@mhmo91/schmancy 0.8.6 → 0.9.2

package/src/utils/search.ts

@@ -0,0 +1,322 @@
+/**
+ * Options for configuring similarity calculation behavior.
+ */
+export interface SimilarityOptions {
+	/**
+	 * Whether to remove accents/diacritics during normalization.
+	 * @default true
+	 */
+	normalizeAccents?: boolean
+
+	/**
+	 * Whether to include word-level Jaccard similarity in fuzzy matching.
+	 * Useful for matching phrases with different word orders.
+	 * @default true
+	 */
+	includeWordJaccard?: boolean
+
+	/**
+	 * Maximum Levenshtein distance threshold for early termination.
+	 * Improves performance for very dissimilar strings.
+	 * Set to 0 to disable early termination.
+	 * @default 0 (disabled)
+	 */
+	maxLevenshteinDistance?: number
+}
+
+// Scoring weights as named constants for clarity and maintainability
+const SCORE_WEIGHT_DICE = 0.45
+const SCORE_WEIGHT_LEVENSHTEIN = 0.45
+const SCORE_WEIGHT_ANAGRAM = 0.3
+const SCORE_WEIGHT_JACCARD = 0.4
+
+// Word splitting pattern - defined once for reuse
+const WORD_SPLIT_PATTERN = /[\s\-_.,;:!?()[\]{}]+/
+
+/**
+ * Calculate similarity score between two strings.
+ * Returns a value between 0 (no match) and 1 (exact match).
+ * Optimized for autocomplete with prioritization of start matches and whole words.
+ *
+ * Scoring tiers (higher scores = better matches):
+ * - 1.00: Exact match
+ * - 0.95-1.00: Target starts with query (autocomplete-style)
+ * - 0.765-0.85: Query matches start of a word in target
+ * - 0.56-0.70: Query is substring of target
+ * - 0.50: Query is subsequence of target
+ * - 0.00-0.50: Fuzzy matching (typos, similar words, character overlap)
+ *
+ * @param query The search query string
+ * @param target The target string to compare against
+ * @param options Optional configuration for similarity calculation
+ * @returns A similarity score from 0 to 1
+ *
+ * @example
+ * similarity('john', 'John Doe') // 0.975+ (starts with)
+ * similarity('doe', 'John Doe') // 0.765+ (word boundary)
+ * similarity('jo', 'John Doe') // 0.95+ (starts with)
+ * similarity('jhn', 'John Doe') // 0.3-0.5 (subsequence/fuzzy)
+ */
+export function similarity(query: string, target: string, options?: SimilarityOptions): number {
+	// Handle edge cases
+	if (!query || !target) return 0
+	if (query === target) return 1
+
+	// Default options
+	const opts: Required<SimilarityOptions> = {
+		normalizeAccents: options?.normalizeAccents ?? true,
+		includeWordJaccard: options?.includeWordJaccard ?? true,
+		maxLevenshteinDistance: options?.maxLevenshteinDistance ?? 0,
+	}
+
+	// Normalize strings for comparison
+	const normalizedQuery = normalizeString(query, opts.normalizeAccents)
+	const normalizedTarget = normalizeString(target, opts.normalizeAccents)
+
+	// 1. Exact match (case-insensitive, accent-insensitive)
+	if (normalizedQuery === normalizedTarget) return 1
+
+	// 2. Target starts with query (highest priority for autocomplete)
+	if (normalizedTarget.startsWith(normalizedQuery)) {
+		// Give higher score to shorter targets (more precise matches)
+		const lengthRatio = normalizedQuery.length / normalizedTarget.length
+		return 0.95 + (lengthRatio * 0.05) // Score between 0.95 and 1.0
+	}
+
+	// 3. Word boundary match (query matches start of any word in target)
+	const words = normalizedTarget.split(WORD_SPLIT_PATTERN).filter(w => w.length > 0)
+	for (let i = 0; i < words.length; i++) {
+		if (words[i].startsWith(normalizedQuery)) {
+			// Score based on which word matched (earlier words score higher)
+			const wordPositionScore = 1 - (i / words.length) * 0.1
+			return 0.85 * wordPositionScore // Score between 0.765 and 0.85
+		}
+	}
+
+	// 4. Direct substring match (query appears anywhere in target)
