• PdfTokenizer: Evaluate making sure data type is set also in case of errors, in DetermineDataType/tryReadDataType. This would help leaving the object operable with the right type even in case of parsing errors. This is done already for dictionaries/arrays, see ReadDictionary/ReadArray
• PdfPredictorDecoder: Support BitsPerComponent != 8
• PdfColorSpaceFilterIndexed::GetSourceScanLineSize() handle bitsPerComponent != 8
• Cleanup PdfTreeNode
• Add subsetting of PdfDifferenceEncoding
• PdfDifferenceList: Validate insertion for the public Add methods (like for example enforce "Adobe Glyph List For New Fonts")
• Optimize compilation SALSprep
• Optimize small allocations in tryGetCodePointsFromCharNameLigatures(), tryGetCodePointsFromUnicodeHexLigatures() in PdfDifferenceEncoding.cpp
• PdfEncoding: Evaluate adding move semantics
• PdfVariant/PdfObject: Evaluate adding a TryGetStringLenient(string_view& str) that catches both PdfString/PdfName
• Add remaining PdfNameTree(s) (also stub)
• Add remaining PdfContentStreamOperators
• According to 5014.CIDFont_Spec page 71-72-73 begincidchar, begincidrange, beginbfchar, beginbfrange have a limit of 100 entries. Fix it where relevant
• noexcept nullable<T> methods
• Check/Review doxygen doc
• If the doc is updated, then should not allow to set an encryption
• If the doc is updated and has an encryption, it should not allow to remove it in a following update
• Evaluate make PdfObjectStream not flate filter by default in PdfMemDocument?
• Evaluate move more utf8::next to utf8::unchecked::next
• PdfToggleButton: Add proper IsChecked/ExportValue handling
• Add version of PdfFont::TryCreateSubstituteFont for rendering (metrics/widths of loaded font override metrics found on /FontFile)
• Add a fallback to search font on the system for text extraction purposes, see https://github.com/podofo/podofo/issues/123
• Check PdfWriter should really update doc trailer when saving. Now the new trailer is written but the doc still has the old one
• PdfMemDocument: Check the DeviceStream is not empty before doing an incremental update/signing operation
• PdfMemDocument: Prevent Save() operation after signing operation
• PdfMemDocument: Evaluate release the device after all objects have been loaded (eg. after a full Save())
• PdfParserObject: Evaluate release the device after loading
• PdfElement: Optimize, keep dictionary/array pointer. Evaluate Add shared_ptr PdfElement::GetObjectPtr()
• Check what do with tools/restore manuals
• Fix/complete handling of text extraction in rotated pages (??? Done?)
• Add method to retrieve shared_ptr from PdfObject, PdfFont (and maybe others) to possibly outlive document destruction
• Do more overflow checks using Chromium numerics, which is now bundled. See comments in utls::DoesMultiplicationOverflow()
• PdfFontManager: Add font hash to cache descriptor
• Add special SetAppearance for PdfSignature respecting "Digital Signature Appearances" document specification
• Add text shaping with Harfbuzz https://github.com/harfbuzz/harfbuzz
• Add fail safe sign/update mechanism, meaning the stream gets trimmed to initial length if there's a crash. Not so easy, especially since we are now using STL streams and it's not easy to trim files without access to native handle and low level I/O operations
• Option to unfold Unicode ligatures to separate codepoints during encoded -> utf8 conversion
• Option to convert Unicode ligatures <-> separate codepoints when drawing strings/converting to encoded
• Optimize charbuff to not initialize memory, keeping std::string compatibility, see https://en.cppreference.com/w/cpp/string/basic_string/resize_and_overwrite
• Add backtrace: https://github.com/boostorg/stacktrace

Ideas:

• PdfFontManager: Consider also statically caching the queries and filepaths. Maybe we could also weakly (weak shared pointer) cache metrics instead of fonts
• PdfName: Evaluate unescape lazily, or offer a way to debug/inspect the unescaped sequence a posteriori
• Consider saving/converting to XRef stream by default