Add page-tree navigation API

README.md

@@ -29,9 +29,10 @@ or too thin (rsc/pdf, ledongthuc/pdf). pdfdisassembler targets PDF 1.x and
 In scope: PDF 1.x and 2.0 reading, classical xref and xref streams,
 indirect-object resolution, stream filters (FlateDecode, ASCII85, ASCIIHex,
 LZW, RunLength), text-string decoding (PDFDocEncoding, UTF-16BE BOM, UTF-8
-BOM), catalog + page tree, DocumentInfo, XMP metadata access, structure
-tree traversal, `/Standard` security handler (V2, V4, V5), defensive
-parsing.
+BOM), catalog + page-tree navigation (page boxes, rotation, resources and
+content streams, with inherited attributes resolved along the `/Parent`
+chain), DocumentInfo, XMP metadata access, structure tree traversal,
+`/Standard` security handler (V2, V4, V5), defensive parsing.
 
 Out of scope: writing PDFs, image filters (DCTDecode/JBIG2/JPX/CCITTFax),
 image rendering, font internals, XFA, public-key encryption, signature
@@ -83,7 +84,8 @@ for entry := range r.Objects() {
 
 More complete examples live under [`examples/`](examples): `inspect`
 prints a summary of a PDF, `structtree` walks the `/StructTreeRoot` as a
-starting point for accessibility tooling.
+starting point for accessibility tooling, and `pageinfo` reports per-page
+boxes, rotation, resources and content size via the page-tree API.
 
 ## Testing
 

doc.go

@@ -10,9 +10,11 @@
 // In scope: PDF 1.x and 2.0 reading, classical xref and xref streams,
 // indirect-object resolution, stream filters (FlateDecode, ASCII85,
 // ASCIIHex, LZW, RunLength), text-string decoding (PDFDocEncoding,
-// UTF-16BE BOM, UTF-8 BOM), catalog + page tree, DocumentInfo, XMP
-// metadata access, structure-tree traversal, the /Standard security
-// handler (V2, V4, V5), defensive xref recovery.
+// UTF-16BE BOM, UTF-8 BOM), catalog + page-tree navigation (page boxes,
+// rotation, resources and content streams, with inherited attributes
+// resolved along the /Parent chain), DocumentInfo, XMP metadata access,
+// structure-tree traversal, the /Standard security handler (V2, V4, V5),
+// defensive xref recovery.
 //
 // Out of scope: writing PDFs, image filters (DCTDecode/JBIG2/JPX/CCITTFax),
 // image rendering, font internals, XFA, public-key encryption, signature

examples/pageinfo/main.go

@@ -0,0 +1,71 @@
+// Command pageinfo prints per-page geometry and content info for a PDF:
+// page count and, for each page, its boxes, rotation, resource categories,
+// and decoded content size. Intended as a starting point for page-importer
+// and layout tooling.
+package main
+
+import (
+	"fmt"
+	"io"
+	"log"
+	"os"
+	"sort"
+
+	"github.com/speedata/pdfdisassembler"
+)
+
+func main() {
+	if len(os.Args) < 2 {
+		fmt.Fprintln(os.Stderr, "usage: pageinfo <file.pdf>")
+		os.Exit(2)
+	}
+	r, err := pdfdisassembler.OpenFile(os.Args[1])
+	if err != nil {
+		log.Fatal(err)
+	}
+	defer r.Close()
+
+	if err := report(os.Stdout, r); err != nil {
+		log.Fatal(err)
+	}
+}
+
+// report writes a human-readable page summary for r to w. The page tree is
+// walked once via Reader.Pages; inherited attributes (boxes, rotation,
+// resources) are resolved by the Page accessors.
+func report(w io.Writer, r *pdfdisassembler.Reader) error {
+	pages, err := r.Pages()
+	if err != nil {
+		return err
+	}
+	fmt.Fprintf(w, "Pages: %d\n", len(pages))
+
+	for _, p := range pages {
+		// Display pages 1-based, matching how readers number them; the API
+		// itself is 0-based (Page.Index).
+		fmt.Fprintf(w, "Page %d:\n", p.Index()+1)
+
+		if box, ok := p.Box(pdfdisassembler.MediaBox); ok {
+			fmt.Fprintf(w, "  MediaBox: %g x %g pt\n", box.Width(), box.Height())
+		}
+		if box, ok := p.Box(pdfdisassembler.CropBox); ok {
+			fmt.Fprintf(w, "  CropBox:  [%g %g %g %g]\n", box.LLX, box.LLY, box.URX, box.URY)
+		}
+		if rot := p.Rotation(); rot != 0 {
+			fmt.Fprintf(w, "  Rotation: %d\n", rot)
+		}
+		if res, ok := p.Resources(); ok {
+			keys := res.Keys()
+			sort.Strings(keys)
+			fmt.Fprintf(w, "  Resources: %v\n", keys)
+		}
+
+		content, err := p.Content()
+		if err != nil {
+			fmt.Fprintf(w, "  Content: (error: %v)\n", err)
+			continue
+		}
+		fmt.Fprintf(w, "  Content: %d bytes decoded\n", len(content))
+	}
+	return nil
+}

