func FetchAndDecodeJSON(r *http.Request, surl, knownProtocol string, lg loghttp.FuncBufUniv, fs fsi.FileSystem) []repo.FullArticle {
fullURL := fmt.Sprintf("%s%s?%s=%s&cnt=%v&prot=%v", routes.AppHost(), routes.FetchSimilarURI,
routes.URLParamKey, surl, numTotal-1, knownProtocol)
// fullURL = fmt.Sprintf("%s%s?%s=%s&cnt=%v", r.URL.Host, repo.routes.FetchSimilarURI,
// routes.URLParamKey, surl, numTotal-1)
lg("lo fetching %v", fullURL)
start := time.Now()
fo := fetch.Options{}
fo.URL = fullURL
bJSON, inf, err := fetch.UrlGetter(r, fo)
_ = inf
lg(err)
if err != nil {
lg("msg %v", inf.Msg)
return nil
}
if len(bJSON) == 0 {
lg("empty bJSON")
return nil
}
lg("\t\tfetch resp complete after %4.2v secs; %vkB", time.Now().Sub(start).Seconds(), len(bJSON)/1024)
var mp map[string][]byte
err = json.Unmarshal(bJSON, &mp)
lg(err)
if err != nil {
if _, ok := mp["msg"]; ok {
lg("%s", mp["msg"])
} else {
lg("%s", bJSON)
}
return nil
}
smaxFound := string(mp["lensimilar"])
maxFound := util.Stoi(smaxFound)
if maxFound < numTotal-1 {
lg("not enough files returned by FetchSimilar 1 - mp[lensimilar] too small: %s", mp["lensimilar"])
return nil
}
least3Files := make([]repo.FullArticle, maxFound+1)
_, ok1 := mp["url_self"]
_, ok2 := mp["mod_self"]
_, ok3 := mp["bod_self"]
if ok1 && ok2 && ok3 {
least3Files[0].Url = string(mp["url_self"])
least3Files[0].Mod, err = time.Parse(http.TimeFormat, string(mp["mod_self"]))
lg(err)
least3Files[0].Body = mp["bod_self"]
if len(least3Files[0].Body) < 200 {
if !bytes.Contains(least3Files[0].Body, []byte(fetch.MsgNoRdirects)) {
lg("found base but its a redirect")
return nil
}
}
}
lg("found base")
for k, v := range mp {
if k == "msg" {
continue
}
if strings.HasSuffix(k, "self") {
continue
}
if strings.HasPrefix(k, "url__") {
sval := strings.TrimPrefix(k, "url__")
val := util.Stoi(sval)
// lg("%v %v %s", sval, val, v)
least3Files[val+1].Url = string(v)
}
if strings.HasPrefix(k, "mod__") {
sval := strings.TrimPrefix(k, "mod__")
val := util.Stoi(sval)
// lg("%v %v %s", sval, val, v)
least3Files[val+1].Mod, err = time.Parse(http.TimeFormat, string(v))
lg(err)
}
if strings.HasPrefix(k, "bod__") {
sval := strings.TrimPrefix(k, "bod__")
val := util.Stoi(sval)
least3Files[val+1].Body = v //html.EscapeString(string(v)
}
}
lg("found %v similar; decoding complete after %4.2v secs", maxFound, time.Now().Sub(start).Seconds())
for _, v := range least3Files {
lg("%v %v", v.Url, len(v.Body))
}
return least3Files
}
func DomClean(b []byte, opt CleaningOptions) (*html.Node, error) {
lg, lge := loghttp.Logger(nil, nil)
_ = lg
b = globFixes(b)
doc, err := html.Parse(bytes.NewReader(b))
if err != nil {
lge(err)
return nil, err
}
if opt.FNamer != nil {
osutilpb.Dom2File(opt.FNamer()+".html", doc)
}
//
//
cleanseDom(doc, 0)
removeCommentsAndIntertagWhitespace(NdX{doc, 0})
fileDump(doc, opt.FNamer)
//
//
condenseTopDown(doc, 0, 0)
removeEmptyNodes(doc, 0)
fileDump(doc, opt.FNamer)
//
//
