Spaces:
Sleeping
Sleeping
File size: 14,597 Bytes
da6a0a4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 | # search_engine_barrels.py
# BARREL-OPTIMIZED SEARCH ENGINE - Loads only required barrels per query
import csv
import json
import math
import os
import re
import time
from collections import defaultdict
# ---------- CONFIG & PATHS ----------
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
PROJECT_ROOT = os.path.dirname(SCRIPT_DIR)
INDEX_DIR = os.path.join(PROJECT_ROOT, "data", "index")
BARREL_DIR = os.path.join(INDEX_DIR, "barrels")
LEXICON_PATH = os.path.join(INDEX_DIR, "lexicon_complete.json")
FORWARD_INDEX_PATH = os.path.join(INDEX_DIR, "forward_index_termid.json")
TERM_TO_BARREL_MAP_PATH = os.path.join(BARREL_DIR, "term_to_barrel_map.json")
MARKET_VALUE_PATH = os.path.join(os.path.dirname(__file__), "..", "data", "raw", "player_latest_market_value", "player_latest_market_value.csv")
PROFILE_DATA_PATH = os.path.join(os.path.dirname(__file__), "..", "data", "processed", "complete_player_profiles.json")
# BM25 parameters
K1 = 1.2
B = 0.75
# Scoring boosts
NAME_TOKEN_WEIGHT = 0.75
NAME_PREFIX_BONUS = 1.25
EXACT_NAME_BONUS = 3.0
RAW_SUBSTRING_BONUS = 0.25
MARKET_VALUE_WEIGHT = 12.0
PROFILE_LENGTH_WEIGHT = 4.0
NON_NAME_MATCH_PENALTY = 1.5
# ---------- TEXT NORMALIZATION ----------
COMPREHENSIVE_STOP_WORDS = {
"the", "and", "in", "for", "with", "on", "at", "from", "by", "as", "is", "was",
"are", "were", "be", "been", "have", "has", "had", "to", "of", "a", "an", "that",
"this", "these", "those", "it", "its", "or", "but", "not", "what", "which", "who",
"when", "where", "why", "how", "all", "any", "both", "each", "few", "more", "most",
"other", "some", "such", "no", "nor", "only", "own", "same", "so", "than", "too",
"very", "can", "will", "just", "should", "now", "player", "club", "team", "football",
"soccer", "match", "game", "season", "league", "cup", "champions", "premier", "la",
"bundesliga", "serie", "current", "main", "position", "nationality", "birth", "place",
# Universal terms that appear in ALL documents (filtering for memory/performance)
"comprehensive", "international", "performance", "transfermarkt", "injury",
"summary", "market", "history", "database", "value",
# Stemmed versions and other universal terms
"data", "teammat", "sourc", "career", "assist", "app", "minut",
"available", "national", "significant", "teammate", "transfer", "goal"
}
def simple_stemmer(word: str) -> str:
if word.endswith("ing") and len(word) > 5:
return word[:-3]
elif word.endswith("ed") and len(word) > 4:
return word[:-2]
elif word.endswith("es") and len(word) > 4:
return word[:-2]
elif word.endswith("s") and len(word) > 3:
return word[:-1]
return word
def normalize_and_tokenize(text: str):
text = text.lower()
tokens = re.findall(r"\b[a-z]+\b", text)
result = []
for w in tokens:
if w in COMPREHENSIVE_STOP_WORDS or len(w) <= 2:
continue
result.append(simple_stemmer(w))
return result
def normalize_name_tokens(value: str):
if not isinstance(value, str):
return []
tokens = re.findall(r"[a-z]+", value.lower())
return [simple_stemmer(tok) for tok in tokens if tok]
def build_name_metadata(name: str):
tokens = normalize_name_tokens(name)
token_set = set(tokens)
normalized = " ".join(tokens)
return {
"tokens": tokens,
"token_set": token_set,
"normalized": normalized,
"raw_lower": name.lower() if isinstance(name, str) else "",
}
def load_market_values(path: str):
values = {}
try:
with open(path, "r", encoding="utf-8") as handle:
reader = csv.DictReader(handle)
for row in reader:
try:
player_id = int(row.get("player_id", ""))
except (TypeError, ValueError):
continue
raw_value = row.get("value")
try:
value = float(raw_value)
except (TypeError, ValueError):
continue
date_key = row.get("date_unix", "") or ""
current = values.get(player_id)
if current is None or date_key > current[0]:
values[player_id] = (date_key, value)
except FileNotFoundError:
print(f"[warn] Market value file not found at {path}")
return {}
return {pid: info[1] for pid, info in values.items()}
def load_profile_lengths(path: str):
try:
with open(path, "r", encoding="utf-8") as handle:
data = json.load(handle)
except FileNotFoundError:
print(f"[warn] Profile data file not found at {path}")
return {}
lengths = {}
for entry in data:
player_id = entry.get("player_id")
if not isinstance(player_id, int):
continue
detailed = entry.get("detailed_content")
if isinstance(detailed, str) and detailed:
lengths[player_id] = len(detailed)
return lengths
# ---------- LOAD STATIC INDEXES (NOT INVERTED INDEX) ----------
print("[init] Loading lexicon...")
