feat(字体处理): 修复多版本字体裁剪

ofdgo_compression_fonts.go

@@ -36,7 +36,15 @@ func (r *Reader) compressionFontUsage(ctx context.Context, documents map[string]
 	if err != nil || scoped == nil {
 		return nil, err
 	}
+	return resourceFontUsage(ctx, resources, scoped, read)
+}
+
+// resourceFontUsage 按已确认的文档作用域合并各字体文件的字符和字形引用
+// 入参: ctx 取消上下文, resources 资源索引, scoped 各索引的用字范围, read 输出文件读取方法
+// 返回: map[string]*editorFontUsage 各字体文件的用字, error 读取或取消错误
+func resourceFontUsage(ctx context.Context, resources []string, scoped map[string]*editorResourceRefs, read func(string) ([]byte, error)) (map[string]*editorFontUsage, error) {
 	result := make(map[string]*editorFontUsage)
+	protected := make(map[string]bool)
 	for _, name := range resources {
 		if err := ctx.Err(); err != nil {
 			return nil, err
@@ -66,12 +74,33 @@ func (r *Reader) compressionFontUsage(ctx context.Context, documents map[string]
 				return nil, err
 			}
 			location := editorResourceLocation(name, root.attr("BaseLoc"), value)
+			alternate := cleanPackagePath(resolveResourcePath(name, root.attr("BaseLoc"), value))
 			id, charset := editorResourceID(font.attr("ID")), font.attr("Charset")
-			unsafe := id == "" || charset != "" && !strings.EqualFold(charset, "unicode") || location != cleanPackagePath(resolveResourcePath(name, root.attr("BaseLoc"), value))
-			if result[location] == nil {
-				result[location] = newEditorFontUsage()
+			unsafe := id == "" || charset != "" && !strings.EqualFold(charset, "unicode") || location != alternate
+			for _, path := range []string{location, alternate} {
+				if result[path] == nil {
+					result[path] = newEditorFontUsage()
+				}
+				result[path].merge(refs.fonts[id], unsafe)
 			}
-			result[location].merge(refs.fonts[id], unsafe)
+		}
+		for media := range editorResourceNodes(root, "MultiMedias", "MultiMedia") {
+			file := media.child("MediaFile")
+			if file == nil {
+				continue
+			}
+			var value string
+			if err := xml.Unmarshal(data[file.start:file.end], &value); err != nil {
+				return nil, err
+			}
+			for _, location := range []string{editorResourceLocation(name, root.attr("BaseLoc"), value), cleanPackagePath(resolveResourcePath(name, root.attr("BaseLoc"), value))} {
+				protected[location] = true
+			}
+		}
+	}
+	for location := range protected {
+		if usage := result[location]; usage != nil {
+			usage.unsafe = true
 		}
 	}
 	return result, ctx.Err()

ofdgo_editor_font_cleanup.go

@@ -23,10 +23,10 @@ import (
 	"slices"
 )
 