removeCommentsAndIntertagWhitespace(NdX{doc, 0}) // prevent spacey textnodes around singl child images
breakoutImagesFromAnchorTrees(doc)
recurseImg2Link(doc)
fileDump(doc, opt.FNamer)
//
//
condenseBottomUpV3(doc, 0, 7, map[string]bool{"div": true})
condenseBottomUpV3(doc, 0, 6, map[string]bool{"div": true})
condenseBottomUpV3(doc, 0, 5, map[string]bool{"div": true})
condenseBottomUpV3(doc, 0, 4, map[string]bool{"div": true})
condenseTopDown(doc, 0, 0)
removeEmptyNodes(doc, 0)
removeEmptyNodes(doc, 0)
fileDump(doc, opt.FNamer)
//
//
if opt.Proxify {
if opt.ProxyHost == "" {
opt.ProxyHost = routes.AppHost()
}
proxify(doc, opt.ProxyHost, &url.URL{Scheme: "http", Host: opt.RemoteHost})
fileDump(doc, opt.FNamer)
}
if opt.Beautify {
removeCommentsAndIntertagWhitespace(NdX{doc, 0})
reIndent(doc, 0)
}
//
//
if opt.AddOutline {
addOutlineAttr(doc, 0, []int{0})
}
if opt.AddID {
addIdAttr(doc, 0, 1)
}
if opt.AddOutline || opt.AddID {
fileDump(doc, opt.FNamer)
}
//
computeXPathStack(doc, 0)
if opt.FNamer != nil {
osutilpb.Bytes2File(opt.FNamer()+".txt", xPathDump)
}
return doc, nil
}
"github.com/pbberlin/tools/net/http/fetch"
"github.com/pbberlin/tools/net/http/fileserver"
"github.com/pbberlin/tools/net/http/loghttp"
"github.com/pbberlin/tools/net/http/repo"
"github.com/pbberlin/tools/net/http/routes"
"github.com/pbberlin/tools/sort/sortmap"
"github.com/pbberlin/tools/stringspb"
)
const numTotal = 3 // comparable html docs
const stageMax = 3 // weedstages
const cTestHostOwn = "localhost:63222"
var hostWithPref = routes.AppHost() + repo.UriMountNameY
func prepare(t *testing.T) aetest.Context {
lg, lge := loghttp.Logger(nil, nil)
_ = lg
c, err := aetest.NewContext(nil)
if err != nil {
lge(err)
t.Fatal(err)
}
serveFile := func(w http.ResponseWriter, r *http.Request, m map[string]interface{}) {
fs1 := repo.GetFS(c)
fileserver.FsiFileServer(w, r, fileserver.Options{FS: fs1, Prefix: repo.UriMountNameY})
var docRoot = "" // no relative path, 'cause working dir too flippant
var whichType = 0 // which type of filesystem, default is dsfs
var memMapFileSys = memfs.New(memfs.DirSort("byDateDesc")) // package variable required as "persistence"
var httpFSys = &httpfs.HttpFs{SourceFs: fsi.FileSystem(memMapFileSys)} // memMap is always ready
var fileserver1 = http.FileServer(httpFSys.Dir(docRoot))
const mountName = "mntftch"
const uriSetType = "/fetch/set-fs-type"
const UriMountNameY = "/" + mountName + "/serve-file/"
const uriFetchCommandReceiver = "/fetch/command-receive"
const uriFetchCommandSender = "/fetch/command-send"
var RepoURL = routes.AppHost() + UriMountNameY
var msg = []byte(`<p>This is an embedded static http server.</p>
<p>
It serves previously downloaded pages<br>
i.e. from handelsblatt or economist.
</p>`)
var testCommands = []FetchCommand{
FetchCommand{
Host: "www.handelsblatt.com",
SearchPrefix: "/politik/deutschland/aa/bb",
},
FetchCommand{
Host: "www.handelsblatt.com",
SearchPrefix: "/politik/international/aa/bb",