GroupDocs.Comparison for Java 26.8 Release Notes

Full list of changes in this release

PDF side-by-side comparison view (COMPARISONNET-4726, COMPARISONNET-4727)

PDF-specific options are now available in a dedicated PdfCompareOptions class (extends CompareOptions). Its DisplayMode controls the layout of the result document:

• INLINE — default. One merged document; deletions and insertions are highlighted on the same pages. Content from both documents may overlap when they differ significantly.
• SIDE_BY_SIDE — each result page shows the source page and its matching target page next to each other; deletions on the left, insertions on the right. Content never overlaps.
• INTERLEAVED — alternating pages: odd pages come from the source (deletions), even pages from the target (insertions). Open the result with “Two Page View” to see each pair together.

In the two non-inline modes, text is compared by Aspose’s side-by-side comparer, so no ChangeInfo objects are produced; images, annotations and watermarks are still compared and marked with rectangle border annotations. Pages are aligned before comparison — when page counts differ, blank pages are inserted at the positions found by a Needleman-Wunsch alignment over per-page word-set similarity. PdfCompareOptions also carries CompareImagesPdf, ImagesInheritanceMode and AnnotationAuthorName; the first two remain on CompareOptions but are deprecated.

Comparing specific PDF pages (COMPARISONNET-4731)

PdfCompareOptions.PagesSetup restricts comparison to a page range. StartPage and EndPage are 1-based and inclusive; either may be left null to mean “from the first page” / “to the last page”. The range is clamped to the actual page count and applied to both documents before any other processing.

Cell location in ChangeInfo (COMPARISONNET-4767)

For Cells comparison (XLSX, CSV, ODS, …) ChangeInfo now reports where the change happened: getRow() and getColumn() return zero-based indices, and getColumnHeader() returns the text of the corresponding column from the first worksheet row when present. All three are null for comparers that do not provide cell coordinates.

CSV result highlighting (COMPARISONNET-4769)

CSV output carries no character styling, so changes could not be seen in the result file. Inserted runs are now wrapped in parentheses and deleted runs in square brackets when the save format is CSV. Other formats are unaffected.

Inline diff for JSON and text (COMPARISONNET-4772)

Text and JSON results are rendered as a true inline diff. The body uses white-space: pre-wrap; every newline in the compared content starts a new <div>, and within a line unchanged, inserted and deleted runs are emitted as sibling <span>, <ins> and <del> elements so changes appear in context. The old rendering — a line-number prefix plus background-coloured spans, followed by a regex pass that rewrote those spans into <ins>/<del> — has been removed entirely.

PDF paragraphs and tables (COMPARISONNET-4751, COMPARISONNET-4763)

Paragraph rendering no longer appends a single TextParagraph and lets Aspose lay it out. The source fragment is built once, its segments are grouped into visual lines using measured segment widths against the available page width, and every segment is then drawn at its own (X, baselineY). This removes both vertical line collapse and horizontal collision between adjacent segments. The wrap budget is the distance from the paragraph’s left edge to the page’s right text boundary, so text spliced in by the diff can exceed the original bounding box and still wrap where it would really wrap.

Table detection now prefers Aspose’s TableAbsorber results and only falls back to reconstructing tables from graphic operators when Aspose finds none. On the rendering side, the default black cell grid is gone — per-cell borders come from the detected source colours, and cells or tables without detected borders draw nothing instead of a white rectangle that was still visible against the page.

Aspose dependency upgrade

Ten Aspose libraries were upgraded. Six needed no source changes at all (Cells, Diagram, Email, Tasks, PSD, BarCode). Four had breaking API changes:

• Words 26.3 — CompositeNode.getChildNodes() was removed; all 14 call sites use getChildNodes(NodeType.ANY, false), which returns the same live collection of immediate children of any type. Range.getFields() now declares throws Exception.
• PDF 26.2 — TextState.getStrikeOut() renamed to isStrikeOut(). The upgrade also enabled the side-by-side change colours (setDeleteColor / setInsertColor), which the previously pinned 24.10 did not expose.
• Slides 26.2 — ISlide.getThumbnail(...) was replaced by getImage(...), returning IImage instead of a BufferedImage. Slide preview generation and slide-to-raster conversion now write straight to the target stream, dropping an intermediate bitmap round-trip.
• CAD 24.6.1 — FileFormat changed from long constants to a real Java enum and Image.getFileFormat() returns it. The DWT constant set was renumbered (DwtCadR11 → DwtCadR1_1, DwtCadR021 → DwtCadR2_21, DwtCadR012 → DwtCadR011_012, …) with no documented one-to-one mapping; since all of them denote DWT, the full published set is mapped. As a side effect DwtCadR2000–DwtCadR2018 are now recognised as DWT, where before they fell through as undetected.
• Imaging 25.12 — FileFormat.Undefined renamed to FileFormat.Unknown.

aspose-pdf 26.2 publishes no jdk17-classified artifact, so the classifier was dropped. That exposed two stale aspose-pdf:23.4:jdk17 declarations in the parent POM which would otherwise have put a second Aspose.PDF version on the classpath; they were removed (the child POM declares the dependency itself).