+	if (normalizedTarget.includes(normalizedQuery)) {
+		// Score based on position (earlier position scores higher)
+		const position = normalizedTarget.indexOf(normalizedQuery)
+		const positionScore = 1 - (position / normalizedTarget.length) * 0.2
+		return 0.7 * positionScore // Score between 0.56 and 0.7
+	}
+
+	// 5. Subsequence check (all query chars appear in order)
+	if (isSubsequence(normalizedQuery, normalizedTarget)) {
+		return 0.5
+	}
+
+	// 6. Fuzzy matching for typos and similar words
+	// Calculate all scores and return the maximum (avoid array allocation)
+	let maxScore = 0
+
+	// 6a. Dice coefficient (good for character-level similarity)
+	const diceScore = diceCoefficient(normalizedQuery, normalizedTarget) * SCORE_WEIGHT_DICE
+	if (diceScore > maxScore) maxScore = diceScore
+
+	// 6b. Levenshtein distance (good for typos)
+	const maxLength = Math.max(normalizedQuery.length, normalizedTarget.length)
+	const levenshteinDistance = calculateLevenshtein(
+		normalizedQuery,
+		normalizedTarget,
+		opts.maxLevenshteinDistance
+	)
+	const levenshteinScore = maxLength ? (1 - levenshteinDistance / maxLength) * SCORE_WEIGHT_LEVENSHTEIN : 0
+	if (levenshteinScore > maxScore) maxScore = levenshteinScore
+
+	// 6c. Character frequency match (anagram-like)
+	if (hasAllCharacters(normalizedQuery, normalizedTarget)) {
+		if (SCORE_WEIGHT_ANAGRAM > maxScore) maxScore = SCORE_WEIGHT_ANAGRAM
+	}
+
+	// 6d. Word-level Jaccard similarity (good for phrase matching)
+	if (opts.includeWordJaccard && words.length > 1) {
+		const queryWords = normalizedQuery.split(WORD_SPLIT_PATTERN).filter(w => w.length > 0)
+		if (queryWords.length > 0) {
+			const jaccardScore = wordJaccardSimilarity(queryWords, words) * SCORE_WEIGHT_JACCARD
+			if (jaccardScore > maxScore) maxScore = jaccardScore
+		}
+	}
+
+	return maxScore
+}
+
+/**
+ * Normalize a string for comparison.
+ * - Converts to lowercase
+ * - Trims whitespace
+ * - Optionally removes accents/diacritics
+ *
+ * @param str The string to normalize
+ * @param removeAccents Whether to remove accents/diacritics
+ * @returns The normalized string
+ */
+function normalizeString(str: string, removeAccents: boolean): string {
+	let normalized = str.toLowerCase().trim()
+
+	if (removeAccents) {
+		// Remove accents using NFD normalization and removing combining marks
+		normalized = normalized.normalize('NFD').replace(/[\u0300-\u036f]/g, '')
+	}
+
+	return normalized
+}
+
+/**
+ * Calculate Jaccard similarity between two sets of words.
+ * Returns a value between 0 (no overlap) and 1 (identical sets).
+ * Optimized to avoid unnecessary array conversions.
+ *
+ * @param words1 Array of words from first string
+ * @param words2 Array of words from second string
+ * @returns Jaccard similarity score (0-1)
+ */
+function wordJaccardSimilarity(words1: string[], words2: string[]): number {
+	if (words1.length === 0 || words2.length === 0) return 0
+
+	const set1 = new Set(words1)
+	const set2 = new Set(words2)
+
+	// Count intersection
+	let intersectionSize = 0
+	set1.forEach(word => {
+		if (set2.has(word)) intersectionSize++
+	})
+
+	// Union size = size1 + size2 - intersection
+	const unionSize = set1.size + set2.size - intersectionSize
+
+	return unionSize > 0 ? intersectionSize / unionSize : 0
+}
+
+/**
+ * Check if string 'sub' is a subsequence of string 'str'.
+ * All characters in 'sub' must appear in order in 'str'.
+ */
+function isSubsequence(sub: string, str: string): boolean {
+	let i = 0,
+		j = 0
+	while (i < sub.length && j < str.length) {
+		if (sub[i] === str[j]) i++
+		j++
+	}
+	return i === sub.length
+}
+
+/**
+ * Check if all characters in 'query' are present in 'target'.
