1"""Derive 5-minute (or other) bars from validated 1-minute bars, per session_date (delta layout, supersede)."""2from __future__ import annotations34from datetime import date5from typing import Any67import pyarrow as pa8from findata_core.db import connection, transaction9from findata_quality import aggregate_bars10from findata_schema import Interval1112from findata_ingest.lake import Lake, live_files13from findata_ingest.reader import read_bars14from findata_ingest.tasks.common import live_delta_dates, live_delta_paths_for_dates, upsert_coverage151617def run(job: dict[str, Any], lake: Lake, log) -> dict[str, Any]:18 p = job.get("params") or {}19 src = p["source_dataset"]20 dst = job.get("dataset_id") or src.replace(".1m.", ".5m.")21 target = Interval(p.get("target_interval", "5m"))22 dates = [date.fromisoformat(d) for d in p.get("dates", [])]23 symbols = p.get("symbols")24 run_id = job.get("run_id")25 with connection() as conn:26 if not dates:27 # all live source dates for requested symbols28 fl = live_files(conn, src, symbols)29 ds = set()30 for f in fl:31 d = f["date_min"]32 while d <= f["date_max"]:33 ds.add(d)34 d = d.fromordinal(d.toordinal() + 1)35 dates = sorted(ds)36 all_syms = symbols or [r["symbol"] for r in conn.execute("SELECT DISTINCT symbol FROM coverage WHERE dataset_id=%s AND session_date = ANY(%s)", (src, dates)).fetchall()]37 written, rows = [], 038 tables = {}39 if symbols:40 # backfill path: aggregate per symbol over all its sessions and write base files per (symbol, year)41 import pyarrow.compute as pc42 retire: list[str] = []43 for s in sorted(set(x.upper() for x in symbols)):44 t, _ = read_bars(src, [s], None, None, lake=lake)45 if t.num_rows == 0:46 continue47 agg = aggregate_bars(t, target, Interval.m1)48 years = pc.year(agg.column("session_date"))49 with connection() as conn:50 old = conn.execute("SELECT path FROM lake_files WHERE dataset_id=%s AND layout='base' AND version_removed IS NULL AND symbol_min=%s AND symbol_max=%s",51 (dst, s, s)).fetchall()52 delta_dates = live_delta_dates(conn, dst)53 if delta_dates:54 agg = agg.filter(pc.invert(pc.is_in(agg.column("session_date"), value_set=pa.array(sorted(delta_dates), pa.date32()))))55 years = pc.year(agg.column("session_date"))56 for y in sorted(set(years.to_pylist())):57 ty = agg.filter(pc.equal(years, y))58 written.append(lake.write_table(dst, ty, "base", {"symbol": s, "year": str(y)}))59 retire.extend(o["path"] for o in old)60 rows += agg.num_rows61 with transaction() as conn:62 upsert_coverage(conn, dst, agg, target)63 del t, agg64 import gc65 gc.collect()66 pa.default_memory_pool().release_unused()67 if not written:68 return {"rows": 0, "symbols": len(symbols)}69 pub = lake.publish(dst, written, sorted(set(retire)), run_id, note=f"derive {target.value} (base) from {src} for {len(symbols)} symbols")70 log.info("derive.done", rows=rows, files=len(written), **pub)71 return {"rows": rows, "files": len(written), "symbols": len(symbols), **pub}72 for d in dates:73 t, _ = read_bars(src, all_syms, d, d, lake=lake)74 if t.num_rows == 0:75 continue76 agg = aggregate_bars(t, target, Interval.m1)77 tables[d] = agg78 rows += agg.num_rows79 written.append(lake.write_table(dst, agg, "delta", {"session_date": d.isoformat()}))80 if not written:81 return {"rows": 0, "dates": len(dates)}82 with connection() as conn:83 retire = live_delta_paths_for_dates(conn, dst, list(tables)) if not symbols else []84 pub = lake.publish(dst, written, retire, run_id, note=f"derive {target.value} from {src}")85 with transaction() as conn:86 for t in tables.values():87 upsert_coverage(conn, dst, t, target)88 log.info("derive.done", rows=rows, files=len(written), **pub)89 return {"rows": rows, "files": len(written), "dates": len(tables), **pub}90