golden/pdfxform a pdf rasterization server

golden/go/pdfxform/main.go

Index: golden/go/pdfxform/main.go
diff --git a/golden/go/pdfxform/main.go b/golden/go/pdfxform/main.go
new file mode 100644
index 0000000000000000000000000000000000000000..92b93b3e5ce096ee14f9f76d2c17b119b33aa5bd
--- /dev/null
+++ b/golden/go/pdfxform/main.go
@@ -0,0 +1,526 @@
+// pdfxform is a server that rasterizes PDF documents into PNG
+package main
+
+import (
+	"bytes"
+	"crypto/md5"
+	"encoding/hex"
+	"encoding/json"
+	"flag"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"net/http"
+	"os"
+	"os/user"
+	"path"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"time"
+
+	"github.com/skia-dev/glog"
+	"go.skia.org/infra/go/auth"
+	"go.skia.org/infra/go/gs"
+	"go.skia.org/infra/go/pdf"
+	"go.skia.org/infra/go/util"
+	"google.golang.org/api/storage/v1"
+)
+
+////////////////////////////////////////////////////////////////////////////////
+
+const (
+	pngExt = "png"
+	pdfExt = "pdf"
+)
+
+////////////////////////////////////////////////////////////////////////////////
+
+type md5digest [md5.Size]byte

[stephana, 2015/06/26 17:25:06]

I think treating MD5 hashes as byte slices (instead of hex strings) is overkill.
Using strings throughout for MD5s is fine IMO and it avoids having function like
decodeMd5(...). 
This is a largely IO bound process, if CPU matters at all it's how fast the
rasterizers can convert a PDF to PNG. 

[hal.canary, 2015/06/26 18:44:12]

The pdfXformer.results map can grow without constraint as the process lives
longer.  But I'll do it your way until we notice problems.

Done.

+
+func (digest md5digest) String() string {
+	return hex.EncodeToString(digest[:])
+}
+
+// decodeMd5 turns a hexadecimal string into a fixed-size binary array for more efficient storage.
+func decodeMd5(md5Str string) md5digest {
+	md5ByteSlice, err := hex.DecodeString(md5Str)
+	if err != nil || len(md5ByteSlice) != md5.Size {
+		glog.Errorf("Unable to decode base64-encoded MD5: %v: %s", md5Str, err)
+		return md5digest{}
+	}
+	var ret md5digest
+	copy(ret[:], md5ByteSlice)
+	return ret
+}
+
+// md5OfFile calculates the MD5 checksum of a file.
+func md5OfFile(path string) (sum md5digest, err error) {
+	md5 := md5.New()
+	f, err := os.Open(path)
+	if err != nil {
+		return
+	}
+	defer util.Close(f)
+	if _, err = io.Copy(md5, f); err != nil {
+		return
+	}
+	copy(sum[:], md5.Sum(nil))
+	return
+}
+
+////////////////////////////////////////////////////////////////////////////////
+
+// removeIf is like util.Remove, but logs no error if the file does not exist.
+func removeIf(path string) {
+	if err := os.Remove(path); err != nil {
+		if !os.IsNotExist(err) {
+			glog.Errorf("Failed to Remove(%s): %v", path, err)
+		}
+	}
+}
+
+// isPDF returns true if the path appears to point to a PDF file.
+func isPDF(path string) bool {
+	f, err := os.Open(path)
+	if err != nil {
+		return false
+	}
+	defer util.Close(f)
+	var buffer [4]byte
+	if n, err := f.Read(buffer[:]); n != 4 || err != nil {
+		return false
+	}
+	var magic = [4]byte{'%', 'P', 'D', 'F'}
+	return bytes.Equal(magic[:], buffer[:])
+}
+
+// writeTo opens a file and dumps the contents of the reader into it.
+func writeTo(path string, reader *io.ReadCloser) error {
+	defer util.Close(*reader)
+	file, err := os.Create(path)
+	if err == nil {
+		_, err = io.Copy(file, *reader)
+	}
+	return err
+}
+
+// assertNil logs the err and exits if it is not nil.
+func assertNil(err error) {
+	if err != nil {
+		errMsg := ""
+		if _, fileName, line, ok := runtime.Caller(1); ok {
+			errMsg = fmt.Sprintf("-called from: %s:%d", fileName, line)
+		}
+		glog.Fatalf("Unexpected error %s: %s", errMsg, err)
+	}
+}
+
+////////////////////////////////////////////////////////////////////////////////
+
+// The dm struct mirrors the structure of the json files that will be
+// parsed and prooduced.

