Implemented the ability to change the number of search threads at any time.
Implemented an option to replace unpaired surrogate characters.
Implemented an option to skip terms consisting only of blended characters.
Implemented index version update to the new index format without reindexing the documents.
Implemented optimization of disk space consumption of an index.
Implemented optimization of RAM consumption of a loaded index.
Implemented optimization of index loading speed.
Implemented optimization of search speed in large indexes.
Implemented optimization of fuzzy search.
Implemented optimization of wildcard search.
Implemented optimization of deleting and updating documents.
Implemented optimization of the index optimization operation.
You can now allow searching while the index is being modified.
Improved index reliability.
Implemented optimization of memory usage during indexing.
Implemented optimization of retrieving document text.
Implemented optimization of image search.
Implemented handling letter case identically on all platforms.
Implemented optimization of search by chunks.
修复
Fixed the default number of search threads.
Prevent inconsistent indexing after changing term dictionaries.
Prevent merging indexes that process text differently.
GroupDocs.Search for Java V26.9.12026年9月28日
修复
Windows font lookup no longer fails with UnsatisfiedLinkError in WindowsNativeCall.readRegistryStringValues.
GroupDocs.Search for Java V26.9主版本2026年9月24日
功能
Added a new index format 26.7, built on a multi-array tree; existing indexes are upgraded instead of re-indexed.
HTML, HTM and XHTML documents are now extracted by the product itself instead of through GroupDocs.Parser.
The jar now declares Automatic-Module-Name: com.groupdocs.search, so it resolves as a named automatic module.
The manifest copyright year is now generated at build time instead of being pinned.
Updated GroupDocs.Parser 26.9, GroupDocs.Metadata 26.7 components and current Aspose libraries.
修复
Index merging failed with ClassCastException as soon as a merged node grew past 32 children.
Offsets above 127 were read as signed bytes, so larger indexes silently lost postings.
Deleting documents left the index half updated, because an offset file was held open and could not be renamed.
OverflowException during merging, segment conversion and document removal on large indexes.
ArgumentException while merging, from a stale reference to the index data buffer.
Documents matched inside an archive were returned with an empty inner path.
Text extraction from OneNote .one documents failed in released builds.
Indexing PNG, TIF, TIFF, ASF and MPP aborted on a metadata property GroupDocs.Metadata cannot dispatch.
.csv and .tsv are no longer forced on the parser as spreadsheets.
The published POM declares <licenses> and <developers>.
A license document for the redistributed Bouncy Castle library, and a product-level third-party notice, ship in the jar.
The product jar no longer carries TestNG or resources orphaned by its dependencies.
GroupDocs.Search for .NET V26.8主版本2026年8月5日
功能
Implemented a fast text extractor for HTML documents - This improvement introduces a new, fast extractor for extracting text from HTML and XHTML documents. This extractor, in addition to text, extracts embedded images encoded in Base64.
GroupDocs.Search for .NET V26.7.22026年7月22日
修复
Fixed ArgumentException for optimize operation.
GroupDocs.Search for .NET V26.7.12026年7月21日
功能
You can now implement an index version update without access to the source documents.
GroupDocs.Search for .NET V26.7主版本2026年7月15日
修复
Fixed OverflowException for operations that modify the index.
GroupDocs.Search for Java V26.6主版本2026年6月10日
功能
Added default text extraction for .log and .md files: GroupDocs.Search now treats .log files as plain text out of the box. The built-in plain-text extractor recognises .md, .log and .txt, and a new FileType.LOG value is exposed so .log documents are reported with a dedicated file type in search results. No configuration is required - .log files placed in an indexed folder are picked up automatically.
Throw a clear exception when the maximum array size is exceeded: When indexing very large data sets, the internal character tree could silently overflow the 32-bit (Int32) limit on the size of its backing byte array, producing an obscure failure. The allocation path now validates the required size using 64-bit arithmetic and throws an explicit exception with diagnostic statistics (array length, term count, total and separate character counts) the moment the Int32 limit would be exceeded, making the root cause obvious.
Skip missing documents during search instead of throwing: Previously, composing search results over an index whose document set had changed (for example after a document was deleted or the index was optimized) could fail with a "Document is not found" exception. From this release the lookup returns null for a missing document and the result composition simply skips it, so search returns the remaining matches instead of aborting. The change covers all result kinds, light, heavy, phrase and composite results, as well as image search.