Nothing Special   »   [go: up one dir, main page]
Skip to content

Commit

Permalink
Fix pdfcpu#385
Browse files Browse the repository at this point in the history
  • Loading branch information
@hhrutter
hhrutter committed Jan 18, 2022
1 parent 96a0a30 commit a002745
Show file tree
Hide file tree
Showing 3 changed files with 163 additions and 52 deletions.
5 changes: 5 additions & 0 deletions pkg/pdfcpu/context.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -396,6 +396,9 @@ type OptimizationContext struct {
DuplicateImages map[int]*StreamDict // Registry of duplicate image dicts.
DuplicateImageObjs IntSet // The set of objects that represents the union of the object graphs of all duplicate image dicts.

ContentStreamCache map[int]*StreamDict
FormStreamCache map[int]*StreamDict

DuplicateInfoObjects IntSet // Possible result of manual info dict modification.
NonReferencedObjs []int // Objects that are not referenced.

Expand All @@ -414,6 +417,8 @@ func newOptimizationContext() *OptimizationContext {
DuplicateImages: map[int]*StreamDict{},
DuplicateImageObjs: IntSet{},
DuplicateInfoObjects: IntSet{},
ContentStreamCache: map[int]*StreamDict{},
FormStreamCache: map[int]*StreamDict{},
Cache: map[int]bool{},
}
}
Expand Down
5 changes: 4 additions & 1 deletion pkg/pdfcpu/equal.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -157,6 +157,9 @@ func equalDicts(d1, d2 Dict, xRefTable *XRefTable) (bool, error) {
return false, nil
}

t1, t2 := d1.Type(), d2.Type()
fontDicts := (t1 != nil && *t1 == "Font") && (t2 != nil && *t2 == "Font")

for key, v1 := range d1 {

v2, found := d2[key]
Expand All @@ -166,7 +169,7 @@ func equalDicts(d1, d2 Dict, xRefTable *XRefTable) (bool, error) {
}

// Special treatment for font dicts
if key == "BaseFont" || key == "FontName" || key == "Name" {
if fontDicts && (key == "BaseFont" || key == "FontName" || key == "Name") {

ok, err := equalFontNames(v1, v2, xRefTable)
if err != nil {
Expand Down
205 changes: 154 additions & 51 deletions pkg/pdfcpu/optimize.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@ limitations under the License.
package pdfcpu

import (
"bytes"
"fmt"
"sort"
"strings"
Expand All @@ -25,8 +26,46 @@ import (
"github.com/pkg/errors"
)

// Mark all content streams for a page dictionary (for stats).
func identifyPageContent(xRefTable *XRefTable, pageDict Dict, pageObjNumber int) error {
func optimizeContentStream(ctx *Context, sd *StreamDict, objNr int) (*IndirectRef, error) {

f := ctx.Optimize.ContentStreamCache
if len(f) == 0 {
f[objNr] = sd
return nil, nil
}

if f[objNr] != nil {
return nil, nil
}

cachedObjNrs := []int{}
for objNr, sd1 := range f {
if *sd1.StreamLength == *sd.StreamLength {
cachedObjNrs = append(cachedObjNrs, objNr)
}
}
if len(cachedObjNrs) == 0 {
f[objNr] = sd
return nil, nil
}

for _, objNr := range cachedObjNrs {
sd1 := f[objNr]
if bytes.Equal(sd.Raw, sd1.Raw) {
ir := NewIndirectRef(objNr, 0)
entry, ok := ctx.FindTableEntryForIndRef(ir)
if ok {
entry.RefCount++
}
return ir, nil
}
}

f[objNr] = sd
return nil, nil
}

func optimizePageContent(ctx *Context, pageDict Dict, pageObjNumber int) error {
log.Optimize.Println("identifyPageContent begin")

o, found := pageDict.Find("Contents")
Expand All @@ -39,16 +78,24 @@ func identifyPageContent(xRefTable *XRefTable, pageDict Dict, pageObjNumber int)

if ir, ok := o.(IndirectRef); ok {

entry, found := xRefTable.FindTableEntry(ir.ObjectNumber.Value(), ir.GenerationNumber.Value())
objNr := ir.ObjectNumber.Value()
entry, found := ctx.FindTableEntry(objNr, ir.GenerationNumber.Value())
if !found {
return errors.Errorf("identifyPageContent: obj#:%d illegal indRef for Contents\n", pageObjNumber)
}

contentStreamDict, ok := entry.Object.(StreamDict)
if ok {
ir, err := optimizeContentStream(ctx, &contentStreamDict, objNr)
if err != nil {
return err
}
if ir != nil {
pageDict["Contents"] = *ir
}
contentStreamDict.IsPageContent = true
entry.Object = contentStreamDict
log.Optimize.Printf("identifyPageContent end: ok obj#%d\n", ir.ObjectNumber.Value())
log.Optimize.Printf("identifyPageContent end: ok obj#%d\n", objNr)
return nil
}

