platform-bible-utils
    Preparing search index...

    Type Alias UsjSearchOptions

    Options controlling how IUsjReaderWriter.search performs its search

    type UsjSearchOptions = {
        flexibleWhitespaceAtBlockBoundaries?: boolean;
        markerStylesToInclude?: Set<string>;
        normalizationForm?: "NFD";
        wordEdges?: { end?: RegExp; start?: RegExp };
    }
    Index

    Properties

    flexibleWhitespaceAtBlockBoundaries?: boolean

    When true, a capture group named ${SEARCH_WHITESPACE_GROUP_PREFIX}${n} — the prefix followed by one or more digits, and nothing else — may match zero characters, but only at a block boundary: an offset in the concatenated text where the adjacent text nodes belong to different block-level markers, which is where the editor renders a block break. A match whose whitespace group matched zero characters anywhere else is discarded.

    The name is matched exactly, so a caller's own group keeps its ordinary meaning even when its name begins with the same letters ((?<wsGap>…) is left alone).

    This lets a phrase copied out of an editor match across a rendered block break: the clipboard supplies a space where the concatenated text has nothing between the two words. Most such breaks render as a line break; adjacent cells in one table row are the exception, rendering side by side.

    A chapter transition never counts as a boundary — the gap there is not something the editor renders as a break, and a match spanning one would put a \c marker inside a replacement's removed span.

    When markerStylesToInclude drops text, what was dropped decides whether a boundary survives it. Dropped text that belongs to one of the two adjacent blocks — a note at the end of the first paragraph or the start of the next — leaves the boundary in place. Dropped text that is a block of its own — a section heading, a list, a table standing between the two — removes it, because that block is on screen between them and the two are not adjacent lines.

    The filter needs capture-group offsets, so when the pattern carries such a group and lacks the d flag, search runs a rebuilt copy of it. Either way the caller's own RegExp object is left as they compiled it, lastIndex included — note that this means a search never advances a global regex's cursor for the caller.

    markerStylesToInclude?: Set<string>

    Optional set of marker styles (e.g., 'p', 'q1', 'nd') to include in the search. When provided, only text within markers whose style is in this set will be searched. Text inside markers not in this set (e.g., footnotes, cross-references) will be excluded. When omitted, all text is searched.

    normalizationForm?: "NFD"

    When 'NFD', the concatenated text is normalized to NFD before the regex is applied, and match positions are mapped back to the original string. This is required for correct ignoreDiacritics behaviour on NFC source text: a pre-composed character such as é (U+00E9) will not match e[combining]* unless the text is first decomposed. Returned text values are always slices of the original (non-NFD) string. A match that starts or ends inside an original character (for example part of a Hangul syllable, stored precomposed or decomposed) is not returned.

    wordEdges?: { end?: RegExp; start?: RegExp }

    Zero-width tests that a match's start and end must each pass, unless that end sits where the block changes: an offset in the concatenated text where the adjacent text nodes belong to different block-level markers. A match failing one is discarded.

    This is how a word restriction reaches a word at a block's edge. The regex itself sees neighboring blocks joined with nothing between them, so a lookaround in it reads Isaac at the end of a paragraph and And at the start of the next as IsaacAnd, which is no word boundary. Unlike a lookaround, a test here cannot make the regex try a different length of match.