Format File Changes Since 2025.1.4662

File	Category	Summary
common/images/images.xsl	Transform	When image GetInfo returns no source dimensions (e.g. FrameMaker-rasterized anchored frames with no @source), natural width/height now fall back to the WIF frame's PostScript page size; adds upper-bound scaling so an image that already fits is not upscaled (EPUB2768)
common/locale/locales.xml	Localization	Emptied the search <StopWords> lists for the Japanese, Korean, and Chinese (Simplified and Traditional) locales so CJK terms are no longer dropped from the search index

File	Category	Summary
API/reverb-search.js	Scripting	CJK-aware search highlighting: match guards no longer treat an adjacent Han/Hiragana/Katakana/Hangul character as word-continuation (so CJK terms highlight); word splitter starts a new word on a boundary letter so adjacent CJK characters aren't dropped; highlight wildcard excludes CJK scripts so .* no longer spans a space-less CJK sentence
Pages/scripts/common.js	Scripting	Highlight match guards made script-aware so an adjacent CJK character no longer suppresses term highlighting
Pages/scripts/search-client.js	Scripting	Word splitter starts a new word on a boundary letter (CJK characters no longer dropped at word breaks); highlight wildcard excludes CJK scripts so it stops at the next CJK word boundary
Pages/scripts/unibreak.js	Scripting	Dropped the Katakana keep-together rule (UAX#29 WB13) so each Katakana character is its own word and is individually searchable, consistent with Han/Hiragana
Transforms/baggage-search-index.xsl	Transform	Index-time word breaker drops the Katakana keep-together rule so Katakana breaks per character, matching the client tokenizer
Transforms/search-index.xsl	Transform	Same index-time change — interior Katakana characters now break per character and are indexed

File	Category	Summary
API/reverb-search.js	Scripting	CJK-aware search highlighting: match guards no longer treat an adjacent Han/Hiragana/Katakana/Hangul character as word-continuation (so CJK terms highlight); word splitter starts a new word on a boundary letter so adjacent CJK characters aren't dropped; highlight wildcard excludes CJK scripts so .* no longer spans a space-less CJK sentence
Pages/scripts/common.js	Scripting	Highlight match guards made script-aware so an adjacent CJK character no longer suppresses term highlighting
Pages/scripts/search-client.js	Scripting	Word splitter starts a new word on a boundary letter (CJK characters no longer dropped at word breaks); highlight wildcard excludes CJK scripts so it stops at the next CJK word boundary
Pages/scripts/unibreak.js	Scripting	Dropped the Katakana keep-together rule (UAX#29 WB13) so each Katakana character is its own word and is individually searchable, consistent with Han/Hiragana
Transforms/baggage-search-index.xsl	Transform	Index-time word breaker drops the Katakana keep-together rule so Katakana breaks per character, matching the client tokenizer
Transforms/search-index.xsl	Transform	Same index-time change — interior Katakana characters now break per character and are indexed