with open(LEXICON_PATH, "r", encoding="utf-8") as f:
lexicon_entries = json.load(f)
token_to_id = {entry["token"]: entry["term_id"] for entry in lexicon_entries}
termid_to_token = {entry["term_id"]: entry["token"] for entry in lexicon_entries}
term_document_frequency = {entry["term_id"]: entry["df"] for entry in lexicon_entries}
print(f"[done] Lexicon loaded: {len(token_to_id):,} tokens")
print("[init] Loading forward index...")
with open(FORWARD_INDEX_PATH, "r", encoding="utf-8") as f:
forward_index = json.load(f)
doc_by_id = {doc["player_id"]: doc for doc in forward_index}
N = len(doc_by_id)
avg_doc_len = sum(d["total_terms"] for d in forward_index) / N if N > 0 else 0.0
name_metadata = {doc_id: build_name_metadata(doc.get("player_name"))
for doc_id, doc in doc_by_id.items()}
print(f"[done] Forward index: {N:,} documents (avg_len={avg_doc_len:.2f})")
print("[init] Loading term-to-barrel mapping...")
with open(TERM_TO_BARREL_MAP_PATH, "r", encoding="utf-8") as f:
term_to_barrel = json.load(f)
print(f"[done] Term-to-barrel map loaded: {len(term_to_barrel):,} mappings")
print("[init] Loading market values...")
player_market_value = load_market_values(MARKET_VALUE_PATH)
max_market_value = max(player_market_value.values(), default=0.0)
market_value_log_max = math.log1p(max_market_value) if max_market_value > 0 else 1.0
print(f"[done] Market values loaded for {len(player_market_value):,} players")
print("[init] Loading profile metadata...")
profile_length_by_id = load_profile_lengths(PROFILE_DATA_PATH)
max_profile_length = max(profile_length_by_id.values(), default=0)
profile_length_log_max = math.log1p(max_profile_length) if max_profile_length > 0 else 1.0
print(f"[done] Profile metadata loaded for {len(profile_length_by_id):,} players")
# ---------- BARREL CACHE (LRU-like) ----------
barrel_cache = {}
MAX_CACHED_BARRELS = 10 # Keep only 10 barrels in memory at once
def load_barrel(barrel_name: str):
"""Load a barrel file and cache it. Implements simple LRU eviction."""
if barrel_name in barrel_cache:
return barrel_cache[barrel_name]
barrel_path = os.path.join(BARREL_DIR, f"{barrel_name}.json")
try:
with open(barrel_path, "r", encoding="utf-8") as f:
barrel_data = json.load(f)
# Cache management
if len(barrel_cache) >= MAX_CACHED_BARRELS:
# Remove oldest (first) entry
oldest_key = next(iter(barrel_cache))
del barrel_cache[oldest_key]
barrel_cache[barrel_name] = barrel_data
return barrel_data
except FileNotFoundError:
print(f"[error] Barrel file not found: {barrel_path}")
return None
# ---------- QUERY TO TERM IDs ----------
def tokens_to_term_ids(tokens):
seen = set()
unique_term_ids = []
for tok in tokens:
tid = token_to_id.get(tok)
if tid is None or tid in seen:
continue
seen.add(tid)
unique_term_ids.append(tid)
return unique_term_ids
# ---------- BM25 SCORING ----------
def bm25_score(tf, df, doc_len, N, avg_doc_len, k1=K1, b=B):
idf = math.log((N - df + 0.5) / (df + 0.5) + 1.0)
denom = tf + k1 * (1 - b + b * (doc_len / avg_doc_len))
return idf * (tf * (k1 + 1) / denom)
# ---------- BARREL-BASED SEARCH ----------
def search(query: str, top_k: int = 10, verbose: bool = True):
start_time = time.perf_counter()
log = print if verbose else (lambda *args, **kwargs: None)
log(f"\n[query] {query}")
query_tokens = normalize_and_tokenize(query)
term_ids = tokens_to_term_ids(query_tokens)
if not term_ids:
elapsed = (time.perf_counter() - start_time) * 1000
log(f"No query terms found in lexicon. (took {elapsed:.2f} ms)")
return []
log("Query tokens -> term_ids:",
[(termid_to_token.get(tid, "?"), tid) for tid in term_ids])
# **KEY OPTIMIZATION: Determine which barrels to load**
required_barrels = set()
for tid in term_ids:
barrel_name = term_to_barrel.get(str(tid))
if barrel_name:
required_barrels.add(barrel_name)
log(f"[barrels] Loading {len(required_barrels)} barrel(s): {sorted(required_barrels)}")
