func main() {
var bigtableClientConnData = gbigtable.ClientConnectionData{
Project: "bigdatagarage",
Zone: "us-central1-c",
Cluster: "workshopanalytics",
KeyJsonFilePath: KeyJsonFilePath,
}
// https://godoc.org/google.golang.org/cloud/bigtable#RowRange
// rr := bigtable.PrefixRange("my")
rr := bigtable.InfiniteRange("")
// rr := bigtable.NewRange("begin", "end")
// rr := bigtable.SingleRow("mykey01")
printFilteredRowsFromTable(bigtableClientConnData, "test-tmp", rr, nil)
// https://godoc.org/google.golang.org/cloud/bigtable#Filter
var opts []bigtable.ReadOption
// opts = append(opts, bigtable.RowFilter(bigtable.FamilyFilter(".*fam01.*")));
// opts = append(opts, bigtable.RowFilter(bigtable.ColumnFilter(".*col.*")));
opts = append(opts, bigtable.RowFilter(bigtable.ColumnFilter(".*col02.*")))
// opts = append(opts, bigtable.RowFilter(bigtable.ChainFilters(bigtable.ColumnFilter(".*col.*"), bigtable.ValueFilter(".*val.*"))));
printFilteredRowsFromTable(bigtableClientConnData, "test-tmp", (bigtable.RowRange{}), opts)
}
// rebuildTable deletes the table if it exists, then creates the table, with the index column family.
func rebuildTable() error {
ctx, _ := context.WithTimeout(context.Background(), 5*time.Minute)
adminClient.DeleteTable(ctx, *tableName)
if err := adminClient.CreateTable(ctx, *tableName); err != nil {
return fmt.Errorf("CreateTable: %v", err)
}
time.Sleep(20 * time.Second)
if err := adminClient.CreateColumnFamily(ctx, *tableName, indexColumnFamily); err != nil {
return fmt.Errorf("CreateColumnFamily: %v", err)
}
if err := adminClient.CreateColumnFamily(ctx, *tableName, contentColumnFamily); err != nil {
return fmt.Errorf("CreateColumnFamily: %v", err)
}
// Open the prototype table. It contains a number of documents to get started with.
prototypeTable := client.Open(prototypeTableName)
var (
writeErr error // Set if any write fails.
mu sync.Mutex // Protects writeErr
wg sync.WaitGroup // Used to wait for all writes to finish.
)
copyRowToTable := func(row bigtable.Row) bool {
mu.Lock()
failed := writeErr != nil
mu.Unlock()
if failed {
return false
}
mut := bigtable.NewMutation()
for family, items := range row {
for _, item := range items {
// Get the column name, excluding the column family name and ':' character.
columnWithoutFamily := item.Column[len(family)+1:]
mut.Set(family, columnWithoutFamily, bigtable.Now(), item.Value)
}
}
wg.Add(1)
go func() {
// TODO: should use a semaphore to limit the number of concurrent writes.
if err := table.Apply(ctx, row.Key(), mut); err != nil {
mu.Lock()
writeErr = err
mu.Unlock()
}
wg.Done()
}()
return true
}
// Create a filter that only accepts the column families we're interested in.
filter := bigtable.FamilyFilter(indexColumnFamily + "|" + contentColumnFamily)
// Read every row from prototypeTable, and call copyRowToTable to copy it to our table.
err := prototypeTable.ReadRows(ctx, bigtable.InfiniteRange(""), copyRowToTable, bigtable.RowFilter(filter))
wg.Wait()
if err != nil {
return err
}
return writeErr
}
// copyTable copies data from one table to another.
func copyTable(src, dst string, client *bigtable.Client, adminClient *bigtable.AdminClient) error {
if src == "" || src == dst {
return nil
}
ctx, _ := context.WithTimeout(context.Background(), time.Minute)
// Open the source and destination tables.
srcTable := client.Open(src)
dstTable := client.Open(dst)
var (
writeErr error // Set if any write fails.
mu sync.Mutex // Protects writeErr
wg sync.WaitGroup // Used to wait for all writes to finish.
)
copyRowToTable := func(row bigtable.Row) bool {
mu.Lock()
failed := writeErr != nil
mu.Unlock()
if failed {
return false
}
mut := bigtable.NewMutation()
for family, items := range row {
for _, item := range items {
// Get the column name, excluding the column family name and ':' character.
columnWithoutFamily := item.Column[len(family)+1:]
mut.Set(family, columnWithoutFamily, bigtable.Now(), item.Value)
}
}
wg.Add(1)
go func() {
// TODO: should use a semaphore to limit the number of concurrent writes.
if err := dstTable.Apply(ctx, row.Key(), mut); err != nil {
mu.Lock()
writeErr = err
mu.Unlock()
}
wg.Done()
}()
return true
}
// Create a filter that only accepts the column families we're interested in.
filter := bigtable.FamilyFilter(indexColumnFamily + "|" + contentColumnFamily)
// Read every row from srcTable, and call copyRowToTable to copy it to our table.