+ * For example, "aovc" matches "avocados" (anagram subset matching).
+ * Optimized with Map for O(n+m) single-pass performance.
+ */
+function hasAllCharacters(query: string, target: string): boolean {
+	// Build character frequency map for target (single pass)
+	const targetCount = new Map<string, number>()
+	for (let i = 0; i < target.length; i++) {
+		const char = target[i]
+		targetCount.set(char, (targetCount.get(char) || 0) + 1)
+	}
+
+	// Check query characters against target (single pass)
+	const queryCount = new Map<string, number>()
+	for (let i = 0; i < query.length; i++) {
+		const char = query[i]
+		queryCount.set(char, (queryCount.get(char) || 0) + 1)
+	}
+
+	// Verify all query chars exist in target with sufficient count
+	let hasAll = true
+	queryCount.forEach((count, char) => {
+		if ((targetCount.get(char) || 0) < count) hasAll = false
+	})
+
+	return hasAll
+}
+
+/**
+ * Compute Dice's coefficient for two strings based on bigrams.
+ * Returns a value between 0 (no similarity) and 1 (perfect match).
+ * Optimized with Map for O(n) performance instead of O(n²).
+ */
+function diceCoefficient(s1: string, s2: string): number {
+	if (s1.length < 2 || s2.length < 2) return 0
+
+	// Build bigram frequency map for s2 (O(n))
+	const bigrams2Map = new Map<string, number>()
+	for (let i = 0; i < s2.length - 1; i++) {
+		const bigram = s2.substring(i, i + 2)
+		bigrams2Map.set(bigram, (bigrams2Map.get(bigram) || 0) + 1)
+	}
+
+	// Count intersections while iterating s1 bigrams (O(n))
+	let intersection = 0
+	let bigrams1Count = 0
+	for (let i = 0; i < s1.length - 1; i++) {
+		const bigram = s1.substring(i, i + 2)
+		bigrams1Count++
+		const count = bigrams2Map.get(bigram)
+		if (count && count > 0) {
+			intersection++
+			bigrams2Map.set(bigram, count - 1)
+		}
+	}
+
+	const bigrams2Count = s2.length - 1
+	return (2 * intersection) / (bigrams1Count + bigrams2Count)
+}
+
+/**
+ * Calculate Levenshtein distance between two strings.
+ * Optimized with O(min(n,m)) space complexity using rolling arrays.
+ * Supports early termination when distance exceeds threshold.
+ *
+ * @param a First string
+ * @param b Second string
+ * @param maxDistance Maximum distance threshold for early termination (0 = disabled)
+ * @returns The Levenshtein distance between the strings
+ */
+function calculateLevenshtein(a: string, b: string, maxDistance = 0): number {
+	// Ensure a is the shorter string for space optimization
+	if (a.length > b.length) [a, b] = [b, a]
+
+	if (a.length === 0) return b.length
+
+	// Early termination: if length difference exceeds max distance
+	if (maxDistance > 0 && b.length - a.length > maxDistance) {
+		return maxDistance + 1
+	}
+
+	// Use two rolling arrays instead of full matrix: O(n) space instead of O(n*m)
+	let prevRow = new Array(a.length + 1)
+	let currRow = new Array(a.length + 1)
+
+	// Initialize first row
+	for (let j = 0; j <= a.length; j++) {
+		prevRow[j] = j
+	}
+
+	// Calculate distances row by row
+	for (let i = 1; i <= b.length; i++) {
+		currRow[0] = i
+		let minRowValue = i
+
+		for (let j = 1; j <= a.length; j++) {
+			const cost = a[j - 1] === b[i - 1] ? 0 : 1
+			currRow[j] = Math.min(
+				prevRow[j] + 1,        // deletion
+				currRow[j - 1] + 1,    // insertion
+				prevRow[j - 1] + cost  // substitution
+			)
+			if (currRow[j] < minRowValue) minRowValue = currRow[j]
+		}
+
+		// Early termination: if minimum value in row exceeds threshold
+		if (maxDistance > 0 && minRowValue > maxDistance) {
+			return maxDistance + 1
+		}
+
+		// Swap arrays for next iteration (O(1) operation)
+		;[prevRow, currRow] = [currRow, prevRow]
+	}
+
+	return prevRow[a.length]
+}

package/src/vite-env.d.ts

@@ -0,0 +1,2 @@
+/// <reference types="vite/client" />
+/// <reference path="../mixins/scss.d.ts" />

package/src/window/index.ts

@@ -0,0 +1,10 @@
+export { default as SchmancyWindow } from './window'
+export { windowManager } from './window-manager'
+export type {
+	WindowBounds,
+	WindowRecord,
+	WindowRegistryState,
+	WindowVisualState,
+	SnapCorner,
+	SnapTarget,
+} from './window-registry'

package/src/window/window-manager.ts

@@ -0,0 +1,204 @@
+/**
+ * Window Manager — singleton service for multi-window coordination.