-// compactVersionFonts 清理仅由版本清单保留的无引用字体，同步清单且不修改编辑器快照
-// 入参: ctx 取消上下文, parts 输出修改, removed 待删除条目
+// compactVersionFonts 清理无引用字体并按全部版本的用字裁剪，不修改编辑器快照
+// 入参: ctx 取消上下文, parts 输出修改, removed 待删除条目, generated 本次编码的页面引用
 // 返回: error 读取、结构或取消错误
-func (e *Editor) compactVersionFonts(ctx context.Context, parts map[string][]byte, removed map[string]bool) error {
+func (e *Editor) compactVersionFonts(ctx context.Context, parts map[string][]byte, removed map[string]bool, generated map[string]editorGeneratedReferences) error {
 	reader := e.source.reader
 	versioned := false
 	for _, body := range reader.OFD.DocBody {
@@ -100,12 +100,12 @@ func (e *Editor) compactVersionFonts(ctx context.Context, parts map[string][]byt
 			}
 		}
 	}
-	if len(candidates) == 0 {
-		return nil
-	}
 	if err := pruneExternalFontFiles(ctx, view, slices.Collect(maps.Keys(candidates))...); err != nil {
 		return err
 	}
+	if err := e.subsetVersionFonts(ctx, view, removed, generated); err != nil {
+		return err
+	}
 	for name, data := range view.files {
 		if original, ok := reader.files[name]; !ok || !bytes.Equal(original, data) {
 			parts[name] = data
@@ -119,3 +119,68 @@ func (e *Editor) compactVersionFonts(ctx context.Context, parts map[string][]byt
 	}
 	return nil
 }
+
+// subsetVersionFonts 合并主入口、历史版本和其他文档的用字，保留共享字体的字形编号
+// 入参: ctx 取消上下文, view 独立输出快照, removed 待删除条目, generated 本次编码的页面引用
+// 返回: error 读取、裁剪或取消错误
+func (e *Editor) subsetVersionFonts(ctx context.Context, view *Reader, removed map[string]bool, generated map[string]editorGeneratedReferences) error {
+	if e.backends.FontResources == nil {
+		return nil
+	}
+	names := maps.Clone(view.fileNames)
+	for name := range removed {
+		delete(names, name)
+	}
+	progress := editorProgress(e.OnWriteProgress)
+	refs, resources, references, safe, err := scanEditorResources(ctx, view, view.files, generated, names, progress, true)
+	if err != nil || !safe {
+		return err
+	}
+	scoped, err := resourceReferenceScopes(ctx, refs.documents, resources, references, "")
+	if err != nil || scoped == nil {
+		return err
+	}
+	read := func(name string) ([]byte, error) { return view.readFileViewContext(ctx, name) }
+	usage, err := resourceFontUsage(ctx, resources, scoped, read)
+	if err != nil {
+		return err
+	}
+	for _, source := range references {
+		for name := range source.files {
+			if font := usage[name]; font != nil && !source.versions[name] {
+				font.unsafe = true
+			}
+		}
+		for _, definition := range source.definitions {
+			for name := range definition.files {
+				if font := usage[name]; font != nil {
+					font.unsafe = true
+				}
+			}
+		}
+	}
+	var cache editorSourceFontCache
+	for index, name := range slices.Sorted(maps.Keys(usage)) {
+		if err := progress.report("fonts", index, len(usage)); err != nil {
+			return err
+		}
+		if usage[name].unsafe || len(usage[name].chars)+len(usage[name].glyphs) == 0 {
+			continue
+		}
+		data, err := read(name)
+		if err != nil {
+			return err
+		}
+		subset, err := cache.subset(ctx, e.backends.FontResources, data, usage[name])
+		if err != nil {
+			return err
+		}
+		if subset != nil {
+			view.files[name] = subset
+		}
+	}
+	if err := progress.report("fonts", len(usage), len(usage)); err != nil {
+		return err
+	}
+	return ctx.Err()
+}

ofdgo_editor_package.go

@@ -635,7 +635,7 @@ func (e *Editor) writeSource(writer io.Writer, fonts map[string][]byte, progress
 			return 0, err
 		}
 	}
-	if err := e.compactVersionFonts(ctx, parts, removed); err != nil {
+	if err := e.compactVersionFonts(ctx, parts, removed, generated); err != nil {
 		return 0, err
 	}
 	if len(parts) != 0 || len(removed) != 0 || e.output != nil && e.output.options.Mode != CompressionUnchanged && !e.output.protected {

ofdgo_font_resource_sfnt.go

@@ -20,6 +20,8 @@ import (
 	"fmt"
 	"maps"
 	"slices"
+
+	"github.com/go-text/typesetting/font/opentype/tables"
 )
 
 // SFNTBackend 提供独立的OpenType字体度量、轮廓和资源处理
@@ -209,7 +211,7 @@ func subsetSFNTFont(data []byte, glyphs []uint16, mapped bool) ([]byte, error) {
 }
 
 // subsetSourceSFNTFont 裁剪静态TrueType或CFF原有字体，保留字形编号、组合依赖、度量和提示指令
-// 集合、字形替换、可变、彩色和未知表保持原样，不对不完整的引用或异常字体猜测修复
+// 字形替换结果一并保留，集合、可变、彩色和未知表保持原样
 // 入参: data 字体数据, usage 全包用字记录
 // 返回: []byte 更小的字体子集，无确定收益时为空
 func subsetSourceSFNTFont(data []byte, usage *editorFontUsage) []byte {
@@ -222,7 +224,7 @@ func subsetSourceSFNTFont(data []byte, usage *editorFontUsage) []byte {
 	}
 	for tag := range sfnt.Tables {
 		switch tag {
-		case "cmap", "head", "hhea", "hmtx", "maxp", "OS/2", "post", "name", "glyf", "loca", "CFF ", "cvt ", "fpgm", "prep", "gasp", "kern", "vhea", "vmtx", "VORG", "hdmx", "LTSH", "VDMX", "GDEF", "GPOS", "DSIG", "FFTM":
+		case "cmap", "head", "hhea", "hmtx", "maxp", "OS/2", "post", "name", "glyf", "loca", "CFF ", "cvt ", "fpgm", "prep", "gasp", "kern", "vhea", "vmtx", "VORG", "hdmx", "LTSH", "VDMX", "GDEF", "GPOS", "GSUB", "meta", "DSIG", "FFTM":
 		default:
 			return nil
 		}
@@ -254,6 +256,9 @@ func subsetSourceSFNTFont(data []byte, usage *editorFontUsage) []byte {
 			mapping[char] = id
 		}
 	}
+	if !retainSFNTSubstitutions(sfnt.Tables["GSUB"], glyphs, sfnt.NumGlyphs()) {
+		return nil
+	}
 	if sfnt.IsCFF {
 		count, err := parseCFFAndCountGlyphs(sfnt.Tables["CFF "])
 		if err != nil || count != int(sfnt.NumGlyphs()) {
@@ -318,6 +323,94 @@ func subsetSourceSFNTFont(data []byte, usage *editorFontUsage) []byte {
 	return result
 }
 
