scripts/make_catalog.py

"""Write sessions.csv -- one row per session: its conditions, how it was captured, what was measured,
and the capture analysis (column groups in lvl_running/catalog.py) -- and mount a compact Markdown
version of it in README.md (between the sessions:start / sessions:end markers), so the dataset card
never goes stale.

    python scripts/make_catalog.py

Always rebuilt in full: it is small, text, and diffs cleanly release to release.
"""
from __future__ import annotations

import os
import re

import pandas as pd

from lvl_running.catalog import build_catalog

HERE = os.path.dirname(os.path.abspath(__file__))
ROOT = os.path.dirname(HERE)                     # the dataset folder (scripts/ lives inside it)
MARKERS = re.compile(r"(<!-- sessions:start[^>]*-->).*?(<!-- sessions:end -->)", re.S)
GLANCE = re.compile(r"(<!-- glance:start[^>]*-->).*?(<!-- glance:end -->)", re.S)


def _v(x) -> str:
    return "" if x is None or (isinstance(x, float) and pd.isna(x)) else str(x)


def _num(x, fmt: str) -> str:
    return "" if _v(x) == "" else fmt.format(float(x))


def sessions_markdown(df: pd.DataFrame) -> str:
    """The card's session table: what a visitor needs at a glance; sessions.csv has every column."""
    head = ("| Date | Session | Conditions | Data (min) | Run (min) | Distance (km) | LEVEL sensors | Reference data | Packet loss (%) | Eff. rate (Hz) | Max gap (s) |\n"
            "|---|---|---|---:|---:|---:|---:|---|---:|---:|---:|")
    rows = []
    for r in df.to_dict("records"):
        overview = f"figures/{r['subject']}/{r['session']}/overview.png"
        cond = _v(r.get("setting"))            # indoor / outdoor; surface, event, location are in sessions.csv
        sensors = _v(r.get("sensor_count"))    # placements are in sessions.csv
        refs = ", ".join(n for n, ok in (("chest strap", r.get("h10_hr")), ("watch", r.get("watch"))) if ok is True)
        cells = [_v(r["date"]), f"[{r['session']}]({overview})", cond, _num(r.get("duration_min"), "{:.0f}"), _num(r.get("run_duration_min"), "{:.0f}"),
                 _num(r.get("distance_km"), "{:.2f}"), sensors, refs, _num(r.get("loss_pct"), "{:.1f}"),
                 _num(r.get("effective_rate_hz"), "{:.1f}"), _num(r.get("max_gap_s"), "{:.2f}")]
        rows.append("| " + " | ".join(c.replace("|", "/") for c in cells) + " |")
    return (f"{head}\n" + "\n".join(rows) + f"\n\n*{len(df)} sessions. Click a session for its overview figure; "
            "every column (setup, versions, heart rate, capture quality) is in `sessions.csv`.*")


def glance(df: pd.DataFrame) -> str:
    """One line of totals: what the dataset is, at a glance."""
    n, subs = len(df), df["subject"].nunique()
    hours = df["duration_min"].fillna(0).sum() / 60
    km = df["distance_km"].fillna(0).sum()
    d0, d1 = df["date"].min(), df["date"].max()
    return (f"**{n} session{'s' if n != 1 else ''}, {subs} participant{'s' if subs != 1 else ''}, "
            f"{hours:.1f} h of recording, {km:.0f} km outdoors, {d0} to {d1}.**")


def mount(readme_path: str, table: str, glance_line: str) -> bool:
    """Replace the generated blocks between their markers in README.md; False if a marker is missing."""
    text = open(readme_path, encoding="utf-8").read()
    if not (MARKERS.search(text) and GLANCE.search(text)):
        return False
    new = MARKERS.sub(lambda m: f"{m.group(1)}\n{table}\n{m.group(2)}", text)
    new = GLANCE.sub(lambda m: f"{m.group(1)}\n{glance_line}\n{m.group(2)}", new)
    with open(readme_path, "w", encoding="utf-8", newline="\n") as f:   # same bytes on every OS
        f.write(new)
    return True


if __name__ == "__main__":
    df = build_catalog(os.path.join(ROOT, "data"))
    df.to_csv(os.path.join(ROOT, "sessions.csv"), index=False)
    ok = mount(os.path.join(ROOT, "README.md"), sessions_markdown(df), glance(df))
    print(f"{len(df)} sessions -> sessions.csv" + ("; README.md table updated" if ok else "; README.md markers NOT found"))