+ *
+ * Tracks all schmancy-window instances: positions, z-indices, focus, visual states.
+ * Components register on connectedCallback, unregister on disconnectedCallback.
+ * Focus changes, position updates, and overlap queries all flow through here.
+ *
+ * Follows the ThemeService singleton pattern (theme/theme.service.ts).
+ *
+ * Example:
+ *   windowManager.register('chat-1', bounds, 'bottom-right')
+ *   windowManager.focus('chat-1')  // bumps z-index, updates focusedId
+ *   const overlaps = windowManager.findOverlaps(projectedBounds, 'chat-1')
+ */
+
+import { BehaviorSubject, Observable, distinctUntilChanged, map } from 'rxjs'
+import { overlayStack } from '../utils/overlay-stack.js'
+import type { WindowBounds, WindowRecord, WindowRegistryState, SnapTarget } from './window-registry.js'
+
+const STORAGE_PREFIX = 'schmancy-window-'
+const LEGACY_STORAGE_PREFIX = 'schmancy-float-'
+
+class WindowManagerService {
+	private static instance: WindowManagerService
+
+	private readonly _state$ = new BehaviorSubject<WindowRegistryState>({
+		windows: new Map(),
+		focusedId: null,
+		stackOrder: [],
+	})
+
+	readonly state$ = this._state$.asObservable()
+
+	private constructor() {}
+
+	static getInstance(): WindowManagerService {
+		if (!WindowManagerService.instance) {
+			WindowManagerService.instance = new WindowManagerService()
+		}
+		return WindowManagerService.instance
+	}
+
+	// ── Synchronous accessors ──────────────────────────────────────────
+
+	get windows(): Map<string, WindowRecord> {
+		return this._state$.value.windows
+	}
+
+	get focusedId(): string | null {
+		return this._state$.value.focusedId
+	}
+
+	// ── Registration ───────────────────────────────────────────────────
+
+	register(id: string, initialBounds: WindowBounds, snapTarget: SnapTarget): void {
+		const state = this._state$.value
+		if (state.windows.has(id)) return
+
+		const zIndex = overlayStack.assignZIndex(id)
+		const record: WindowRecord = {
+			id,
+			bounds: initialBounds,
+			visualState: 'normal',
+			zIndex,
+			open: false,
+			snapTarget,
+		}
+
+		const windows = new Map(state.windows)
+		windows.set(id, record)
+		const stackOrder = overlayStack.getStackOrder()
+
+		this._state$.next({ ...state, windows, stackOrder })
+	}
+
+	unregister(id: string): void {
+		const state = this._state$.value
+		if (!state.windows.has(id)) return
+
+		overlayStack.releaseId(id)
+		const windows = new Map(state.windows)
+		windows.delete(id)
+		const stackOrder = overlayStack.getStackOrder()
+		const focusedId = state.focusedId === id ? null : state.focusedId
+
+		this._state$.next({ ...state, windows, stackOrder, focusedId })
+	}
+
+	// ── State mutations ────────────────────────────────────────────────
+
+	updateBounds(id: string, bounds: WindowBounds): void {
+		this._updateRecord(id, { bounds })
+	}
+
+	updateVisualState(id: string, visualState: WindowRecord['visualState']): void {
+		this._updateRecord(id, { visualState })
+	}
+
+	updateOpen(id: string, open: boolean): void {
+		this._updateRecord(id, { open })
+	}
+
+	focus(id: string): void {
+		const state = this._state$.value
+		if (!state.windows.has(id)) return
+		if (state.focusedId === id) return
+
+		const zIndex = overlayStack.bringToFront(id)
+		const windows = new Map(state.windows)
+		const record = windows.get(id)
+		if (record) {
+			windows.set(id, { ...record, zIndex })
+		}
+		const stackOrder = overlayStack.getStackOrder()
+
+		this._state$.next({ ...state, windows, stackOrder, focusedId: id })
+	}
+
+	// ── Queries ────────────────────────────────────────────────────────
+
+	findOverlaps(bounds: WindowBounds, excludeId: string): WindowRecord[] {
