Spaces:
Sleeping
Sleeping
File size: 13,207 Bytes
b336134 | 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 | """
MVCC Executor.
Performs lock-free versioned operations on Parquet files using Polars LazyFrames.
"""
from __future__ import annotations
import os
import polars as pl
from config import DATA_DIR, LAZY_THRESHOLD_BYTES
from services.session_manager import session_manager
from core.column_registry import column_registry
from core.execution.polars_builder import build_polars_expression
def _build_predicate(column: str, condition: str, value) -> pl.Expr:
"""Build a Polars expression from a comparison condition string."""
col = pl.col(column)
if condition == ">":
return col > value
elif condition == "<":
return col < value
elif condition == ">=":
return col >= value
elif condition == "<=":
return col <= value
elif condition == "!=":
return col != value
elif condition == "==" or condition == "=":
if isinstance(value, (int, float)):
return col == value
try:
return col == float(value)
except (ValueError, TypeError):
return col.cast(pl.String).str.to_lowercase() == str(value).lower()
elif condition in ("contains", "~"):
return col.cast(pl.String).str.contains(str(value), literal=True)
else:
return col.cast(pl.String).str.to_lowercase() == str(value).lower()
def _format_number(val) -> str:
"""Pretty-print numbers with commas."""
if val is None:
return "N/A"
if isinstance(val, float):
if val == int(val):
return f"{int(val):,}"
return f"{val:,.2f}"
if isinstance(val, int):
return f"{val:,}"
return str(val)
def execute_mvcc_command(session_id: str, intent: dict) -> dict:
"""Execute the command and return a result dict.
Updates metadata and creates a new parquet version on write.
"""
meta = session_manager.get(session_id)
if meta is None:
return {
"status": "error",
"message": "Session nahi mila. Pehle file upload karo.",
"diff": {}
}
version = getattr(meta, "current_version", 0)
# Resolve read path
# If version is 0, we can fall back to the initial upload path (data/sessions/{session_id}.parquet).
v_path = os.path.join(DATA_DIR, f"{session_id}_v{version}.parquet")
if not os.path.exists(v_path):
if version == 0:
v_path = os.path.join(DATA_DIR, f"{session_id}.parquet")
if not os.path.exists(v_path):
raise FileNotFoundError(f"Initial Parquet file not found for session {session_id}")
else:
raise FileNotFoundError(f"Parquet file not found for version {version} of session {session_id}")
op = intent.get("operation")
if not op:
return {
"message": "Operation specifies nahi kiya gaya",
"diff": {}
}
# ββ Read-only operations (Aggregates) ββ
if op in ("sum", "average", "count", "min", "max"):
lf = pl.scan_parquet(v_path)
if op == "sum":
col = intent["column"]
result = lf.select(pl.col(col).sum()).collect().item()
return {
"message": f"{col} ka sum = {_format_number(result)}",
"diff": {"operation": "sum", "column": col, "result": result},
}
elif op == "average":
col = intent["column"]
result = lf.select(pl.col(col).mean()).collect().item()
return {
"message": f"{col} ka average = {_format_number(result)}",
"diff": {"operation": "average", "column": col, "result": result},
}
elif op == "count":
col = intent.get("column")
if col:
result = lf.select(pl.col(col).count()).collect().item()
return {
"message": f"{col} mein {_format_number(result)} non-null values hain",
"diff": {"operation": "count", "column": col, "result": result},
}
else:
result = lf.select(pl.len()).collect().item()
return {
"message": f"Total {_format_number(result)} rows hain",
"diff": {"operation": "count", "column": None, "result": result},
}
elif op == "min":
col = intent["column"]
result = lf.select(pl.col(col).min()).collect().item()
return {
"message": f"{col} ka minimum = {_format_number(result)}",
"diff": {"operation": "min", "column": col, "result": result},
}
elif op == "max":
col = intent["column"]
result = lf.select(pl.col(col).max()).collect().item()
return {
"message": f"{col} ka maximum = {_format_number(result)}",
"diff": {"operation": "max", "column": col, "result": result},
}
# ββ Write operations ββ
lf = pl.scan_parquet(v_path)
new_version = version + 1
new_path = os.path.join(DATA_DIR, f"{session_id}_v{new_version}.parquet")
before_count = lf.select(pl.len()).collect().item()
if op in ("increase", "decrease"):
col = intent["column"]
value = intent.get("value")
if value is None:
value = 10.0
is_percent = intent.get("is_percent", True)
expr = build_polars_expression(op, col, {"value": value, "is_percent": is_percent})
lf = lf.with_columns(expr)
elif op == "cast_type":
col = intent["column"]
target = intent["target_dtype"]
expr = build_polars_expression("cast_type", col, {"target_dtype": target})
lf = lf.with_columns(expr)
elif op == "find_replace":
col = intent["column"]
old_val = intent["old_value"]
new_val = intent["new_value"]
schema = lf.collect_schema()
dtype = schema[col]
is_numeric = dtype in (pl.Float64, pl.Int64, pl.Int32, pl.Float32)
if is_numeric:
try:
float(old_val)
float(new_val)
except (ValueError, TypeError):
return {
"message": f"'{col}' numeric hai, replace values bhi numbers hone chahiye",
"diff": {},
}
expr = build_polars_expression("find_replace", col, {
"old_value": old_val,
"new_value": new_val,
"is_numeric": is_numeric
})
lf = lf.with_columns(expr)
