// parse is a helper to just get a URL object from a string we know is a safe
// url (ParseURL requires us to deal with potential errors)
func parse(ref string) *walker.URL {
u, err := walker.ParseURL(ref)
if err != nil {
panic("Failed to parse walker.URL: " + ref)
}
return u
}
func (ds *CqlModel) ListLinkHistorical(linkUrl string, seedIndex int, limit int) ([]LinkInfo, int, error) {
if limit <= 0 {
return nil, seedIndex, fmt.Errorf("Bad value for limit parameter %d", limit)
}
db := ds.Db
u, err := walker.ParseURL(linkUrl)
if err != nil {
return nil, seedIndex, err
}
query := `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ? AND subdom = ? AND path = ? AND proto = ?`
tld1, err := u.ToplevelDomainPlusOne()
if err != nil {
return nil, seedIndex, err
}
subtld1, err := u.Subdomain()
if err != nil {
return nil, seedIndex, err
}
itr := db.Query(query, tld1, subtld1, u.RequestURI(), u.Scheme).Iter()
var linfos []LinkInfo
var dom, sub, path, prot, getError string
var crawlTime time.Time
var status int
var robotsExcluded bool
count := 0
for itr.Scan(&dom, &sub, &path, &prot, &crawlTime, &status, &getError, &robotsExcluded) {
if count < seedIndex {
count++
continue
}
url, _ := walker.CreateURL(dom, sub, path, prot, crawlTime)
linfo := LinkInfo{
Url: url.String(),
Status: status,
Error: getError,
RobotsExcluded: robotsExcluded,
CrawlTime: crawlTime,
}
linfos = append(linfos, linfo)
if len(linfos) >= limit {
break
}
}
err = itr.Close()
return linfos, seedIndex + len(linfos), err
}
func TestSeedCommand(t *testing.T) {
u, _ := walker.ParseURL("http://test.com")
datastore := &MockDatastore{}
datastore.On("StoreParsedURL", u, mock.AnythingOfType("*walker.FetchResults")).Return("")
cmd.Datastore(datastore)
orig := os.Args
defer func() { os.Args = orig }()
os.Args = []string{os.Args[0], "seed", "--url=" + u.String()}
go func() {
time.Sleep(5 * time.Millisecond)
syscall.Kill(os.Getpid(), syscall.SIGINT)
}()
cmd.Execute()
datastore.AssertExpectations(t)
}
func (ds *CqlModel) FindLink(link string) (*LinkInfo, error) {
db := ds.Db
u, err := walker.ParseURL(link)
if err != nil {
return nil, err
}
query := `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ? AND
subdom = ? AND
path = ? AND
proto = ?`
tld1, err := u.ToplevelDomainPlusOne()
if err != nil {
return nil, err
}
subtld1, err := u.Subdomain()
if err != nil {
return nil, err
}
itr := db.Query(query, tld1, subtld1, u.RequestURI(), u.Scheme).Iter()
rtimes := map[string]rememberTimes{}
linfos, err := ds.collectLinkInfos(nil, rtimes, itr, 1)
if err != nil {
itr.Close()
return nil, err
}
err = itr.Close()
if err != nil {
return nil, err
}
if len(linfos) == 0 {
return nil, nil
} else {
return &linfos[0], nil
}
}
func TestURLTLD(t *testing.T) {
for _, dt := range tldtests {
u, err := walker.ParseURL(dt.URL)
if err != nil {
if !dt.ErrorExpected {
t.Errorf("Did not expect error parsing %v: %v", dt.URL, err)
}
continue
}
dom, err := u.ToplevelDomainPlusOne()
if err != nil && !dt.ErrorExpected {
t.Errorf("Did not expect error getting TLD+1: %v", err)
}
if dom != dt.ExpectedTLDPlusOne {
t.Errorf("Expected ToplevelDomainPlusOne to be %v\nBut got: %v",
dt.ExpectedTLDPlusOne, dom)
}
subdom, err := u.Subdomain()
if err != nil && !dt.ErrorExpected {
t.Errorf("Did not expect error getting subdomain: %v", err)
}
if subdom != dt.ExpectedSubdomain {
t.Errorf("Expected Subdomain to be %v\nBut got: %v",
dt.ExpectedSubdomain, subdom)
}
dom2, subdom2, err := u.TLDPlusOneAndSubdomain()
if err != nil && !dt.ErrorExpected {
t.Errorf("Did not expect error getting TLD+1 and subdomain: %v", err)
}
if dom2 != dt.ExpectedTLDPlusOne {
t.Errorf("Expected TLDPlusOneAndSubdomain to give domain %v\nBut got: %v",
dt.ExpectedTLDPlusOne, dom2)
}
if subdom2 != dt.ExpectedSubdomain {
t.Errorf("Expected TLDPlusOneAndSubdomain to give subdomain %v\nBut got: %v",
dt.ExpectedSubdomain, subdom2)
}
}
}
func TestURLCreation(t *testing.T) {
url1, err := url.Parse("http://sub1.test.com/thepath?query=blah")
if err != nil {
t.Fatal(err)
}
wurl1, err := walker.ParseURL("http://sub1.test.com/thepath?query=blah")
if err != nil {
t.Fatal(err)
}
if url1.String() != wurl1.String() {
t.Errorf("URLs should be the same: %v\nAnd: %v")