Expand All @@ -68,7 +115,7 @@ func identifyPageContent(xRefTable *XRefTable, pageDict Dict, pageObjNumber int)
return errors.Errorf("identifyPageContent: obj#:%d corrupt page content array entry\n", pageObjNumber)
}

entry, found := xRefTable.FindTableEntry(ir.ObjectNumber.Value(), ir.GenerationNumber.Value())
entry, found := ctx.FindTableEntry(ir.ObjectNumber.Value(), ir.GenerationNumber.Value())
if !found {
return errors.Errorf("identifyPageContent: obj#:%d illegal indRef for Contents\n", pageObjNumber)
}
Expand All @@ -77,7 +124,8 @@ func identifyPageContent(xRefTable *XRefTable, pageDict Dict, pageObjNumber int)
if !ok {
return errors.Errorf("identifyPageContent: obj#:%d page content entry is no stream dict\n", pageObjNumber)
}

// TODO Check against contentStream cache.
// contentArr[i] = *ir
contentStreamDict.IsPageContent = true
entry.Object = contentStreamDict
log.Optimize.Printf("identifyPageContent: ok obj#%d\n", ir.GenerationNumber.Value())
Expand Down Expand Up @@ -364,12 +412,95 @@ func handleDuplicateImageObject(ctx *Context, imageDict *StreamDict, resourceNam
return nil, nil
}

// Get rid of redundant XObjects e.g. embedded images.
func optimizeXObjectImage(ctx *Context, osd *StreamDict, rName string, objNr, pageNumber int, pageImages IntSet) (*IndirectRef, error) {

// Already registered image object that appears in different resources dicts.
if _, found := ctx.Optimize.ImageObjects[objNr]; found {
// This image has already been registered.
//log.Optimize.Printf("optimizeXObjectResourcesDict: Imageobject %d already registered\n", objNr)
pageImages[objNr] = true
return nil, nil
}

// Check if image is a duplicate and if so return the object number of the original.
originalObjNr, err := handleDuplicateImageObject(ctx, osd, rName, objNr, pageNumber)
if err != nil {
return nil, err
}

if originalObjNr != nil {
// We have identified a redundant image!
// Update xobject resource dict so that rName points to the original.
ir := NewIndirectRef(*originalObjNr, 0)
//rDict[rName] = *ir
// Increase refCount for *originalObjNr
entry, ok := ctx.FindTableEntryForIndRef(ir)
if ok {
entry.RefCount++
}
return ir, nil
}

// Register new image dict.
log.Optimize.Printf("optimizeXObjectResourcesDict: adding new image obj#%d\n", objNr)

ctx.Optimize.ImageObjects[objNr] =
&ImageObject{
ResourceNames: []string{rName},
ImageDict: osd,
}

pageImages[objNr] = true
return nil, nil
}

func optimizeXObjectForm(ctx *Context, sd *StreamDict, rName string, objNr int) (*IndirectRef, error) {

f := ctx.Optimize.FormStreamCache
if len(f) == 0 {
f[objNr] = sd
return nil, nil
}

if f[objNr] != nil {
return nil, nil
}

cachedObjNrs := []int{}
for objNr, sd1 := range f {
if *sd1.StreamLength == *sd.StreamLength {
cachedObjNrs = append(cachedObjNrs, objNr)
}
}
if len(cachedObjNrs) == 0 {
f[objNr] = sd
return nil, nil
}

for _, objNr := range cachedObjNrs {
sd1 := f[objNr]
ok, err := equalStreamDicts(sd, sd1, ctx.XRefTable)
if err != nil {
return nil, err
}
if ok {
ir := NewIndirectRef(objNr, 0)
entry, ok := ctx.FindTableEntryForIndRef(ir)
if ok {
entry.RefCount++
}
return ir, nil
}
}

f[objNr] = sd
return nil, nil
}

func optimizeXObjectResourcesDict(ctx *Context, rDict Dict, pageNumber, pageObjNumber int) error {
log.Optimize.Printf("optimizeXObjectResourcesDict page#%dbegin: %s\n", pageObjNumber, rDict)
pageImages := pageImages(ctx, pageNumber)

