Flutter ETS API Documentation v1.0.0

Class: PreviewTextStringMerge

Stateless helpers for preview text merging, tail-shrink picking, and caret mapping.

Constructors

Constructor

new PreviewTextStringMerge(): PreviewTextStringMerge

Returns

PreviewTextStringMerge

Methods

applyFullBodySuffixStitch()

static applyFullBodySuffixStitch(ctx): void

Stitch suffix-only IME full-body when caret inside preview.

Parameters

ctx

PreviewTextMergeContext

Returns

void

applyLocalFullBodyPreviewMerge()

static applyLocalFullBodyPreviewMerge(ctx): void

Full-body preview merge: shrink uses local one-step rules (L'R, getImePreviewBackspaceDeleteRange, pick…). When the IME body is a strict UTF-16 prefix shorter than prev and the caret lies in the removed suffix (offset >= len(ime)), use ime so pre-edit text matches candidates (e.g. z'n'x'h + z'n'x at tail — local bk can wrongly yield z'n'h). If the caret is still inside the kept prefix (offset < len(ime)), keep local bk. Growth / same-length updates use the IME callback string as the whole body. ListenableEditingState forces full IMC sync.

Parameters

ctx

PreviewTextMergeContext

Returns

void

collapseConsecutiveImeApostrophesInPreview()

static collapseConsecutiveImeApostrophesInPreview(text): string

Mid-preview edits must not leave adjacent IME apostrophe-like code units ('', '+U+2019, …); fold every consecutive pair to one ASCII ' (e.g. letter delete at h'|s after skipping ').

Parameters

text

string

Returns

string

collapseConsecutiveImeApostrophesInPreviewAndCaret()

static collapseConsecutiveImeApostrophesInPreviewAndCaret(previewText, caretRel): PreviewCollapseApostropheCaretResult

Same as collapseConsecutiveImeApostrophesInPreview, but keeps a preview-relative caret aligned when code units before it are removed (pairs at runStart,runStart+1 → one ' at runStart).

Parameters

previewText

string

caretRel

number

Returns

PreviewCollapseApostropheCaretResult

computeEffectiveTextAfterShrinkRules()

static computeEffectiveTextAfterShrinkRules(ctx): void

Preview body after merge; no IME-vs-local effective override.

Parameters

ctx

PreviewTextMergeContext

Returns

void

correctDuplicativeOpeningTextSplice()

static correctDuplicativeOpeningTextSplice(ctx): void

Undo mistaken “duplicative opening” range splice for segmented pinyin.

Parameters

ctx

PreviewTextMergeContext

Returns

void

detectSingleInsertOffsetInPreview()

static detectSingleInsertOffsetInPreview(prev, next): number

If prev→next is one contiguous insert of delta code units at some offset, return that offset; else -1.

Parameters

string

Returns

number

dropMisleadingSpanEndSpliceForMidCaret()

static dropMisleadingSpanEndSpliceForMidCaret(ctx): void

Drop misleading span-end range when caret is left of splice (tap mid-preview + Backspace).

Parameters

ctx

PreviewTextMergeContext

Returns

void

fixBareApostropheInsertWithoutSegmentMarker()

static fixBareApostropheInsertWithoutSegmentMarker(ctx): void

prev has no ' but IME full body adds segment apostrophes — rebare-append at caret.

Parameters

ctx

PreviewTextMergeContext

Returns

void

fixUtf16SuffixAppendAtMidCaret()

static fixUtf16SuffixAppendAtMidCaret(ctx): void

Full-body: IME appends at UTF-16 end while caret is mid — rotate insertion to caret. Aligns insert product rule (new text left of caret; caret after insert), not tail-only append.

Parameters

ctx

PreviewTextMergeContext

Returns

void

initPreviewRangeSplice()

static initPreviewRangeSplice(ctx, rangeStart, rangeEnd): void

OHOS: range.start/end ≥ 0 → splice inside previous preview string. When the preview span does not start at document 0 (oldLeft > 0), range is usually in document coordinates; convert splice bounds by subtracting oldLeft when rangeStart >= oldLeft so mid-preview insert (e.g. doc 6..11 + ime) maps to preview 0..5, not mistaken preview indices 6..11 (clamped) which yield z'm'z'y.

Parameters

ctx

PreviewTextMergeContext

rangeStart

number

rangeEnd

number

Returns

void

isImePreviewApostropheLikeChar()

static isImePreviewApostropheLikeChar(segmentChar): boolean

Matches TextUtils.isOnlyImePreviewApostropheRun: common IME segment apostrophe code units (incl. non-ASCII).

Parameters

segmentChar

string

Returns

boolean

pickCaretLeftDeleteForImePreviewTailShrink()

static pickCaretLeftDeleteForImePreviewTailShrink(prev, offset): string

When the caret is not at the preview end: enforce preview invariant 1 (one Backspace step left of the caret only), do not accept IME “tail shrink” substitutes.

If [delPlain,offset) is apostrophes only: take plainBk if it does not introduce illegal '' (x'|j→xj); else try ext so ext does not swallow the syllable’s first letter leaving only j.
After plain deletes one char, if x''z appears: try collapsing first ''→', unless '' bridged a letter in prev; then consider extended Backspace.

