edrlab
in-document publication-wide text search improvements
We already support multiline / line break matching with collapsing of whitespace / DOM space-tab identation. However, we need to implement string normalization of ligatures / unicode pairs, diacritics / accented characters, etc.
See: https://github.com/edrlab/thorium-reader/pull/1086#issue-431414659 https://github.com/edrlab/thorium-reader/pull/1049/files#diff-600356d6a987e6ef5c15d0a38e75c5888e8483d82dd7e9e5fe7c875df4813e50
Good candidate for HTML search algorithm (including diacritics removal / text normalisation with support for back-referencing the original DOM offsets). The highlighting is performed with DOM mutations (insertion of mark element), but we just need the raw output data mapped to our own DOM range serialisation so we can use our own highlighter.
https://github.com/GoogleChromeLabs/text-fragments-polyfill/
- normalise search space as well as query so a simple indexOf() can be used, then de-normalise the matched range to index into the real DOM: https://github.com/GoogleChromeLabs/text-fragments-polyfill/blob/53375fea08665bac009bb0aa01a030e065c3933d/src/text-fragment-utils.js#L924-L934 ==> https://github.com/GoogleChromeLabs/text-fragments-polyfill/blob/53375fea08665bac009bb0aa01a030e065c3933d/src/text-fragment-utils.js#L718-L719 ==> https://github.com/GoogleChromeLabs/text-fragments-polyfill/blob/53375fea08665bac009bb0aa01a030e065c3933d/src/text-fragment-utils.js#L769-L830
- use regexp to avoid normalising search space and query (the regexp defines normalisation rules but the regexp matches are expressed relative to the real DOM text): https://github.com/julkue/mark.js/blob/7f7e9820514e2268918c2259b58aec3bd5f437f6/src/lib/regexpcreator.js#L305-L319 ==> https://github.com/julkue/mark.js/blob/7f7e9820514e2268918c2259b58aec3bd5f437f6/src/lib/mark.js#L837