+		const result: WindowRecord[] = []
+		for (const [id, record] of this._state$.value.windows) {
+			if (id === excludeId) continue
+			if (rectsOverlap(bounds, record.bounds)) {
+				result.push(record)
+			}
+		}
+		return result
+	}
+
+	getNeighbors(id: string): WindowRecord[] {
+		const result: WindowRecord[] = []
+		for (const [otherId, record] of this._state$.value.windows) {
+			if (otherId !== id) result.push(record)
+		}
+		return result
+	}
+
+	// ── Selectors ──────────────────────────────────────────────────────
+
+	selectWindow(id: string): Observable<WindowRecord | undefined> {
+		return this._state$.pipe(
+			map(state => state.windows.get(id)),
+			distinctUntilChanged(),
+		)
+	}
+
+	selectFocused(): Observable<string | null> {
+		return this._state$.pipe(
+			map(state => state.focusedId),
+			distinctUntilChanged(),
+		)
+	}
+
+	// ── Persistence ────────────────────────────────────────────────────
+
+	loadPosition(id: string): { x: number; y: number; anchor: string } | null {
+		try {
+			// Try new key first, fall back to legacy
+			const raw = localStorage.getItem(STORAGE_PREFIX + id) ?? localStorage.getItem(LEGACY_STORAGE_PREFIX + id)
+			if (!raw) return null
+			return JSON.parse(raw) as { x: number; y: number; anchor: string }
+		} catch {
+			return null
+		}
+	}
+
+	savePosition(id: string, data: { x: number; y: number; anchor: string }): void {
+		try {
+			localStorage.setItem(STORAGE_PREFIX + id, JSON.stringify(data))
+		} catch {
+			// ignore
+		}
+	}
+
+	clearPosition(id: string): void {
+		try {
+			localStorage.removeItem(STORAGE_PREFIX + id)
+			localStorage.removeItem(LEGACY_STORAGE_PREFIX + id)
+		} catch {
+			// ignore
+		}
+	}
+
+	// ── Internal ───────────────────────────────────────────────────────
+
+	private _updateRecord(id: string, updates: Partial<WindowRecord>): void {
+		const state = this._state$.value
+		const record = state.windows.get(id)
+		if (!record) return
+
+		const windows = new Map(state.windows)
+		windows.set(id, { ...record, ...updates })
+		this._state$.next({ ...state, windows })
+	}
+}
+
+/** Simple rect overlap check (inline — no circular import with window-position) */
+function rectsOverlap(a: WindowBounds, b: WindowBounds): boolean {
+	return !(a.left >= b.left + b.width || a.left + a.width <= b.left || a.top >= b.top + b.height || a.top + a.height <= b.top)
+}
+
+export const windowManager = WindowManagerService.getInstance()

package/src/window/window-position.ts

@@ -0,0 +1,136 @@
+/**
+ * Window Position — pure functions for overlap resolution, cascade, and snapping.
+ *
+ * All functions are stateless: they take window bounds and viewport dimensions,
+ * return new bounds. No side effects, no DOM access, no subscriptions.
+ *
+ * Example:
+ *   const projected = { left: 100, top: 400, width: 360, height: 500 }
+ *   const neighbors = windowManager.findOverlaps(projected, 'my-id')
+ *   const resolved = resolveOverlap(projected, neighbors, { width: 1440, height: 900 })
+ *   // → { left: 144, top: 444, width: 360, height: 500 } (cascaded away from conflict)
+ */
+
+import type { WindowBounds, WindowRecord, SnapCorner } from './window-registry.js'
+
+const HEAD_HEIGHT = 44
+const EDGE_GAP = 16
+const CASCADE_OFFSET = HEAD_HEIGHT
+const MAX_CASCADE_ATTEMPTS = 10
+
+/** Check if two rectangles overlap (edges touching does NOT count as overlap) */
+export function rectsOverlap(a: WindowBounds, b: WindowBounds): boolean {
+	return !(a.left >= b.left + b.width || a.left + a.width <= b.left || a.top >= b.top + b.height || a.top + a.height <= b.top)
+}
+
+/**
+ * Resolve overlap by cascading diagonally until no conflict.