Parameters

string

offset

number

Returns

string

previewBareForTailShrinkCompare()

static previewBareForTailShrinkCompare(text): string

Strip IME pinyin segment markers ' before deciding if only the preview tail was removed. Otherwise abc'de'f'g → abc'de'f is not a strict UTF-16 prefix (trailing ' vs none), and mid-string caret correction would be missed.

Parameters

text

string

Returns

string

previewCaretAfterDeletionUsingBareStrings()

static previewCaretAfterDeletionUsingBareStrings(prev, next, caretInPrev): number

e.g. a'b'l'c → abc: non-contiguous UTF-16 deletes break previewCaretAfterSingleDeletion; post-Backspace clamp pins mid-string caret to end. Recompute delete geometry on bare strings without ' and map back to next UTF-16.

Parameters

string

caretInPrev

number

Returns

number

previewCaretAfterInsertionAtLocalCaret()

static previewCaretAfterInsertionAtLocalCaret(prev, next, caretInPrev): number

When LCP-based insert detection fails, assume delta code units were inserted at local caret caretInPrev (common IME behavior).

Parameters

string

caretInPrev

number

Returns

number

previewCaretAfterSingleDeletion()

static previewCaretAfterSingleDeletion(prev, next, caretInPrev): number

IME shortens preview once: assume one contiguous delete of delLen code units; returns folded caret after the deleted block under invariant 1; -1 if geometry cannot be recognized.

Parameters

string

caretInPrev

number

Returns

number

previewCaretAfterSingleInsertion()

static previewCaretAfterSingleInsertion(prev, next, caretInPrev): number

IME lengthens preview once: assume delta code units inserted at insStart; returns folded caret under invariant 2 (right of the new insertion); -1 if unrecognized.

Parameters

string

caretInPrev

number

Returns

number

previewHasConsecutiveImeApostropheRun()

static previewHasConsecutiveImeApostropheRun(text): boolean

Whether two adjacent apostrophe-like code units exist (mixing U+2019 with ASCII breaks indexOf("''")).

Parameters

text

string

Returns

boolean

previewIndexOfFirstConsecutiveImeApostropheRun()

static previewIndexOfFirstConsecutiveImeApostropheRun(text): number

Parameters

text

string

Returns

number

previewIntroducesNewConsecutiveApostropheRun()

static previewIntroducesNewConsecutiveApostropheRun(prev, candidate): boolean

Whether candidate introduces a new illegal consecutive-apostrophe run that prev did not have.

Parameters

string

candidate

string

Returns

boolean

previewMapCaretAfterPreviewRangeReplace()

static previewMapCaretAfterPreviewRangeReplace(prevLen, rangeStart, rangeEnd, insLen, caretInPrev): number

After IME replaces preview range [rangeStart, rangeEnd) with insLen UTF-16 code units (this callback’s text), map the old-string folded caret into the new string. When rangeStart===rangeEnd it is pure insert (delSpan=0): caret at and right of the insertion point shifts by insLen.

Parameters

prevLen

number

rangeStart

number

rangeEnd

number

insLen

number

caretInPrev

number

Returns

number

previewSharedUtf16PrefixLen()

static previewSharedUtf16PrefixLen(prev, candidate): number

UTF-16 shared prefix length between prev and candidate (code-unit equality).

Parameters

string

candidate

string

Returns

number

previewStitchSuffixOnlyImeFullBodyIfNeeded()

static previewStitchSuffixOnlyImeFullBodyIfNeeded(prevPreview, imeText, caretRelInPreview): string

In full-preview callbacks the IME sometimes sends only the UTF-16 suffix of the previous preview (not a strict-prefix tail trim); splicing blindly drops the left syllables (e.g. prev=a'h'x'j's, ime=x'j's). When both contain ' segments and the folded caret is near the suffix start, treat text as a suffix and restore full prev.

Parameters

prevPreview

string

imeText

string

caretRelInPreview

number

Returns

string

repairPreviewRangeSpliceWhenDisagreesWithLocalOneStepBackspace()

static repairPreviewRangeSpliceWhenDisagreesWithLocalOneStepBackspace(ctx): void

Mid-preview Backspace: OHOS may deliver a range splice (start/end) + short ime that stitches garbage (e.g. z'm'h'z'y'a + replace [4,9) with z'm → z'm'z'm'a) while the caret is inside the preview. One local getImePreviewBackspaceDeleteRange step at the caret (e.g. delete h → z'm'z'y'a) matches product rules; if it disagrees with the splice merge, drop the splice and use the local string (non–full-body so later steps do not re-run tail-shrink on the wrong IME payload).

Parameters

ctx

PreviewTextMergeContext

Returns

void

shouldKeepSuffixForCaretLeftUnitInLApostropheRThreePreview()

static shouldKeepSuffixForCaretLeftUnitInLApostropheRThreePreview(prevPreview, offsetInPreview): boolean

Exactly L'R (three UTF-16 units: non-apostrophe, IME apostrophe-like, non-apostrophe) with the caret immediately before R. One user unit left of the caret is L' (marker is not deleted alone); result is R (z'|a→a, x'|h→h).

Parameters

prevPreview

string

offsetInPreview

number

Returns

boolean