[stephana, 2015/06/26 17:25:06]

Instead of the defining it's own structures to receive the results. 
It should re-use the ones from the gold ingester (see
golden/go/goldingester/goldingester.go). 

That way any changes in the ingester will be reflected here and help detect
modifications that need
to be made. 

[hal.canary, 2015/06/26 18:44:11]

Done.  Thanks.

+
+type dm struct {
+	Results     []result          `json:"results"`
+	BuildNumber string            `json:"build_number"`
+	GitHash     string            `json:"gitHash"`
+	Key         map[string]string `json:"key"`
+	MaxRSSMB    int               `json:"max_rss_MB"`
+}
+type keyType struct {
+	Config         string `json:"config,omitempty"`
+	Name           string `json:"name,omitempty"`
+	Source_options string `json:"source_options,omitempty"`
+	Source_type    string `json:"source_type,omitempty"`
+	Rasterizer     string `json:"rasterizer,omitempty"`
+}
+type optionsType struct {
+	Ext string `json:"ext,omitempty"`
+}
+type result struct {
+	Key     keyType     `json:"key"`
+	Md5     string      `json:"md5"`
+	Options optionsType `json:"options"`
+}
+
+////////////////////////////////////////////////////////////////////////////////
+
+// storageClient struct is used for uploading to cloud storage
+type storageClient struct {
+	httpClient     *http.Client
+	storageService *storage.Service
+}
+
+// getClient returns an authorized storage.Service and the
+// corresponding http.Client; if anything goes wrong, it logs a fatal
+// error.
+func getClient(cacheFilePath string) storageClient {
+	config := auth.OAuthConfig(cacheFilePath, auth.SCOPE_FULL_CONTROL)
+	client, err := auth.RunFlow(config)
+	assertNil(err)
+	gsService, err := storage.New(client)
+	assertNil(err)
+	return storageClient{httpClient: client, storageService: gsService}
+}
+
+// gsFetch fetch the object's data from google storage
+func gsFetch(object *storage.Object, sc storageClient) (io.ReadCloser, int64, error) {
+	request, err := gs.RequestForStorageURL(object.MediaLink)
+	if err != nil {
+		return nil, -1, err
+	}
+	resp, err := sc.httpClient.Do(request)
+	if err != nil {
+		return nil, -1, err
+	}
+	if resp.StatusCode != 200 {
+		resp.Body.Close()
+		return nil, -1, fmt.Errorf("Failed to retrieve: %s %d %s", object.MediaLink, resp.StatusCode, resp.Status)
+	}
+	return resp.Body, resp.ContentLength, nil
+}
+
+// uploadFile uploads the specified file to the remote dir in Google
+// Storage. It also sets the appropriate ACLs on the uploaded file.
+// If the file already exists on the server, do nothing.
+func uploadFile(sc storageClient, input io.Reader, storageBucket, storagePath, accessControlEntity string) (bool, error) {
+	obj, _ := sc.storageService.Objects.Get(storageBucket, storagePath).Do()
+	if obj != nil {
+		return false, nil // noclobber
+	}
+	fullPath := fmt.Sprintf("gs://%s/%s", storageBucket, storagePath)
+	object := &storage.Object{Name: storagePath}
+	if _, err := sc.storageService.Objects.Insert(storageBucket, object).Media(input).Do(); err != nil {
+		return false, fmt.Errorf("Objects.Insert(%s) failed: %s", fullPath, err)
+	}
+	objectAcl := &storage.ObjectAccessControl{
+		Bucket: storageBucket, Entity: accessControlEntity, Object: storagePath, Role: "READER",
+	}
+	if _, err := sc.storageService.ObjectAccessControls.Insert(storageBucket, storagePath, objectAcl).Do(); err != nil {
+		return false, fmt.Errorf("Could not update ACL of %s: %s", fullPath, err)
+	}
+	return true, nil
+}
+
+////////////////////////////////////////////////////////////////////////////////
+
+// The pdfXformer struct holds state (results, counter) and constants (bucket, directories)
+type pdfXformer struct {
+	client                 storageClient
+	storageBucket          string
+	storageJsonDirectory   string
+	storageImagesDirectory string
+	accessControlEntity    string
+	rasterizers            []pdf.Rasterizer
+	results                map[md5digest]map[int]md5digest
+	counter                int
+	identifier             string
+}
+
+var errorImageMd5 = md5digest{