}
created, err := walker.CreateURL("test.com", "sub1", "thepath?query=blah", "http",
walker.NotYetCrawled)
if err != nil {
t.Fatal(err)
}
if created.String() != wurl1.String() {
t.Errorf("Expected CreateURL to return %v\nBut got: %v", wurl1, created)
}
}
func (ds *CqlModel) ListLinks(domain string, seedUrl string, limit int) ([]LinkInfo, error) {
if limit <= 0 {
return nil, fmt.Errorf("Bad value for limit parameter %d", limit)
}
db := ds.Db
var linfos []LinkInfo
rtimes := map[string]rememberTimes{}
var table []queryEntry
if seedUrl == "" {
table = []queryEntry{
queryEntry{
query: `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ?`,
args: []interface{}{domain},
},
}
} else {
u, err := walker.ParseURL(seedUrl)
if err != nil {
return linfos, err
}
dom, err := u.ToplevelDomainPlusOne()
if err != nil {
return linfos, err
}
sub, err := u.Subdomain()
if err != nil {
return linfos, err
}
pat := u.RequestURI()
pro := u.Scheme
table = []queryEntry{
queryEntry{
query: `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ? AND
subdom = ? AND
path = ? AND
proto > ?`,
args: []interface{}{dom, sub, pat, pro},
},
queryEntry{
query: `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ? AND
subdom = ? AND
path > ?`,
args: []interface{}{dom, sub, pat},
},
queryEntry{
query: `SELECT dom, subdom, path, proto, time, stat, err, robot_ex
FROM links
WHERE dom = ? AND
subdom > ?`,
args: []interface{}{dom, sub},
},
}
}
var err error
for _, qt := range table {
itr := db.Query(qt.query, qt.args...).Iter()
linfos, err = ds.collectLinkInfos(linfos, rtimes, itr, limit)
if err != nil {
return linfos, err
}
err = itr.Close()
if err != nil {
return linfos, err
} else if len(linfos) >= limit {
return linfos, nil
}
}
return linfos, nil
}
//NOTE: InsertLinks should try to insert as much information as possible
//return errors for things it can't handle
func (ds *CqlModel) InsertLinks(links []string) []error {
//
// Collect domains
//
var domains []string
var errList []error
var urls []*walker.URL
for i := range links {
link := links[i]
url, err := walker.ParseURL(link)
if err != nil {
errList = append(errList, fmt.Errorf("%v # ParseURL: %v", link, err))
domains = append(domains, "")
urls = append(urls, nil)
continue
} else if url.Scheme == "" {
errList = append(errList, fmt.Errorf("%v # ParseURL: undefined scheme (http:// or https://)", link))
domains = append(domains, "")
urls = append(urls, nil)
continue
}
domain, err := url.ToplevelDomainPlusOne()
if err != nil {
errList = append(errList, fmt.Errorf("%v # ToplevelDomainPlusOne: bad domain: %v", link, err))
domains = append(domains, "")
urls = append(urls, nil)
continue
}
domains = append(domains, domain)
urls = append(urls, url)
}
//
// Push domain information to table. The only trick to this, is I don't add links unless
// the domain can be added
//
db := ds.Db
var seen = map[string]bool{}
for i := range links {
link := links[i]
d := domains[i]
u := urls[i]
// if you already had an error, keep going
if u == nil {
continue
}
if !seen[d] {
err := ds.addDomainIfNew(d)
if err != nil {
errList = append(errList, fmt.Errorf("%v # addDomainIfNew: %v", link, err))
continue
}
}
seen[d] = true
subdom, err := u.Subdomain()
if err != nil {
errList = append(errList, fmt.Errorf("%v # Subdomain(): %v", link, err))
continue
}
err = db.Query(`INSERT INTO links (dom, subdom, path, proto, time)
VALUES (?, ?, ?, ?, ?)`, d, subdom,
u.RequestURI(), u.Scheme, walker.NotYetCrawled).Exec()
if err != nil {
errList = append(errList, fmt.Errorf("%v # `insert query`: %v", link, err))
continue
}
}
return errList
}
func init() {
walkerCommand := &cobra.Command{
Use: "walker",
}
var config string
walkerCommand.PersistentFlags().StringVarP(&config,
"config", "c", "", "path to a config file to load")
readConfig := func() {
if config != "" {
if err := walker.ReadConfigFile(config); err != nil {
panic(err.Error())
}
}
}
var noConsole bool = false
crawlCommand := &cobra.Command{
Use: "crawl",
Short: "start an all-in-one crawler",
Run: func(cmd *cobra.Command, args []string) {
readConfig()
if commander.Datastore == nil {
ds, err := walker.NewCassandraDatastore()
if err != nil {
fatalf("Failed creating Cassandra datastore: %v", err)