# Load only required barrels
barrel_load_start = time.perf_counter()
loaded_barrels = {}
for barrel_name in required_barrels:
barrel_data = load_barrel(barrel_name)
if barrel_data:
loaded_barrels[barrel_name] = barrel_data
barrel_load_time = (time.perf_counter() - barrel_load_start) * 1000
log(f"[barrels] Loaded in {barrel_load_time:.2f} ms")
# BM25 scoring using barrel data
scores = defaultdict(float)
for tid in term_ids:
df = term_document_frequency.get(tid, 0)
if df == 0:
continue
# Get barrel for this term
barrel_name = term_to_barrel.get(str(tid))
if not barrel_name or barrel_name not in loaded_barrels:
continue
barrel_data = loaded_barrels[barrel_name]
inverted_index_part = barrel_data.get("inverted_index", {})
# Get postings for this term
term_data = inverted_index_part.get(str(tid))
if not term_data:
continue
postings = term_data.get("postings", {})
for doc_id_str, info in postings.items():
doc_id = int(doc_id_str)
tf = info["tf"]
doc_len = doc_by_id[doc_id]["total_terms"]
scores[doc_id] += bm25_score(tf, df, doc_len, N, avg_doc_len)
# Metadata boosting (same as before)
if scores:
query_name_tokens = normalize_name_tokens(query)
query_name = " ".join(query_name_tokens)
raw_query_lower = query.lower().strip()
for doc_id in scores:
boost = 0.0
meta = name_metadata.get(doc_id)
has_name_match = False
match_count = 0
if meta:
if query_tokens:
match_count = sum(1 for tok in query_tokens if tok in meta["token_set"])
if match_count:
boost += NAME_TOKEN_WEIGHT * match_count
has_name_match = True
if query_name:
if meta["normalized"] == query_name:
boost += EXACT_NAME_BONUS
has_name_match = True
elif meta["normalized"].startswith(query_name):
boost += NAME_PREFIX_BONUS
has_name_match = True
if raw_query_lower and raw_query_lower in meta["raw_lower"]:
boost += RAW_SUBSTRING_BONUS
has_name_match = True
if not has_name_match and query_tokens:
boost -= NON_NAME_MATCH_PENALTY
if has_name_match:
value = player_market_value.get(doc_id)
if value and market_value_log_max > 0.0:
boost += MARKET_VALUE_WEIGHT * (math.log1p(value) / market_value_log_max)
length = profile_length_by_id.get(doc_id)
if length and profile_length_log_max > 0.0:
boost += PROFILE_LENGTH_WEIGHT * (math.log1p(length) / profile_length_log_max)
scores[doc_id] += boost
if not scores:
elapsed = (time.perf_counter() - start_time) * 1000
log(f"No documents matched these terms. (took {elapsed:.2f} ms)")
return []
ranked = sorted(scores.items(), key=lambda x: x[1], reverse=True)[:top_k]
results = []
for rank, (doc_id, score) in enumerate(ranked, start=1):
doc = doc_by_id[doc_id]
results.append({
"rank": rank,
"doc_id": doc_id,
"player_id": doc["player_id"],
"player_name": doc["player_name"],
"score": score,
"market_value": player_market_value.get(doc_id),
})
elapsed = (time.perf_counter() - start_time) * 1000
log("\n[top] Results:")
for r in results:
value = r["market_value"]
length = profile_length_by_id.get(r["doc_id"])
extras = []
if value:
extras.append(f"market_value~{value:,.0f} EUR")
if length:
extras.append(f"profile_chars={length}")
extra_text = f" [{', '.join(extras)}]" if extras else ""
log(f"{r['rank']:2d}. [{r['score']:.3f}] {r['player_name']} (player_id={r['player_id']}){extra_text}")
log(f"\n[time] {elapsed:.2f} ms (barrel_load={barrel_load_time:.2f} ms)")
log(f"[memory] {len(barrel_cache)} barrels cached, {len(required_barrels)} loaded for this query")
if elapsed < 500:
log("[perf]Under 500 ms goal")
else:
log("[perf]Above 500 ms goal")
return results
# ---------- CLI ----------
if __name__ == "__main__":
print("\n[ready] BARREL-OPTIMIZED search engine ready.")
print(f"[info] System loads only required barrels per query (max {MAX_CACHED_BARRELS} cached)")
print("[info] Type a query or press Enter to exit.\n")
while True:
q = input("Query> ").strip()
if not q:
break
search(q, top_k=10)
print("\n[exit] Exiting search engine.")
|