[stephana, 2015/06/26 17:25:06]

Another argument to use strings instead of md5digest. 

[hal.canary, 2015/06/26 18:44:12]

Done.

+	0x45, 0xAA, 0x8A, 0xF2,
+	0x65, 0xD1, 0x68, 0x39,
+	0x40, 0x25, 0x83, 0xDF,
+	0x57, 0x56, 0xA7, 0xC6,
+} // data/45aa8af265d16839402583df5756a7c6.png
+
+// rasterizeOnce applies a single rastetizer to the given pdf file.
+// If the rasterizer fails, use the errorImage.  If everything
+// succeeds, upload the PNG.
+func (xformer *pdfXformer) rasterizeOnce(pdfPath string, rasterizerIndex int) (md5digest, error) {
+	rasterizer := xformer.rasterizers[rasterizerIndex]
+	tempdir := filepath.Dir(pdfPath)
+	pngPath := path.Join(tempdir, fmt.Sprintf("%s.%s", rasterizer.String(), pngExt))
+	defer removeIf(pngPath)
+	glog.Infof("> > > > rasterizing with %s", rasterizer)
+	err := rasterizer.Rasterize(pdfPath, pngPath)
+	if err != nil {
+		glog.Warningf("rasterizing %s with %s failed: %s", filepath.Base(pdfPath), rasterizer.String(), err)
+		return errorImageMd5, nil
+	}
+	md5, err := md5OfFile(pngPath)
+	if err != nil {
+		return md5digest{}, err
+	}
+	f, err := os.Open(pngPath)
+	if err != nil {
+		return md5digest{}, err
+	}
+	defer util.Close(f)
+	pngUploadPath := fmt.Sprintf("%s/%s.%s", xformer.storageImagesDirectory, md5.String(), pngExt)
+	didUpload, err := uploadFile(xformer.client, f, xformer.storageBucket, pngUploadPath, xformer.accessControlEntity)
+	if err != nil {
+		return md5digest{}, err
+	}
+	if didUpload {
+		glog.Infof("> > > > uploaded %s", pngUploadPath)
+	}
+	return md5, nil
+}
+
+// makeTmpDir returns a nicely-named directory for temp files in $TMPDIR
+func (xformer *pdfXformer) makeTmpDir() string {
+	if xformer.identifier == "" {
+		var host, userName string
+		if h, err := os.Hostname(); err == nil {
+			host = h
+			if i := strings.Index(host, "."); i >= 0 {
+				host = host[:i]
+			}
+		}
+		if currentUser, err := user.Current(); err == nil {
+			userName = currentUser.Username
+		}
+		userName = strings.Replace(userName, `\`, "_", -1)
+		xformer.identifier = fmt.Sprintf("%s.%s.%s.tmp.%d.", filepath.Base(os.Args[0]), host, userName, os.Getpid())
+	}
+	tempdir, err := ioutil.TempDir("", xformer.identifier)
+	assertNil(err)
+	return tempdir
+}
+
+// processResult rasterizes a single PDF result and returns a set of new results.
+func (xformer *pdfXformer) processResult(res result) []result {
+	rasterizedResults := []result{}
+	resultMap, found := xformer.results[decodeMd5(res.Md5)]
+	if found {
+		// Skip rasterizion steps: big win.
+		for index, rasterizer := range xformer.rasterizers {
+			keyCopy := res.Key
+			keyCopy.Rasterizer = rasterizer.String()
+			md5, ok := resultMap[index]
+			if ok {
+				rasterizedResults = append(rasterizedResults, result{Key: keyCopy, Md5: md5.String(), Options: optionsType{pngExt}})