examples/pageinfo/main_test.go

@@ -0,0 +1,108 @@
+package main
+
+import (
+	"bytes"
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/speedata/pdfdisassembler"
+)
+
+// buildObjPDF assembles 1-based object bodies into a PDF with a classical xref
+// and the given trailer dictionary body (without the surrounding << >>).
+func buildObjPDF(t *testing.T, objs []string, trailer string) []byte {
+	t.Helper()
+	var buf bytes.Buffer
+	off := func() int { return buf.Len() }
+	fmt.Fprint(&buf, "%PDF-1.7\n%\xE2\xE3\xCF\xD3\n")
+	offsets := make([]int, len(objs)+1)
+	for i, body := range objs {
+		offsets[i+1] = off()
+		fmt.Fprintf(&buf, "%d 0 obj\n%s\nendobj\n", i+1, body)
+	}
+	xrefOff := off()
+	fmt.Fprintf(&buf, "xref\n0 %d\n%010d %05d f \n", len(objs)+1, 0, 65535)
+	for i := 1; i <= len(objs); i++ {
+		fmt.Fprintf(&buf, "%010d %05d n \n", offsets[i], 0)
+	}
+	fmt.Fprintf(&buf, "trailer\n<< /Size %d %s >>\nstartxref\n%d\n%%%%EOF\n",
+		len(objs)+1, trailer, xrefOff)
+	return buf.Bytes()
+}
+
+// buildPagesPDF builds a two-page document: page 1 inherits its MediaBox and
+// Resources from the /Pages root and carries a content stream; page 2 overrides
+// MediaBox, adds a CropBox and /Rotate, and has no content.
+func buildPagesPDF(t *testing.T) []byte {
+	const content = "BT (Hi) Tj ET" // 13 bytes
+	return buildObjPDF(t, []string{
+		"<< /Type /Catalog /Pages 2 0 R >>",
+		"<< /Type /Pages /Count 2 /Kids [ 3 0 R 4 0 R ] /MediaBox [0 0 612 792] /Resources << /Font << /F1 5 0 R >> >> >>",
+		"<< /Type /Page /Parent 2 0 R /Contents 6 0 R >>",
+		"<< /Type /Page /Parent 2 0 R /MediaBox [0 0 200 200] /CropBox [5 5 195 195] /Rotate 90 >>",
+		"<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>",
+		fmt.Sprintf("<< /Length %d >>\nstream\n%s\nendstream", len(content), content),
+	}, "/Root 1 0 R")
+}
+
+func TestReport(t *testing.T) {
+	r, err := pdfdisassembler.Open(bytes.NewReader(buildPagesPDF(t)))
+	if err != nil {
+		t.Fatalf("Open: %v", err)
+	}
+	defer r.Close()
+
+	var out bytes.Buffer
+	if err := report(&out, r); err != nil {
+		t.Fatalf("report: %v", err)
+	}
+	got := out.String()
+
+	wants := []string{
+		"Pages: 2",
+		"Page 1:",
+		"MediaBox: 612 x 792 pt", // inherited from /Pages root
+		"Resources: [Font]",      // inherited
+		"Content: 13 bytes decoded",
+		"Page 2:",
+		"MediaBox: 200 x 200 pt", // overridden locally
+		"CropBox:  [5 5 195 195]",
+		"Rotation: 90",
+		"Content: 0 bytes decoded", // no /Contents
+	}
+	for _, w := range wants {
+		if !strings.Contains(got, w) {
+			t.Errorf("output missing %q; got:\n%s", w, got)
+		}
+	}
+
+	// Page 1 has no rotation, so no Rotation line should appear before "Page 2".
+	page1 := got[strings.Index(got, "Page 1:"):strings.Index(got, "Page 2:")]
+	if strings.Contains(page1, "Rotation:") {
+		t.Errorf("page 1 should not print a Rotation line; got:\n%s", page1)
+	}
+}
+
+// TestReportCyclicKidsTerminates feeds a page tree whose /Kids cycles back on
+// itself; report must return rather than recurse until the stack overflows.
+func TestReportCyclicKidsTerminates(t *testing.T) {
+	data := buildObjPDF(t, []string{
+		"<< /Type /Catalog /Pages 2 0 R >>",
+		"<< /Type /Pages /Count 1 /Kids [ 3 0 R ] >>",
+		"<< /Type /Pages /Kids [ 2 0 R ] >>", // cycle back to obj 2
+	}, "/Root 1 0 R")
+	r, err := pdfdisassembler.Open(bytes.NewReader(data))
+	if err != nil {
+		t.Fatalf("Open: %v", err)
+	}
+	defer r.Close()
+
+	var out bytes.Buffer
+	if err := report(&out, r); err != nil {
+		t.Fatalf("report: %v", err)
+	}
+	if got := out.String(); !strings.Contains(got, "Pages: 0") {
+		t.Errorf("want \"Pages: 0\", got:\n%s", got)
+	}
+}

