V3: Write page annots from loaded PDFs (#452)

* Add Annots when serializing page object
* Handle multiple dicts with pending objects per object (writer)

pdf/model/annotations.go

@@ -306,6 +306,8 @@ type PdfAnnotationWidget struct {
 	AA     core.PdfObject
 	BS     core.PdfObject
 	Parent core.PdfObject
+
+	parent *PdfField
 }
 
 // PdfAnnotationWatermark represents Watermark annotations.
@@ -1796,7 +1798,12 @@ func (widget *PdfAnnotationWidget) ToPdfObject() core.PdfObject {
 	d.SetIfNotNil("A", widget.A)
 	d.SetIfNotNil("AA", widget.AA)
 	d.SetIfNotNil("BS", widget.BS)
-	d.SetIfNotNil("Parent", widget.Parent)
+
+	if widget.parent != nil {
+		d.SetIfNotNil("Parent", widget.parent.GetContainingPdfObject())
+	} else if widget.Parent != nil {
+		d.SetIfNotNil("Parent", widget.Parent)
+	}
 
 	return container
 }

pdf/model/fields.go

@@ -547,16 +547,16 @@ func (f *PdfField) SetFlag(flag FieldFlag) {
 
 // newPdfFieldFromIndirectObject load a field from an indirect object containing the field dictionary.
 func (r *PdfReader) newPdfFieldFromIndirectObject(container *core.PdfIndirectObject, parent *PdfField) (*PdfField, error) {
-	d, isDict := container.PdfObject.(*core.PdfObjectDictionary)
-	if !isDict {
-		return nil, fmt.Errorf("PdfField indirect object not containing a dictionary")
-	}
-
 	// If already processed and cached - return processed model.
 	if field, cached := r.modelManager.GetModelFromPrimitive(container).(*PdfField); cached {
 		return field, nil
 	}
 
+	d, isDict := core.GetDict(container)
+	if !isDict {
+		return nil, fmt.Errorf("PdfField indirect object not containing a dictionary")
+	}
+
 	field := NewPdfField()
 
 	// Field type (required in terminal fields).
@@ -661,7 +661,8 @@ func (r *PdfReader) newPdfFieldFromIndirectObject(container *core.PdfIndirectObj
 			if !ok {
 				return nil, errors.New("invalid widget annotation")
 			}
-			widget.Parent = field.GetContainingPdfObject()
+			widget.parent = field
+			widget.Parent = field.container
 
 			field.Annotations = append(field.Annotations, widget)
 
@@ -695,6 +696,7 @@ func (r *PdfReader) newPdfFieldFromIndirectObject(container *core.PdfIndirectObj
 				if !ok {
 					return nil, ErrTypeCheck
 				}
+				wa.parent = field
 				field.Annotations = append(field.Annotations, wa)
 			} else {
 				childf, err := r.newPdfFieldFromIndirectObject(container, field)

pdf/model/page.go

@@ -495,6 +495,8 @@ func (p *PdfPage) GetPageDict() *core.PdfObjectDictionary {
 			}
 		}
 		d.Set("Annots", arr)
+	} else if p.Annots != nil {
+		d.SetIfNotNil("Annots", p.Annots)
 	}
 
 	return d

pdf/model/writer.go

@@ -157,7 +157,7 @@ type PdfWriter struct {
 	// for writing.
 	// The map stores the object and the dictionary it is contained in.
 	// Only way so we can access the dictionary entry later.
-	pendingObjects map[core.PdfObject]*core.PdfObjectDictionary
+	pendingObjects map[core.PdfObject][]*core.PdfObjectDictionary
 
 	// Forms.
 	acroForm *PdfAcroForm
@@ -179,7 +179,7 @@ func NewPdfWriter() PdfWriter {
 
 	w.objectsMap = map[core.PdfObject]struct{}{}
 	w.objects = []core.PdfObject{}
-	w.pendingObjects = map[core.PdfObject]*core.PdfObjectDictionary{}
+	w.pendingObjects = map[core.PdfObject][]*core.PdfObjectDictionary{}
 	w.traversed = map[core.PdfObject]struct{}{}
 
 	// PDF Version. Can be changed if using more advanced features in PDF.
@@ -466,8 +466,8 @@ func (w *PdfWriter) addObjects(obj core.PdfObject) error {
 				}
 
 				if hasObj := w.hasObject(v); !hasObj {
-					common.Log.Debug("Parent obj is missing!! %T %p %v", v, v, v)
-					w.pendingObjects[v] = dict
+					common.Log.Debug("Parent obj not added yet!! %T %p %v", v, v, v)
+					w.pendingObjects[v] = append(w.pendingObjects[v], dict)
 					// Although it is missing at this point, it could be added later...
 				}
 				// How to handle the parent? Make sure it is present?
@@ -889,15 +889,17 @@ func (w *PdfWriter) Write(writer io.Writer) error {
 	}
 