+			} else {
+				glog.Errorf("missing rasterizer %s on %s", rasterizer.String(), res.Md5)
+			}
+		}
+		return rasterizedResults
+	}
+
+	tempdir := xformer.makeTmpDir()
+	defer util.RemoveAll(tempdir)
+	pdfPath := path.Join(tempdir, fmt.Sprintf("%s.pdf", res.Md5))
+	objectName := fmt.Sprintf("%s/%s.pdf", xformer.storageImagesDirectory, res.Md5)
+	storageURL := fmt.Sprintf("gs://%s/%s", xformer.storageBucket, objectName)
+	object, err := xformer.client.storageService.Objects.Get(xformer.storageBucket, objectName).Do()
+	if err != nil {
+		glog.Errorf("unable to find %s: %s", storageURL, err)
+		return []result{}
+	}
+	pdfData, _, err := gsFetch(object, xformer.client)
+	if err != nil {
+		glog.Errorf("unable to retrieve %s: %s", storageURL, err)
+		return []result{}
+	}
+	writeTo(pdfPath, &pdfData)
+	if !isPDF(pdfPath) {
+		glog.Errorf("%s is not a PDF", objectName)
+		return []result{}
+	}
+	resultMap = map[int]md5digest{}
+	for index, rasterizer := range xformer.rasterizers {
+		md5, err := xformer.rasterizeOnce(pdfPath, index)
+		if err != nil {
+			glog.Errorf("rasterizer %s failed on %s.pdf: %s", rasterizer, res.Md5, err)
+			continue
+		}
+		keyCopy := res.Key
+		keyCopy.Rasterizer = rasterizer.String()
+		rasterizedResults = append(rasterizedResults, result{Key: keyCopy, Md5: md5.String(), Options: optionsType{pngExt}})
+		resultMap[index] = md5
+	}
+	xformer.results[decodeMd5(res.Md5)] = resultMap
+	return rasterizedResults
+}
+
+// processJsonFile reads a json file and produces a new json file
+// with rasterized results.
+func (xformer *pdfXformer) processJsonFile(jsonFileObject *storage.Object) {
+	jsonURL := fmt.Sprintf("gs://%s/%s", xformer.storageBucket, jsonFileObject.Name)
+	// if jsonFileObject.Metadata["rasterized"] == "true" {
+	// 	glog.Infof("> > skipping %s (already processed) {%d}", jsonURL, xformer.counter)
+	// 	return
+	// }
+	body, length, err := gsFetch(jsonFileObject, xformer.client)
+	if err != nil {
+		glog.Errorf("Failed to fetch %s", jsonURL)
+		return
+	}
+	if 0 == length {
+		util.Close(body)
+		glog.Infof("> > skipping %s (empty file) {%d}", jsonURL, xformer.counter)
+		return
+	}
+	dmstruct := dm{}
+	err = json.NewDecoder(body).Decode(&dmstruct)
+	util.Close(body)
+	if err != nil {
+		glog.Errorf("Failed to parse %s", jsonURL)
+		return
+	}
+	countPdfResults := 0
+	for _, res := range dmstruct.Results {
+		if res.Options.Ext == pdfExt {
+			countPdfResults++
+		}
+	}
+	if 0 == countPdfResults {
+		glog.Infof("> > 0 PDFs found %s {%d}", jsonURL, xformer.counter)
+		xformer.setRasterized(jsonFileObject)
+		return
+	}
+
+	glog.Infof("> > processing %d pdfs of %d results {%d}", countPdfResults, len(dmstruct.Results), xformer.counter)
+	rasterizedResults := []result{}
+	i := 0
+	for _, res := range dmstruct.Results {
