func CrawlPerson(httpClient doc.HttpClient, id string) (*Person, error) {
site, user := ParsePersonId(id)
switch site {
case "github.com":
u, err := GithubSpider.ReadUser(user)
if err != nil {
return nil, errorsp.WithStacksAndMessage(err, "ReadUser %s failed", id)
}
p := &Person{Id: id}
for name, ri := range u.Repos {
path := user + "/" + name
p.Packages = append(p.Packages, "github.com/"+path)
if err := store.UpdatePackage(site, path, func(info *stpb.PackageInfo) error {
info.RepoInfo = ri
return nil
}); err != nil {
log.Printf("UpdatePackage %v %v failed: %v", site, path, err)
} else {
log.Printf("UpdatePackage %v %v success", site, path)
}
}
return p, nil
case "bitbucket.org":
p, err := doc.GetBitbucketPerson(httpClient, map[string]string{"owner": user})
if err != nil {
return nil, errorsp.WithStacks(err)
}
return &Person{
Id: id,
Packages: p.Projects,
}, nil
}
return nil, nil
}
func CrawlRepoInfo(site, user, name string) *sppb.RepoInfo {
// Check cache in store.
path := user + "/" + name
p, err := store.ReadPackage(site, path)
if err != nil {
log.Printf("ReadPackage %v %v failed: %v", site, path, err)
} else {
if p.RepoInfo != nil && store.RepoInfoAge(p.RepoInfo) < maxRepoInfoAge {
log.Printf("Repo cache of %s/%s hit", site, path)
bi.Inc("crawler.repocache.hit")
return p.RepoInfo
}
}
bi.Inc("crawler.repocache.miss")
ri, err := GithubSpider.ReadRepository(user, name)
if err != nil {
if errorsp.Cause(err) == github.ErrInvalidRepository {
if err := store.DeletePackage(site, path); err != nil {
log.Printf("DeleteRepoInfo %v %v failed: %v", site, path, err)
}
}
return nil
}
if err := store.UpdatePackage(site, path, func(info *stpb.PackageInfo) error {
info.RepoInfo = ri
return nil
}); err != nil {
log.Printf("UpdatePackage %v %v failed: %v", site, path, err)
} else {
log.Printf("UpdatePackage %s %s success", site, path)
}
return ri
}
func appendNewPackage(pkg, foundWay string) {
cDB.AppendPackage(pkg, allDocsPkgs.Contain)
site, path := utils.SplitPackage(pkg)
if err := store.UpdatePackage(site, path, func(*stpb.PackageInfo) error {
return nil
}); err != nil {
log.Printf("UpdatePackage %s %s failed: %v", site, path, err)
}
if err := store.AppendPackageEvent(site, path, foundWay, time.Now(), sppb.HistoryEvent_Action_None); err != nil {
log.Printf("UpdatePackageHistory %s %s failed: %v", site, path, err)
}
}
// OnlyMapper.Map
func (pc *PackageCrawler) Map(key, val sophie.SophieWriter, c []sophie.Collector) error {
if time.Now().After(AppStopTime) {
log.Printf("[Part %d] Timeout(key = %v), PackageCrawler returns EOM",
pc.part, key)
return mr.EOM
}
pkg := string(*key.(*sophie.RawString))
ent := val.(*gcse.CrawlingEntry)
if ent.Version < gcse.CrawlerVersion {
// if gcse.CrawlerVersion is larger than Version, Etag is ignored.
ent.Etag = ""
}
log.Printf("[Part %d] Crawling package %v with etag %s\n", pc.part, pkg, ent.Etag)
p, flds, err := gcse.CrawlPackage(pc.httpClient, pkg, ent.Etag)
for _, fld := range flds {
if spider.LikeGoSubFolder(fld.Name) {
appendNewPackage(pkg+"/"+fld.Path, "parent")
}
}
site, path := utils.SplitPackage(pkg)
if err != nil && errorsp.Cause(err) != gcse.ErrPackageNotModifed {
log.Printf("[Part %d] Crawling pkg %s failed: %v", pc.part, pkg, err)
if gcse.IsBadPackage(err) {
utils.LogError(store.AppendPackageEvent(site, path, "", time.Now(), sppb.HistoryEvent_Action_Invalid), "AppendPackageEvent %v %v failed", site, path)
bi.AddValue(bi.Sum, "crawler.package.wrong-package", 1)
// a wrong path
nda := gcse.NewDocAction{
Action: gcse.NDA_DEL,
}
c[0].Collect(sophie.RawString(pkg), &nda)
cDB.PackageDB.Delete(pkg)
log.Printf("[Part %d] Remove wrong package %s", pc.part, pkg)
} else {
utils.LogError(store.AppendPackageEvent(site, path, "", time.Now(), sppb.HistoryEvent_Action_Failed), "AppendPackageEvent %v %v failed", site, path)
bi.Inc("crawler.package.failed")
if strings.HasPrefix(pkg, "github.com/") {
bi.Inc("crawler.package.failed.github")
}
pc.failCount++
cDB.SchedulePackage(pkg, time.Now().Add(12*time.Hour), ent.Etag)
if pc.failCount >= 10 || strings.Contains(err.Error(), "403") {
durToSleep := 10 * time.Minute
if time.Now().Add(durToSleep).After(AppStopTime) {
log.Printf("[Part %d] Timeout(key = %v), PackageCrawler returns EOM",
pc.part, key)
return mr.EOM
}
log.Printf("[Part %d] Last ten crawling packages failed, sleep for a while...(current: %s)",
pc.part, pkg)
time.Sleep(durToSleep)
pc.failCount = 0
}
}
return nil
}
utils.LogError(store.AppendPackageEvent(site, path, "", time.Now(), sppb.HistoryEvent_Action_Success), "AppendPackageEvent %v %v failed", site, path)
pc.failCount = 0
if errorsp.Cause(err) == gcse.ErrPackageNotModifed {
// TODO crawling stars for unchanged project
log.Printf("[Part %d] Package %s unchanged!", pc.part, pkg)
schedulePackageNextCrawl(pkg, ent.Etag)
bi.AddValue(bi.Sum, "crawler.package.not-modified", 1)
return nil
}
bi.AddValue(bi.Sum, "crawler.package.success", 1)
if strings.HasPrefix(pkg, "github.com/") {
bi.AddValue(bi.Sum, "crawler.package.success.github", 1)
}
log.Printf("[Part %d] Crawled package %s success!", pc.part, pkg)
var pkgInfo *stpb.PackageInfo
if err := store.UpdatePackage(site, path, func(pi *stpb.PackageInfo) error {
fillPackageInfo(p, pi)
pkgInfo = pi
return nil
}); err != nil {
log.Printf("UpdatePackage %v %v failed: %v", site, path, err)
}
saveRelatedInfo(pkgInfo)
nda := gcse.NewDocAction{
Action: gcse.NDA_UPDATE,
DocInfo: packageToDoc(p),
}
c[0].Collect(sophie.RawString(pkg), &nda)
log.Printf("[Part %d] Package %s saved!", pc.part, pkg)
if !strings.HasPrefix(pkg, "github.com/") {
// github.com throttling is done within the GithubSpider.
time.Sleep(10 * time.Second)
}
return nil
}