func TestCM(t *testing.T) {
var err error
check := func() {
if err != nil {
t.Fatal(err)
}
}
cm, _ := countmin.New(5, 16)
db, err := MakeDB(":memory:", true, &Settings{"foowiki.xml.bz2", 8})
check()
for _, i := range []uint32{1, 6, 13, 7, 8, 20, 44} {
cm.Add(i, i+5)
}
err = StoreCM(db, cm)
check()
got, err := LoadCM(db)
check()
if !reflect.DeepEqual(cm.Counts(), got.Counts()) {
t.Errorf("expected %v, got %v", cm.Counts(), got)
}
}
func processPages(articles <-chan *wikidump.Page,
linkch chan<- *processedLink, narticles *uint32,
nrows, ncols, maxN int) *countmin.Sketch {
ngramcount, err := countmin.New(nrows, ncols)
if err != nil {
// Shouldn't happen; we already constructed a count-min sketch
// with the exact same size in main.
panic(err)
}
for a := range articles {
text := wikidump.Cleanup(a.Text)
links := wikidump.ExtractLinks(text)
for link, freq := range links {
linkch <- processLink(&link, freq, maxN)
}
tokens := nlp.Tokenize(text)
for _, h := range hash.NGrams(tokens, 1, maxN) {
ngramcount.Add1(h)
}
atomic.AddUint32(narticles, 1)
}
return ngramcount
}
func makeSemanticizer() Semanticizer {
cm, _ := countmin.New(10, 4)
db, _ := storage.MakeDB(":memory:", true, &storage.Settings{MaxNGram: 2})
allq, _ := prepareAllQuery(db)
sem := Semanticizer{db: db, ngramcount: cm, maxNGram: 2, allQuery: allq}
for _, h := range hash.NGrams([]string{"Hello", "world"}, 2, 2) {
_, err := db.Exec(`insert into linkstats values (?, 0, 1)`, h)
if err == nil {
_, err = db.Exec(`insert into titles values (0, "dmr")`)
}
if err != nil {
panic(err)
}
}
return sem
}
func main() {
kingpin.Parse()
log.SetPrefix("dumpparser ")
var err error
check := func() {
if err != nil {
log.Fatal(err)
}
}
if *download != "" {
*dumppath, err = wikidump.Download(*download, *dumppath, true)
check()
} else if *dumppath == "" {
log.Fatal("no --download and no dumppath specified (try --help)")
}
f, err := open(*dumppath)
check()
defer f.Close()
log.Printf("Creating database at %s", *dbpath)
db, err := storage.MakeDB(*dbpath, true,
&storage.Settings{*dumppath, uint(*maxNGram)})
check()
// The numbers here are completely arbitrary.
nworkers := runtime.GOMAXPROCS(0)
articles := make(chan *wikidump.Page, 10*nworkers)
links := make(chan map[wikidump.Link]int, 10*nworkers)
redirects := make(chan *wikidump.Redirect, 100)
var wg sync.WaitGroup
// Collect redirects.
wg.Add(1)
redirmap := make(map[string]string)
go func() {
for r := range redirects {
redirmap[r.Title] = r.Target
}
wg.Done()
}()
// Clean up and tokenize articles, extract links, count n-grams.
maxN := int(*maxNGram)
counters := make([]*countmin.Sketch, nworkers)
var worker sync.WaitGroup
worker.Add(nworkers)
log.Printf("%d workers", nworkers)
for i := 0; i < nworkers; i++ {
counters[i], err = countmin.New(int(*nrows), int(*ncols))
check()
go func(ngramcount *countmin.Sketch) {
for a := range articles {
text := wikidump.Cleanup(a.Text)
links <- wikidump.ExtractLinks(text)
tokens := nlp.Tokenize(text)
for _, h := range hash.NGrams(tokens, 1, maxN) {
ngramcount.Add1(h)
}
}
worker.Done()
}(counters[i])
}
wg.Add(1)
go func() {
worker.Wait()
close(links)
for i := 1; i < nworkers; i++ {
counters[0].Sum(counters[i])
}
counters = counters[:1]
wg.Done()
}()
// Collect links and store them in the database.
wg.Add(1)
done := make(chan struct{})
go func() {
if slerr := storeLinks(db, links, maxN); slerr != nil {
panic(slerr)
}
wg.Done()
}()
go wikidump.GetPages(f, articles, redirects)
wg.Wait()
close(done)
log.Printf("Processing %d redirects", len(redirmap))
storage.ProcessRedirects(db, redirmap, true)
err = storage.StoreCM(db, counters[0])
check()
log.Println("Finalizing database")
err = storage.Finalize(db)
check()
err = db.Close()
check()
}
func realMain(dbpath, dumppath, download string, nrows, ncols, maxNGram int,
logger *log.Logger) {
var err error
check := func() {
if err != nil {
panic(err)
}
}
if download != "" {
dumppath, err = wikidump.Download(download, dumppath, true)
check()
} else if dumppath == "" {
panic("no --download and no dumppath specified (try --help)")
}
f, err := open(dumppath)
check()
defer f.Close()
logger.Printf("Creating database at %s", dbpath)
db, err := storage.MakeDB(dbpath, true,
&storage.Settings{Dumpname: dumppath, MaxNGram: uint(maxNGram)})
check()
// The numbers here are completely arbitrary.
nworkers := runtime.GOMAXPROCS(0)
articles := make(chan *wikidump.Page, 10*nworkers)
linkch := make(chan *processedLink, 10*nworkers)
redirch := make(chan *wikidump.Redirect, 10*nworkers)
// Clean up and tokenize articles, extract links, count n-grams.
counters := make(chan *countmin.Sketch, nworkers)
counterTotal, err := countmin.New(nrows, ncols)
check()
go wikidump.GetPages(f, articles, redirch)
logger.Printf("processing dump with %d workers", nworkers)
var narticles uint32
for i := 0; i < nworkers; i++ {
// These signal completion by sending on counters.
go func() {
counters <- processPages(articles, linkch, &narticles,
nrows, ncols, maxNGram)
}()
}
var wg sync.WaitGroup
wg.Add(1)
go func() {
for i := 0; i < nworkers; i++ {
counterTotal.Sum(<-counters)
}
close(counters) // Force panic for programmer error.
close(linkch) // We know the workers are done now.
wg.Done()
}()
// Collect redirects. We store these in nworkers slices to avoid having
// to copy them into a single structure.
// The allRedirects channel MUST be buffered.
wg.Add(nworkers)
allRedirects := make(chan []wikidump.Redirect, nworkers)
var nredirs uint32
for i := 0; i < nworkers; i++ {
go func() {
slice := collectRedirects(redirch)
atomic.AddUint32(&nredirs, uint32(len(slice)))
allRedirects <- slice
wg.Done()
}()
}
go pageProgress(&narticles, logger, &wg)
err = storeLinks(db, linkch)
wg.Wait()
close(allRedirects)
// Check error from storeLinks now, after goroutines have stopped.
check()
logger.Printf("Processing redirects")
bar := pb.StartNew(int(nredirs))
for slice := range allRedirects {
err = storage.StoreRedirects(db, slice, bar)
check()
}
bar.Finish()
err = storage.StoreCM(db, counterTotal)
check()
logger.Println("Finalizing database")
err = storage.Finalize(db)
check()
err = db.Close()
check()
}