err := srcTable.ReadRows(ctx, bigtable.InfiniteRange(""), copyRowToTable, bigtable.RowFilter(filter))
wg.Wait()
if err != nil {
return err
}
return writeErr
}
func doCount(ctx context.Context, args ...string) {
if len(args) != 1 {
log.Fatal("usage: cbt count <table>")
}
tbl := getClient().Open(args[0])
n := 0
err := tbl.ReadRows(ctx, bigtable.InfiniteRange(""), func(_ bigtable.Row) bool {
n++
return true
}, bigtable.RowFilter(bigtable.StripValueFilter()))
if err != nil {
log.Fatalf("Reading rows: %v", err)
}
fmt.Println(n)
}
func main() {
var err error
config, err = cbtrc.Load()
if err != nil {
log.Fatal(err)
}
config.RegisterFlags()
flag.Parse()
if err := config.CheckFlags(); err != nil {
log.Fatal(err)
}
if config.Creds != "" {
os.Setenv("GOOGLE_APPLICATION_CREDENTIALS", config.Creds)
}
if flag.NArg() != 0 {
flag.Usage()
os.Exit(1)
}
log.Printf("Dialing connections...")
client, err = bigtable.NewClient(context.Background(), config.Project, config.Zone, config.Cluster)
if err != nil {
log.Fatalf("Making bigtable.Client: %v", err)
}
defer client.Close()
adminClient, err = bigtable.NewAdminClient(context.Background(), config.Project, config.Zone, config.Cluster)
if err != nil {
log.Fatalf("Making bigtable.AdminClient: %v", err)
}
defer adminClient.Close()
// Create a scratch table.
log.Printf("Setting up scratch table...")
if err := adminClient.CreateTable(context.Background(), *scratchTable); err != nil {
log.Fatalf("Making scratch table %q: %v", *scratchTable, err)
}
if err := adminClient.CreateColumnFamily(context.Background(), *scratchTable, "f"); err != nil {
log.Fatalf("Making scratch table column family: %v", err)
}
// Upon a successful run, delete the table. Don't bother checking for errors.
defer adminClient.DeleteTable(context.Background(), *scratchTable)
log.Printf("Starting load test... (run for %v)", *runFor)
tbl := client.Open(*scratchTable)
sem := make(chan int, 100) // limit the number of requests happening at once
var reads, writes stats
stopTime := time.Now().Add(*runFor)
var wg sync.WaitGroup
for time.Now().Before(stopTime) {
sem <- 1
wg.Add(1)
go func() {
defer wg.Done()
defer func() { <-sem }()
ok := true
opStart := time.Now()
var stats *stats
defer func() {
stats.Record(ok, time.Since(opStart))
}()
row := fmt.Sprintf("row%d", rand.Intn(100)) // operate on 1 of 100 rows
switch rand.Intn(10) {
default:
// read
stats = &reads
_, err := tbl.ReadRow(context.Background(), row, bigtable.RowFilter(bigtable.LatestNFilter(1)))
if err != nil {
log.Printf("Error doing read: %v", err)
ok = false
}
case 0, 1, 2, 3, 4:
// write
stats = &writes
mut := bigtable.NewMutation()
mut.Set("f", "col", bigtable.Now(), bytes.Repeat([]byte("0"), 1<<10)) // 1 KB write
if err := tbl.Apply(context.Background(), row, mut); err != nil {
log.Printf("Error doing mutation: %v", err)
ok = false
}
}
}()
}
wg.Wait()
log.Printf("Reads (%d ok / %d tries):\n%v", reads.ok, reads.tries, newAggregate(reads.ds))
log.Printf("Writes (%d ok / %d tries):\n%v", writes.ok, writes.tries, newAggregate(writes.ds))
}
// handleSearch responds to search queries, returning links and snippets for matching documents.
func handleSearch(w http.ResponseWriter, r *http.Request, table *bigtable.Table) {
ctx, _ := context.WithTimeout(context.Background(), 10*time.Second)
query := r.FormValue("q")
// Split the query into words.
words := tokenize(query)
if len(words) == 0 {
http.Error(w, "Empty query.", http.StatusBadRequest)
return
}
// readRows reads from many rows concurrently.
readRows := func(rows []string) ([]bigtable.Row, error) {
results := make([]bigtable.Row, len(rows))
errors := make([]error, len(rows))
var wg sync.WaitGroup
for i, row := range rows {
wg.Add(1)
go func(i int, row string) {
defer wg.Done()
results[i], errors[i] = table.ReadRow(ctx, row, bigtable.RowFilter(bigtable.LatestNFilter(1)))
}(i, row)
}
wg.Wait()
for _, err := range errors {
if err != nil {
return nil, err
}
}
return results, nil
}
// For each query word, get the list of documents containing it.
results, err := readRows(words)
if err != nil {
http.Error(w, "Error reading index: "+err.Error(), http.StatusInternalServerError)
return
}
// Count how many of the query words each result contained.
hits := make(map[string]int)
for _, r := range results {
for _, r := range r[indexColumnFamily] {
hits[r.Column]++
}
}
// Build a slice of all the documents that matched every query word.
var matches []string
for doc, count := range hits {
if count == len(words) {
matches = append(matches, doc[len(indexColumnFamily+":"):])
}
}
// Fetch the content of those documents from the Bigtable.
content, err := readRows(matches)
if err != nil {
http.Error(w, "Error reading results: "+err.Error(), http.StatusInternalServerError)
return
}
type result struct{ Title, Snippet string }
data := struct {
Query string
Results []result
}{query, nil}
// Output links and snippets.
for i, doc := range matches {
var text string
c := content[i][contentColumnFamily]
if len(c) > 0 {
text = string(c[0].Value)
}
if len(text) > 100 {
text = text[:100] + "..."
}
data.Results = append(data.Results, result{doc, text})
}
var buf bytes.Buffer
if err := searchTemplate.ExecuteTemplate(&buf, "", data); err != nil {
http.Error(w, "Error executing HTML template: "+err.Error(), http.StatusInternalServerError)
return
}
io.Copy(w, &buf)
}