobs, job)
}
opt := distrib.NewDefaultOptions()
opt.TimeOutDur = 3500 * time.Millisecond
opt.Want = int32(countSimilar - len(selecteds) + 4) // get some more, in case we have "redirected" bodies
opt.NumWorkers = int(opt.Want) // 5s query limit; => hurry; spawn as many as we want
if numWorkers > 0 {
opt.NumWorkers = numWorkers
}
lg("Preparing %v simultaneous, wanting %v fetches; at %4.2v secs.", opt.NumWorkers, opt.Want, time.Now().Sub(start).Seconds())
opt.CollectRemainder = false // 5s query limit; => hurry; dont wait for stragglers
ret, msg := distrib.Distrib(jobs, opt)
lg("Distrib returned at %4.2v secs with %v results.", time.Now().Sub(start).Seconds(), len(ret))
lg("\n" + msg.String())
for _, v := range ret {
v1, _ := v.Worker.(*MyWorker)
if v1.FA != nil {
age := time.Now().Sub(v1.FA.Mod)
if age.Hours() < 10 {
lg("\t\tusing fetched file with age %4.2v hrs", age.Hours())
nonExistFetched = append(nonExistFetched, *v1.FA)
if len(nonExistFetched) > (countSimilar - len(selecteds)) {
break
}
}
}
if v1.err != nil {
lg(err)
}
}
lg("tried %v links - yielding %v fetched - jobs %v", len(nonExisting), len(nonExistFetched), len(jobs))
selecteds = append(selecteds, nonExistFetched...)
//
//
// Extract links
for _, v := range nonExistFetched {
// lg("links -> memory dirtree for %q", v.Url)
addAnchors(lg, cmd.Host, v.Body, dirTree)
}
}
//
if time.Now().Sub(dirTree.LastFound).Seconds() < 10 {
lg("saving accumulated (new) links to digest")
saveDigest(lg, fs1, fnDigest, dirTree)
}
lg("\t\t%4.2v secs so far 3", time.Now().Sub(start).Seconds())
mp := map[string][]byte{}
mp["msg"] = b.Bytes()
mp["url_self"] = []byte(condenseTrailingDir(ourl.Path, cmd.CondenseTrailingDirs))
mp["mod_self"] = []byte(modSrc.Format(http.TimeFormat))
mp["bod_self"] = btsSrc
for i, v := range selecteds {
mp["url__"+spf("%02v", i)] = []byte(v.Url)
mp["mod__"+spf("%02v", i)] = []byte(v.Mod.Format(http.TimeFormat))
mp["bod__"+spf("%02v", i)] = v.Body
}
mp["lensimilar"] = []byte(spf("%02v", len(selecteds)))
//
smp, err := json.MarshalIndent(mp, "", "\t")
if err != nil {
lg(b, "marshalling mp to []byte failed\n")
return
}
r.Header.Set("X-Custom-Header-Counter", "nocounter")
w.Header().Set("Content-Type", "application/json")
w.Write(smp)
b.Reset() // this keeps the buf pointer intact; outgoing defers are still heeded
b = new(bytes.Buffer) // creates a *new* buf pointer; outgoing defers write into the *old* buf
lg("\t\t%4.2v secs so far 4 (json resp written as []byte)", time.Now().Sub(start).Seconds())
return
}
// +build p4
// go test -tags=p4
package test
import (
"fmt"
"math/rand"
"testing"
"time"
"github.com/pbberlin/tools/distrib"
)
var opt = distrib.NewDefaultOptions()
type MyWorker struct {
Inp, Res int
}
func (m *MyWorker) Work() {
m.Res = 500 + m.Inp
time.Sleep(time.Millisecond * time.Duration(rand.Intn(20)))
}
func Test4(t *testing.T) {
opt.CollectRemainder = true
do(t, 3)
}