+ * Shifts by (HEAD_HEIGHT, HEAD_HEIGHT) per attempt, clamped to viewport.
+ */
+export function resolveOverlap(
+	candidate: WindowBounds,
+	existing: WindowRecord[],
+	viewport: { width: number; height: number },
+): WindowBounds {
+	let bounds = { ...candidate }
+
+	for (let attempt = 0; attempt < MAX_CASCADE_ATTEMPTS; attempt++) {
+		const hasOverlap = existing.some(w => rectsOverlap(bounds, w.bounds))
+		if (!hasOverlap) break
+
+		bounds = {
+			...bounds,
+			left: bounds.left + CASCADE_OFFSET,
+			top: bounds.top + CASCADE_OFFSET,
+		}
+	}
+
+	return clampToViewport(bounds, viewport)
+}
+
+/**
+ * Snap window bounds to a corner with standard edge gap.
+ */
+export function snapToCorner(
+	bounds: WindowBounds,
+	corner: SnapCorner,
+	viewport: { width: number; height: number },
+	bottomOffset: number,
+): WindowBounds {
+	const isRight = corner.includes('right')
+	const isBottom = corner.includes('bottom')
+
+	return {
+		width: bounds.width,
+		height: bounds.height,
+		left: isRight ? viewport.width - bounds.width - EDGE_GAP : EDGE_GAP,
+		top: isBottom ? viewport.height - bounds.height - EDGE_GAP - bottomOffset : EDGE_GAP,
+	}
+}
+
+/**
+ * Place a window flush beside a neighbor.
+ */
+export function snapToNeighbor(
+	bounds: WindowBounds,
+	neighbor: WindowBounds,
+	side: 'left' | 'right',
+): WindowBounds {
+	return {
+		...bounds,
+		left: side === 'right' ? neighbor.left + neighbor.width + EDGE_GAP : neighbor.left - bounds.width - EDGE_GAP,
+		top: neighbor.top,
+	}
+}
+
+/**
+ * Clamp bounds so the window stays fully within the viewport.
+ */
+export function clampToViewport(
+	bounds: WindowBounds,
+	viewport: { width: number; height: number },
+): WindowBounds {
+	return {
+		width: bounds.width,
+		height: bounds.height,
+		left: Math.max(0, Math.min(bounds.left, viewport.width - bounds.width)),
+		top: Math.max(0, Math.min(bounds.top, viewport.height - bounds.height)),
+	}
+}
+
+/**
+ * Convert corner-relative offset (used by the old float) to absolute viewport bounds.
+ */
+export function cornerOffsetToAbsolute(
+	corner: SnapCorner,
+	offset: { x: number; y: number },
+	size: { width: number; height: number },
+	viewport: { width: number; height: number },
+	bottomOffset: number,
+): WindowBounds {
+	const isRight = corner.includes('right')
+	const isBottom = corner.includes('bottom')
+	return {
+		width: size.width,
+		height: size.height,
+		left: isRight ? viewport.width - offset.x - size.width : offset.x,
+		top: isBottom ? viewport.height - offset.y - size.height - bottomOffset : offset.y,
+	}
+}
+
+/**
+ * Convert absolute viewport bounds back to corner-relative offset.
+ */
+export function absoluteToCornerOffset(
+	corner: SnapCorner,
+	bounds: WindowBounds,
+	viewport: { width: number; height: number },
+	bottomOffset: number,
+): { x: number; y: number } {
+	const isRight = corner.includes('right')
+	const isBottom = corner.includes('bottom')
+	return {
+		x: isRight ? viewport.width - bounds.left - bounds.width : bounds.left,
+		y: isBottom ? viewport.height - bounds.top - bounds.height - bottomOffset : bounds.top,
+	}
+}