}
commander.Datastore = ds
commander.Dispatcher = &walker.CassandraDispatcher{}
}
if commander.Handler == nil {
commander.Handler = &walker.SimpleWriterHandler{}
}
manager := &walker.FetchManager{
Datastore: commander.Datastore,
Handler: commander.Handler,
}
go manager.Start()
if commander.Dispatcher != nil {
go func() {
err := commander.Dispatcher.StartDispatcher()
if err != nil {
panic(err.Error())
}
}()
}
if !noConsole {
console.Start()
}
sig := make(chan os.Signal)
signal.Notify(sig, syscall.SIGINT)
<-sig
if commander.Dispatcher != nil {
commander.Dispatcher.StopDispatcher()
}
manager.Stop()
},
}
crawlCommand.Flags().BoolVarP(&noConsole, "no-console", "C", false, "Do not start the console")
walkerCommand.AddCommand(crawlCommand)
fetchCommand := &cobra.Command{
Use: "fetch",
Short: "start only a walker fetch manager",
Run: func(cmd *cobra.Command, args []string) {
readConfig()
if commander.Datastore == nil {
ds, err := walker.NewCassandraDatastore()
if err != nil {
fatalf("Failed creating Cassandra datastore: %v", err)
}
commander.Datastore = ds
commander.Dispatcher = &walker.CassandraDispatcher{}
}
if commander.Handler == nil {
commander.Handler = &walker.SimpleWriterHandler{}
}
manager := &walker.FetchManager{
Datastore: commander.Datastore,
Handler: commander.Handler,
}
go manager.Start()
sig := make(chan os.Signal)
signal.Notify(sig, syscall.SIGINT)
<-sig
manager.Stop()
},
}
walkerCommand.AddCommand(fetchCommand)
dispatchCommand := &cobra.Command{
Use: "dispatch",
Short: "start only a walker dispatcher",
Run: func(cmd *cobra.Command, args []string) {
readConfig()
if commander.Dispatcher == nil {
commander.Dispatcher = &walker.CassandraDispatcher{}
}
go func() {
err := commander.Dispatcher.StartDispatcher()
if err != nil {
panic(err.Error())
}
}()
sig := make(chan os.Signal)
signal.Notify(sig, syscall.SIGINT)
<-sig
commander.Dispatcher.StopDispatcher()
},
}
walkerCommand.AddCommand(dispatchCommand)
var seedURL string
seedCommand := &cobra.Command{
Use: "seed",
Short: "add a seed URL to the datastore",
Long: `Seed is useful for:
- Adding starter links to bootstrap a broad crawl
- Adding links when add_new_domains is false
- Adding any other link that needs to be crawled soon
This command will insert the provided link and also add its domain to the
crawl, regardless of the add_new_domains configuration setting.`,
Run: func(cmd *cobra.Command, args []string) {
readConfig()
orig := walker.Config.AddNewDomains
defer func() { walker.Config.AddNewDomains = orig }()
walker.Config.AddNewDomains = true
if seedURL == "" {
fatalf("Seed URL needed to execute; add on with --url/-u")
}
u, err := walker.ParseURL(seedURL)
if err != nil {
fatalf("Could not parse %v as a url: %v", seedURL, err)
}
if commander.Datastore == nil {
ds, err := walker.NewCassandraDatastore()
if err != nil {
fatalf("Failed creating Cassandra datastore: %v", err)
}
commander.Datastore = ds
}
commander.Datastore.StoreParsedURL(u, nil)
},
}
seedCommand.Flags().StringVarP(&seedURL, "url", "u", "", "URL to add as a seed")
walkerCommand.AddCommand(seedCommand)
var outfile string
schemaCommand := &cobra.Command{
Use: "schema",
Short: "output the walker schema",
Long: `Schema prints the walker schema to stdout, substituting
schema-relevant configuration items (ex. keyspace, replication factor).
Useful for something like:
$ <edit walker.yaml as desired>
$ walker schema -o schema.cql
$ <edit schema.cql further as desired>
$ cqlsh -f schema.cql
`,
Run: func(cmd *cobra.Command, args []string) {
readConfig()
if outfile == "" {
fatalf("An output file is needed to execute; add with --out/-o")
}
out, err := os.Create(outfile)
if err != nil {
panic(err.Error())
}
defer out.Close()
schema, err := walker.GetCassandraSchema()
if err != nil {
panic(err.Error())
}
fmt.Fprint(out, schema)
},
}
schemaCommand.Flags().StringVarP(&outfile, "out", "o", "", "File to write output to")
walkerCommand.AddCommand(schemaCommand)
consoleCommand := &cobra.Command{
Use: "console",
Short: "Start up the walker console",
Run: func(cmd *cobra.Command, args []string) {
readConfig()
console.Run()
},
}
walkerCommand.AddCommand(consoleCommand)
commander.Command = walkerCommand
}