elif op == "delete_column":
col = intent["column"]
lf = lf.drop(col)
elif op == "rename_column":
old = intent["column"]
new = intent["new_name"]
lf = lf.rename({old: new})
elif op == "add_column":
new_col = intent.get("new_column_name") or intent.get("column", "new_col")
default_val = intent.get("value", 0)
lf = lf.with_columns(pl.lit(default_val).alias(new_col))
elif op == "remove_duplicates":
col = intent.get("column")
if col:
lf = lf.unique(subset=[col], keep="first")
else:
lf = lf.unique(keep="first")
elif op == "filter":
col = intent["column"]
condition = intent["condition"]
filter_value = intent["filter_value"]
predicate = _build_predicate(col, condition, filter_value)
lf = lf.filter(predicate)
elif op in ("sort_asc", "sort_desc"):
col = intent["column"]
descending = (op == "sort_desc")
lf = lf.sort(col, descending=descending)
else:
return {
"message": f"'{op}' operation supported nahi hai",
"diff": {},
}
# Lock-free versioned write
prev_size = os.path.getsize(v_path)
try:
if prev_size > LAZY_THRESHOLD_BYTES:
lf.sink_parquet(new_path)
else:
lf.collect().write_parquet(new_path)
except Exception:
# Fallback to eager write if streaming sink is not supported for this query plan (e.g. sort/unique)
lf.collect().write_parquet(new_path)
# Read updated metadata
lf_new = pl.scan_parquet(new_path)
new_schema = lf_new.collect_schema()
after_count = lf_new.select(pl.len()).collect().item()
columns_meta = [{"name": name, "dtype": str(dtype)} for name, dtype in new_schema.items()]
# Format result message and diff
message = ""
diff = {}
if op in ("increase", "decrease"):
col = intent["column"]
value = intent.get("value")
if value is None:
value = 10.0
is_percent = intent.get("is_percent", True)
op_label = f"{value}% badha diya" if op == "increase" else f"{value}% ghata diya"
if not is_percent:
op_label = f"{value} joda" if op == "increase" else f"{value} ghata diya"
message = f"{col} ko {op_label} ({after_count:,} rows updated)"
diff = {
"operation": op,
"column": col,
"value": value,
"is_percent": is_percent,
"affected_rows": after_count,
}
elif op == "cast_type":
col = intent["column"]
target = intent["target_dtype"]
message = f"'{col}' ka type {target} mein change ho gaya"
diff = {"operation": "cast_type", "column": col, "target_dtype": target}
elif op == "find_replace":
col = intent["column"]
old_val = intent["old_value"]
new_val = intent["new_value"]
dtype = new_schema[col]
is_numeric = dtype in (pl.Float64, pl.Int64, pl.Int32, pl.Float32)
if is_numeric:
try:
count = lf_new.filter(pl.col(col) == float(new_val)).select(pl.len()).collect().item()
except (ValueError, TypeError):
count = "N/A"
else:
count = lf_new.filter(pl.col(col).cast(pl.String) == str(new_val)).select(pl.len()).collect().item()
message = f"'{col}' mein '{old_val}' ko '{new_val}' se replace kiya ({count} rows changed)"
diff = {
"operation": "find_replace",
"column": col,
"old_value": old_val,
"new_value": new_val,
"affected_rows": count,
}
elif op == "delete_column":
col = intent["column"]
message = f"'{col}' column delete ho gaya"
diff = {"operation": "delete_column", "column": col}
elif op == "rename_column":
old = intent["column"]
new = intent["new_name"]
message = f"'{old}' ka naam badal ke '{new}' ho gaya"
diff = {"operation": "rename_column", "old_name": old, "new_name": new}
elif op == "add_column":
new_col = intent.get("new_column_name") or intent.get("column", "new_col")
default_val = intent.get("value", 0)
message = f"Naya column '{new_col}' add ho gaya (default: {default_val})"
diff = {"operation": "add_column", "column": new_col}
elif op == "remove_duplicates":
col = intent.get("column")
removed = before_count - after_count
if col:
message = f"'{col}' ke duplicate rows hata diye ({removed} rows removed, {after_count:,} remaining)"
else:
message = f"Duplicate rows hata diye ({removed} rows removed, {after_count:,} remaining)"
diff = {
"operation": "remove_duplicates",
"column": col,
"rows_removed": removed,
"rows_remaining": after_count,
}
elif op == "filter":
col = intent["column"]
condition = intent["condition"]
filter_value = intent["filter_value"]
message = f"{col} {condition} {filter_value} β {after_count:,} rows bach gaye"
diff = {
"operation": "filter",
"column": col,
"condition": condition,
"filter_value": filter_value,
"rows_after": after_count,
}
elif op in ("sort_asc", "sort_desc"):
col = intent["column"]
descending = (op == "sort_desc")
order = "descending (bada se chhota)" if descending else "ascending (chhota se bada)"
message = f"{col} ko {order} sort kiya ({after_count:,} rows)"
diff = {
"operation": "sort",
"column": col,
"descending": descending,
"affected_rows": after_count,
}
# Update session metadata in session_manager
meta.current_version = new_version
meta.row_count = after_count
meta.columns = columns_meta
meta.touch()
# Update column registry
column_registry.register(session_id, [c["name"] for c in columns_meta])
# Clean up the previous intermediate versioned file
if version > 0:
prev_version_path = os.path.join(DATA_DIR, f"{session_id}_v{version}.parquet")
if os.path.exists(prev_version_path):
try:
os.remove(prev_version_path)
except Exception as e:
print(f"[MVCC Executor] Warning cleaning up version {version}: {e}")
return {
"message": message,
"diff": diff,
}
|