Skip to content

search.md

Latest commit

 

History

History
79 lines (55 loc) · 4.57 KB

search.md

File metadata and controls

79 lines (55 loc) · 4.57 KB

Search Module

Overview

Full-text search across Bible verses with token-based intersection, book/testament filtering, and search history.

Key Files

  • Alkitab/src/main/java/yuku/alkitab/base/ac/SearchActivity.kt — Search UI with history autocomplete
  • Alkitab/src/main/java/yuku/alkitab/base/util/SearchEngine.kt — Core search engine (grep-based, token intersection)
  • Alkitab/src/main/java/yuku/alkitab/base/util/QueryTokenizer.kt — Tokenizes queries with quote and plus-sign support
  • Alkitab/src/main/java/yuku/alkitab/base/util/SearchEngineQuery.kt — Data class holding query string and optional book filter (bookIds: SparseBooleanArray?)

Query Tokenization

QueryTokenizer.tokenize(query) splits a query string into tokens using these rules:

  1. Space-separated terms — each term becomes a separate token. All tokens must match (AND logic).

    • Example: love grace → tokens ["love", "grace"]

  2. Quoted phrases — text inside quotes becomes a single token with a + prefix (triggers whole-word matching). Supports ASCII "..." and Unicode curly quotes \u201c...\u201d (and reversed).

    • Example: "in the beginning" → token ["+in the beginning"]

  3. Plus-prefixed words — a + before a word triggers whole-word matching for that word. Multiple consecutive +word tokens are joined into a single multi-word token.

    • Example: +love → token ["+love"] (whole-word: won't match "beloved")
    • Example: +in +the → token ["+in the"] (multi-word phrase)

Matching Modes

The + prefix on a token determines the matching mode:

Token Mode Behavior
love Substring Matches "love", "beloved", "loves", "gloved"
+love Whole-word Matches only "love" surrounded by non-letter/non-digit chars or string boundaries
+"in the beginning" Multi-word phrase Whole-word match of each word in sequence, tolerant of formatting tags and punctuation between words

Multi-word Token Processing

When a +-prefixed token contains multiple words, QueryTokenizer.tokenizeMultiwordToken() splits it on word boundaries ([\p{javaLetterOrDigit}'-]+). The search then uses indexOfWholeMultiword() which:

  • Finds each word as a whole-word match
  • Strips inline formatting tags (@<...@>...@/) between words
  • Skips punctuation between words
  • Requires all words to appear in sequence within the same verse

Search Algorithm

SearchEngine.searchByGrep() performs a grep-based search across all verses:

  1. Tokenize the query via QueryTokenizer.tokenize()
  2. Sort tokens by length (longest first), then alphabetically — this is an optimization so the most selective token is searched first
  3. Remove duplicate tokens
  4. For each token sequentially:
    • Iterate over all books (or the filtered subset from SearchEngineQuery.bookIds)
    • For each book, iterate over all chapters
    • Load chapter text (lowercased, formatting codes intact) via version.loadChapterTextLowercasedWithoutSplit(book, chapter_1)
    • For each verse in the chapter, check if the token matches:
      • Substring tokens: indexOf() on lowercased text
      • Whole-word tokens: indexOfWholeWord() — checks that match boundaries are non-letter/non-digit
      • Multi-word tokens: indexOfWholeMultiword() — sequential word matching with tag/punctuation tolerance
    • Collect matching verse ARIs into an IntArrayList
    • If this is not the first token, intersect the new results with the previous token's results (only verses matching ALL tokens survive)
  5. Return final intersection as the result set

Key detail: the algorithm searches one token at a time across all verses, then intersects results — it does NOT check all tokens per verse. This means the first (longest) token prunes the result set early, making subsequent token searches faster.

Filtering

  • Testament: Old Testament only, New Testament only, or all
  • Single book: Restrict search to a specific Bible book
  • Filters are passed via SearchEngineQuery.bookIds (SparseBooleanArray mapping bookId → included)

Search History

Recent searches are stored as JSON in preferences (Prefkey.searchHistory), bounded to a maximum of 20 entries. New searches are added to the front; duplicates are removed before re-adding. History entries appear as autocomplete suggestions in SearchActivity.

Result Display

Results show verse reference and text with keyword highlighting. TextColorUtil adapts highlight colors to the current theme's background brightness. Users can select results to navigate to that verse in IsiActivity.