// Iterate over XObject resource dict.
for rName, v := range rDict {

indRef, ok := v.(IndirectRef)
Expand All @@ -381,9 +512,6 @@ func optimizeXObjectResourcesDict(ctx *Context, rDict Dict, pageNumber, pageObjN
objNr := int(indRef.ObjectNumber)
log.Optimize.Printf("optimizeXObjectResourcesDict: objectNumber = %d\n", objNr)

// We are dealing with a new XObject..
// Dereference the XObject stream dict.

osd, _, err := ctx.DereferenceStreamDict(indRef)
if err != nil {
return err
Expand All @@ -399,55 +527,28 @@ func optimizeXObjectResourcesDict(ctx *Context, rDict Dict, pageNumber, pageObjN
}

if *osd.Dict.Subtype() == "Image" {

// Already registered image object that appears in different resources dicts.
if _, found := ctx.Optimize.ImageObjects[objNr]; found {
// This image has already been registered.
//log.Optimize.Printf("optimizeXObjectResourcesDict: Imageobject %d already registered\n", objNr)
pageImages[objNr] = true
continue
}

// Check if image is a duplicate and if so return the object number of the original.
originalObjNr, err := handleDuplicateImageObject(ctx, osd, rName, objNr, pageNumber)
ir, err := optimizeXObjectImage(ctx, osd, rName, objNr, pageNumber, pageImages)
if err != nil {
return err
}

if originalObjNr != nil {
// We have identified a redundant image!
// Update xobject resource dict so that rName points to the original.
ir := NewIndirectRef(*originalObjNr, 0)
if ir != nil {
rDict[rName] = *ir
// Increase refCount for *originalObjNr
entry, ok := ctx.FindTableEntryForIndRef(ir)
if ok {
entry.RefCount++
}
continue
}

// Register new image dict.
log.Optimize.Printf("optimizeXObjectResourcesDict: adding new image obj#%d\n", objNr)

ctx.Optimize.ImageObjects[objNr] =
&ImageObject{
ResourceNames: []string{rName},
ImageDict: osd,
}

pageImages[objNr] = true
continue
}

if *osd.Subtype() != "Form" {
log.Optimize.Printf("optimizeXObjectResourcesDict: unexpected stream dict Subtype %s\n", *osd.Dict.Subtype())
if *osd.Subtype() == "Form" {
ir, err := optimizeXObjectForm(ctx, osd, rName, objNr)
if err != nil {
return err
}
if ir != nil {
rDict[rName] = *ir
}
continue
}

// Process form dict
log.Optimize.Printf("optimizeXObjectResourcesDict: parsing form dict obj:%d\n", objNr)
parseResourcesDict(ctx, osd.Dict, pageNumber, objNr)
log.Optimize.Printf("optimizeXObjectResourcesDict: unexpected stream dict Subtype %s\n", *osd.Dict.Subtype())
}

log.Optimize.Println("optimizeXObjectResourcesDict end")
Expand Down Expand Up @@ -600,8 +701,7 @@ func parsePagesDict(ctx *Context, pagesDict Dict, pageNumber int) (int, error) {

// Process page dict.

// Mark page content streams for stats.
if err = identifyPageContent(ctx.XRefTable, pageNodeDict, int(ir.ObjectNumber)); err != nil {
if err = optimizePageContent(ctx, pageNodeDict, int(ir.ObjectNumber)); err != nil {
return 0, err
}

Expand Down Expand Up @@ -748,6 +848,9 @@ func optimizeFontAndImages(ctx *Context) error {
return err
}

ctx.Optimize.ContentStreamCache = map[int]*StreamDict{}
ctx.Optimize.FormStreamCache = map[int]*StreamDict{}

// Identify all duplicate objects.
if err = calcRedundantObjects(ctx); err != nil {
return err
Expand Down

0 comments on commit a002745

Please sign in to comment.