+// retainSFNTSubstitutions 保留全部替换结果，使原字形替换表在裁剪后仍可使用
+// 入参: data 替换表, glyphs 待保留字形, count 字形总数
+// 返回: bool 替换表及结果编号是否有效
+func retainSFNTSubstitutions(data []byte, glyphs map[uint16]bool, count uint16) bool {
+	if len(data) == 0 {
+		return true
+	}
+	if len(data) < 10 || binary.BigEndian.Uint16(data) != 1 || binary.BigEndian.Uint16(data[2:]) > 1 {
+		return false
+	}
+	if binary.BigEndian.Uint16(data[2:]) == 1 && (len(data) < 14 || binary.BigEndian.Uint32(data[10:]) != 0) {
+		return false
+	}
+	layout, _, err := tables.ParseLayout(data)
+	if err != nil {
+		return false
+	}
+	for _, lookup := range layout.LookupList.Lookups {
+		subtables, err := lookup.AsGSUBLookups()
+		if err != nil {
+			return false
+		}
+		for _, subtable := range subtables {
+			if extension, ok := subtable.(tables.ExtensionSubs); ok {
+				if extension.ExtensionOffset < 8 || uint64(extension.ExtensionOffset) > uint64(len(extension.RawData)) {
+					return false
+				}
+				subtable, err = extension.Resolve()
+				if err != nil {
+					return false
+				}
+			}
+			if !retainSFNTSubstitution(subtable, glyphs, count) {
+				return false
+			}
+		}
+	}
+	return true
+}
+
+// retainSFNTSubstitution 收集替换子表的输出字形，上下文替换的结果由被引用的查找表提供
+// 入参: subtable 替换子表, glyphs 待保留字形, count 字形总数
+// 返回: bool 子表类型及输出字形是否有效
+func retainSFNTSubstitution(subtable tables.GSUBLookup, glyphs map[uint16]bool, count uint16) bool {
+	var outputs []tables.GlyphID
+	switch subtable := subtable.(type) {
+	case tables.SingleSubs:
+		switch data := subtable.Data.(type) {
+		case tables.SingleSubstData1:
+			for id := uint16(0); id < count; id++ {
+				if _, ok := data.Coverage.Index(tables.GlyphID(id)); ok {
+					outputs = append(outputs, tables.GlyphID(id+uint16(data.DeltaGlyphID)))
+				}
+			}
+		case tables.SingleSubstData2:
+			outputs = data.SubstituteGlyphIDs
+		default:
+			return false
+		}
+	case tables.MultipleSubs:
+		for _, sequence := range subtable.Sequences {
+			outputs = append(outputs, sequence.SubstituteGlyphIDs...)
+		}
+	case tables.AlternateSubs:
+		for _, set := range subtable.AlternateSets {
+			outputs = append(outputs, set.AlternateGlyphIDs...)
+		}
+	case tables.LigatureSubs:
+		for _, set := range subtable.LigatureSets {
+			for _, ligature := range set.Ligatures {
+				outputs = append(outputs, ligature.LigatureGlyph)
+			}
+		}
+	case tables.ReverseChainSingleSubs:
+		outputs = subtable.SubstituteGlyphIDs
+	case tables.ContextualSubs, tables.ChainedContextualSubs:
+	default:
+		return false
+	}
+	for _, id := range outputs {
+		if uint16(id) >= count {
+			return false
+		}
+		glyphs[uint16(id)] = true
+	}
+	return true
+}
+
 // subsetCompactSFNTFont 重排新建静态字体的字形，同步复合引用、字符映射和度量
 // 入参: sfnt 字体, glyphs 使用字形
 // 返回: []byte 紧凑字体子集, error 字形数据错误