 	// Check pending objects prior to write.
-	for pendingObj, pendingObjDict := range w.pendingObjects {
+	for pendingObj, pendingObjDicts := range w.pendingObjects {
 		if !w.hasObject(pendingObj) {
-			common.Log.Debug("ERROR Pending object %+v %T (%p) never added for writing", pendingObj, pendingObj, pendingObj)
-			for _, key := range pendingObjDict.Keys() {
-				val := pendingObjDict.Get(key)
-				if val == pendingObj {
-					common.Log.Debug("Pending object found! and replaced with null")
-					pendingObjDict.Set(key, core.MakeNull())
-					break
+			common.Log.Debug("WARN Pending object %+v %T (%p) never added for writing", pendingObj, pendingObj, pendingObj)
+			for _, pendingObjDict := range pendingObjDicts {
+				for _, key := range pendingObjDict.Keys() {
+					val := pendingObjDict.Get(key)
+					if val == pendingObj {
+						common.Log.Debug("Pending object found! and replaced with null")
+						pendingObjDict.Set(key, core.MakeNull())
+						break
+					}
 				}
 			}
 		}

pdf/model/writer_test.go

@@ -0,0 +1,102 @@
+/*
+ * This file is subject to the terms and conditions defined in
+ * file 'LICENSE.md', which is part of this source code package.
+ */
+
+package model
+
+import (
+	"bytes"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/unidoc/unidoc/pdf/core"
+)
+
+// Tests loading annotations from file, writing back out and reloading.
+func TestReadWriteAnnotations(t *testing.T) {
+	f, err := os.Open(`testdata/OoPdfFormExample.pdf`)
+	require.NoError(t, err)
+	defer f.Close()
+
+	reader, err := NewPdfReaderLazy(f)
+	require.NoError(t, err)
+
+	checkAnnots := func(reader *PdfReader, formExpected bool) {
+		// Check Acroform and fields.
+		if formExpected {
+			require.NotNil(t, reader.AcroForm)
+			fields := reader.AcroForm.AllFields()
+			require.Len(t, fields, 17)
+			require.Nil(t, fields[0].Parent)
+		} else {
+			require.Nil(t, reader.AcroForm)
+		}
+
+		// Check annotations.
+		numPages, err := reader.GetNumPages()
+		require.NoError(t, err)
+		require.Equal(t, 1, numPages)
+
+		page, err := reader.GetPage(1)
+		require.NoError(t, err)
+
+		require.NotNil(t, page.Annots)
+		annots, err := page.GetAnnotations()
+		require.NoError(t, err)
+		require.Len(t, annots, 17)
+
+		wa, ok := annots[0].GetContext().(*PdfAnnotationWidget)
+		require.True(t, ok)
+		if formExpected {
+			require.NotNil(t, wa.parent)
+			require.NotNil(t, wa.Parent)
+		} else {
+			require.Nil(t, wa.parent)
+			require.True(t, core.IsNullObject(wa.Parent))
+		}
+	}
+	checkAnnots(reader, true)
+
+	// Write out and reload. With the AcroForm in place.
+	{
+		w := NewPdfWriter()
+		page, err := reader.GetPage(1)
+		require.NoError(t, err)
+		err = w.AddPage(page)
+		require.NoError(t, err)
+		err = w.SetForms(reader.AcroForm)
+		require.NoError(t, err)
+
+		var buf bytes.Buffer
+		err = w.Write(&buf)
+		require.NoError(t, err)
+
+		bufReader := bytes.NewReader(buf.Bytes())
+		reader, err = NewPdfReaderLazy(bufReader)
+		require.NoError(t, err)
+
+		checkAnnots(reader, true)
+	}
+
+	// Write out and reload without setting the AcroForm.
+	{
+		w := NewPdfWriter()
+		page, err := reader.GetPage(1)
+		require.NoError(t, err)
+		err = w.AddPage(page)
+		require.NoError(t, err)
+
+		var buf bytes.Buffer
+		err = w.Write(&buf)
+		require.NoError(t, err)
+
+		bufReader := bytes.NewReader(buf.Bytes())
+		reader, err = NewPdfReaderLazy(bufReader)
+		require.NoError(t, err)
+
+		checkAnnots(reader, false)
+	}
+}