|
|
from pathlib import Path
|
|
|
import pyarrow.parquet as pq
|
|
|
import pandas as pd
|
|
|
|
|
|
def summarize_parquet_metadata(path_str: str) -> pd.DataFrame:
|
|
|
path = Path(path_str)
|
|
|
files = [path] if path.is_file() else list(path.rglob("*.parquet"))
|
|
|
cols = {}
|
|
|
for f in files:
|
|
|
pf = pq.ParquetFile(f)
|
|
|
schema = pf.schema_arrow
|
|
|
logical_types = {name: schema.field(i).type for i, name in enumerate(schema.names)}
|
|
|
md = pf.metadata
|
|
|
for rg_idx in range(md.num_row_groups):
|
|
|
rg = md.row_group(rg_idx)
|
|
|
for c_idx in range(rg.num_columns):
|
|
|
c_md = rg.column(c_idx)
|
|
|
name = c_md.path_in_schema
|
|
|
stats = c_md.statistics
|
|
|
if stats is None:
|
|
|
|
|
|
continue
|
|
|
entry = cols.setdefault(name, {
|
|
|
"dtype": str(logical_types.get(name, "")),
|
|
|
"min": None, "max": None, "nulls": 0, "rows": 0
|
|
|
})
|
|
|
entry["nulls"] += stats.null_count or 0
|
|
|
entry["rows"] += rg.num_rows
|
|
|
vmin, vmax = stats.min, stats.max
|
|
|
|
|
|
if isinstance(vmin, (bytes, bytearray)):
|
|
|
try: vmin = vmin.decode("utf-8", "replace")
|
|
|
except Exception: pass
|
|
|
if isinstance(vmax, (bytes, bytearray)):
|
|
|
try: vmax = vmax.decode("utf-8", "replace")
|
|
|
except Exception: pass
|
|
|
if entry["min"] is None or (vmin is not None and vmin < entry["min"]):
|
|
|
entry["min"] = vmin
|
|
|
if entry["max"] is None or (vmax is not None and vmax > entry["max"]):
|
|
|
entry["max"] = vmax
|
|
|
|
|
|
if not cols:
|
|
|
return pd.DataFrame(columns=["dtype", "min", "max", "nulls", "rows"])
|
|
|
return pd.DataFrame.from_dict(cols, orient="index") \
|
|
|
.rename_axis("column") \
|
|
|
.sort_index()
|
|
|
|
|
|
summary = summarize_parquet_metadata("fr-astrotemp-1h_1900-12_to_2025-08_norm_v1.parquet")
|
|
|
print(summary)
|
|
|
|