+		if res.Options.Ext == pdfExt {
+			i++
+			glog.Infof("> > > processing %s.pdf [%d/%d] {%d}", res.Md5, i, countPdfResults, xformer.counter)
+			rasterizedResults = append(rasterizedResults, xformer.processResult(res)...)
+		}
+	}
+	newDMStruct := dm{
+		BuildNumber: dmstruct.BuildNumber,
+		GitHash:     dmstruct.GitHash,
+		Key:         dmstruct.Key,
+		MaxRSSMB:    dmstruct.MaxRSSMB,
+		Results:     rasterizedResults,
+	}
+	newJson, err := json.Marshal(newDMStruct)
+	assertNil(err)
+
+	now := time.Now()
+	// Change the date; leave most of the rest of the path components.
+	jsonPathComponents := strings.Split(jsonFileObject.Name, "/") // []string
+	if len(jsonPathComponents) < 4 {
+		fmt.Errorf("unexpected number of path components %q", jsonPathComponents)
+		return
+	}
+	jsonPathComponents = jsonPathComponents[len(jsonPathComponents)-4:]
+	jsonPathComponents[1] += "-pdfxformer"
+	jsonUploadPath := fmt.Sprintf("%s/%d/%02d/%02d/%02d/%s",
+		xformer.storageJsonDirectory,
+		now.Year(),
+		int(now.Month()),
+		now.Day(),
+		now.Hour(),
+		strings.Join(jsonPathComponents, "/"))
+
+	_, err = uploadFile(xformer.client, bytes.NewReader(newJson), xformer.storageBucket, jsonUploadPath, xformer.accessControlEntity)
+	glog.Infof("> > wrote gs://%s/%s", xformer.storageBucket, jsonUploadPath)
+	newJsonFileObject, err := xformer.client.storageService.Objects.Get(xformer.storageBucket, jsonUploadPath).Do()
+	if err != nil {
+		glog.Errorf("Failed to find %s: %s", jsonUploadPath, err)
+	} else {
+		xformer.setRasterized(newJsonFileObject)
+	}
+	xformer.setRasterized(jsonFileObject)
+}
+
+// setRasterized sets the rasterized metadata flag of the given storage.Object
+func (xformer *pdfXformer) setRasterized(jsonFileObject *storage.Object) {
+	if nil == jsonFileObject.Metadata {
+		jsonFileObject.Metadata = map[string]string{}
+	}
+	jsonFileObject.Metadata["rasterized"] = "true"
+	_, err := xformer.client.storageService.Objects.Patch(xformer.storageBucket, jsonFileObject.Name, jsonFileObject).Do()
+	if err != nil {
+		glog.Errorf("Failed to update metadata of %s: %s", jsonFileObject.Name, err)
+	} else {
+		glog.Infof("> > Updated metadata of %s", jsonFileObject.Name)
+	}
+}
+
+// processTimeRange calls gs.GetLatestGSDirs to get a list of
+func (xformer *pdfXformer) processTimeRange(start time.Time, end time.Time) {
+	glog.Infof("Processing time range: (%s, %s)", start.Truncate(time.Second), end.Truncate(time.Second))
+	for _, dir := range gs.GetLatestGSDirs(start.Unix(), end.Unix(), xformer.storageJsonDirectory) {
+		glog.Infof("> Reading gs://%s/%s\n", xformer.storageBucket, dir)
+		requestedObjects := xformer.client.storageService.Objects.List(xformer.storageBucket).Prefix(dir).Fields(
+			"nextPageToken", "items/updated", "items/md5Hash", "items/mediaLink", "items/name", "items/metadata")