filter.go

@@ -10,11 +10,31 @@ import (
 // the file, applies decryption if enabled, then runs the declared filter
 // chain.
 func (r *Reader) decodeStream(s *Stream) ([]byte, error) {
+	raw, err := r.rawStreamSlice(s)
+	if err != nil {
+		return nil, err
+	}
+	return r.applyFilters(s, raw, true)
+}
+
+// rawStreamSlice returns the stream's raw bytes as a sub-slice of the file
+// buffer (no copy), after bounds-checking the declared extent.
+func (r *Reader) rawStreamSlice(s *Stream) ([]byte, error) {
 	if s.rawOffset < 0 || s.rawOffset+s.rawLength > int64(len(r.buf)) {
 		return nil, fmt.Errorf("pdfdisassembler: stream %d %d R: bytes out of range", s.objNumber, s.objGeneration)
 	}
-	raw := r.buf[s.rawOffset : s.rawOffset+s.rawLength]
-	return r.applyFilters(s, raw, true)
+	return r.buf[s.rawOffset : s.rawOffset+s.rawLength], nil
+}
+
+// rawStreamBytes returns a copy of the stream's raw bytes (see Stream.RawBytes).
+func (r *Reader) rawStreamBytes(s *Stream) ([]byte, error) {
+	raw, err := r.rawStreamSlice(s)
+	if err != nil {
+		return nil, err
+	}
+	out := make([]byte, len(raw))
+	copy(out, raw)
+	return out, nil
 }
 
 // applyFilters decrypts (if encrypted is true and an encryption context

object.go

@@ -297,6 +297,13 @@ func (s *Stream) RawLength() int64 {
 	return s.rawLength
 }
 
+// RawBytes returns a copy of the stream's raw, undecoded bytes exactly as they
+// appear in the file — before any filter or decryption is applied. For the
+// decoded content, use Content.
+func (s *Stream) RawBytes() ([]byte, error) {
+	return s.reader.rawStreamBytes(s)
+}
+
 // ObjectEntry is yielded by Reader.Objects: an in-use indirect object plus
 // its resolved value.
 type ObjectEntry struct {