+		for requestedObjects != nil {
+			responseObjects, err := requestedObjects.Do()
+			if err != nil {
+				glog.Errorf("request %#v failed: %s", requestedObjects, err)
+				continue
+			}
+			for _, jsonObject := range responseObjects.Items {
+				xformer.counter++
+				glog.Infof("> > Processing object:  gs://%s/%s {%d}", xformer.storageBucket, jsonObject.Name, xformer.counter)
+				xformer.processJsonFile(jsonObject)
+			}
+			if len(responseObjects.NextPageToken) > 0 {
+				requestedObjects.PageToken(responseObjects.NextPageToken)
+			} else {
+				requestedObjects = nil
+			}
+		}
+	}
+	glog.Infof("finished time range.")
+}
+
+// uploadErrorImage should be run once to verify that the
+func (xformer *pdfXformer) uploadErrorImage() {
+	_, thisFile, _, ok := runtime.Caller(0)
+	if !ok {
+		glog.Fatalf("Could not find data dir: runtime.Caller() failed.")
+	}
+	filename := fmt.Sprintf("%s.png", errorImageMd5.String())
+	f, err := os.Open(path.Join(path.Dir(thisFile), "data", filename))
+	if err != nil {
+		glog.Fatalf("Could not open data/%s", filename)
+	}
+	defer util.Close(f)
+	errorImagePath := fmt.Sprintf("%s/%s", xformer.storageImagesDirectory, filename)
+	_, err = uploadFile(xformer.client, f, xformer.storageBucket, errorImagePath, xformer.accessControlEntity)
+	assertNil(err) // If we can't upload this, we can't upload anything.
+}
+
+// Environment variables: we respect $TMPDIR
+// Arguments: glog uses -logtostderr, -log_dir
+func main() {
+	flag.Parse()
+
+	// TODO(halcanary): where should this file exist?
+	configDir := path.Join(os.Getenv("HOME"), ".config")
+	assertNil(os.MkdirAll(configDir, 0700))
+
+	xformer := pdfXformer{
+		client: getClient(path.Join(configDir, "google_storage_token.data")),
+		// storageBucket:          "chromium-skia-gm",
+		// storageJsonDirectory:   "dm-json-v1",
+		// storageImagesDirectory: "dm-images-v1",
+		accessControlEntity:    "domain-google.com",
+		results:                map[md5digest]map[int]md5digest{},
+		storageBucket:          "skia-infra-testdata",
+		storageJsonDirectory:   "pdfxformer-testdata/json",
+		storageImagesDirectory: "pdfxformer-testdata/img",
+	}
+
+	xformer.uploadErrorImage()
+
+	for _, rasterizer := range []pdf.Rasterizer{pdf.Pdfium{}, pdf.Poppler{}} {
+		if rasterizer.Enabled() {
+			xformer.rasterizers = append(xformer.rasterizers, rasterizer)
+		} else {
+			glog.Infof("rasterizer %s is disabled", rasterizer.String())
+		}
+	}
+	if len(xformer.rasterizers) == 0 {
+		glog.Fatalf("no rasterizers found")
+	}
+
+	end := time.Now()
+	start := end.Add(-72 * time.Hour)
+	timeTicker := time.Tick(time.Minute)
+	for {
+		xformer.processTimeRange(start, end)
+		glog.Flush() // Flush before waiting for next tick; it may be a while.
+		_ = <-timeTicker
+		start = end
+		end = time.Now()
+	}
+}