make_tables.py

"""
make_tables.py --- regenerate the LaTeX body rows of the manuscript tables.

Reads the four raw batch_idemo log files and the ablation summary JSON
produced by `aggregate.py`, and emits a single .tex file with the body
rows for every per-instance and per-family table in Sections 6.1, 6.2,
and 6.3 of the manuscript.

The emitted .tex is a drop-in replacement for the body rows --- the
caller is expected to wrap them in the table/tabular environments
already present in the manuscript source.

Usage:

    python make_tables.py \
        --frb-log     experiment/npbench/benchmarks_npbench/benchmarks_npbench.txt \
        --dimacs-log  experiment/npbench/clique_complement_npbench/clique_complement_npbench.txt \
        --network-log experiment/network/network.txt \
        --hardest-log experiment/hardest/hardest.txt \
        --ablation    ablation_summary.json \
        --out         tables.tex
"""

import argparse
import json
import re
import sys
from collections import defaultdict


def parse_log(path):
    """Parse a batch_idemo log into a list of {name, algo_ms, parse_ms, size}."""
    if not path:
        return []
    rows = []
    cur_parse = None
    cur_algo = None
    with open(path) as f:
        for line in f:
            m = re.search(r"Parsing the Input File done in: ([\d.]+) milliseconds", line)
            if m:
                cur_parse = float(m.group(1))
                continue
            m = re.search(r"Our Algorithm with an approximate solution done in: ([\d.]+) milliseconds", line)
            if m:
                cur_algo = float(m.group(1))
                continue
            m = re.search(r"INFO - ([\w./\\-]+\.[\w.-]+): Vertex Cover Size (\d+)", line)
            if m:
                rows.append({"name": m.group(1), "parse_ms": cur_parse,
                             "algo_ms": cur_algo, "size": int(m.group(2))})
                cur_parse = None
                cur_algo = None
    return rows


def fmt_time(ms):
    if ms is None:
        return "--"
    if ms < 1000:
        return "{:.1f}ms".format(ms)
    return "{:.2f}s".format(ms / 1000)


def emit_per_instance_rows(rows, out, label):
    """Emit one LaTeX row per instance: name & size & time \\ (without OPT
    or ratio columns, which require a reference value the caller supplies)."""
    out.write("% {} -- {} rows\n".format(label, len(rows)))
    for r in rows:
        out.write("{} & {} & {} \\\\\n".format(
            r["name"].replace("_", r"\_"), r["size"], fmt_time(r["algo_ms"])))


def emit_ablation_rows(summary, out, families, label):
    """Emit one LaTeX row per family for the ablation table."""
    out.write("\n% Ablation -- {}\n".format(label))
    for fam in families:
        if fam not in summary["per_family"]:
            continue
        d = summary["per_family"][fam]
        mean = d["mean"]
        pct = d["min_pct"]
        fam_disp = fam.replace("_", r"\_")
        out.write(
            "\\texttt{{{}}} & {} & {:.1f} & {:.1f} & {:.1f} & {:.1f} & "
            "{:.0f} & {:.0f} & {:.0f} & {:.0f} \\\\\n".format(
                fam_disp, d["n"], mean["c1"], mean["c2"], mean["c3"], mean["c4"],
                pct["c1"], pct["c2"], pct["c3"], pct["c4"]))
    o = summary.get("overall")
    if o:
        out.write(
            "\\midrule\n\\textbf{{Overall}} & \\textbf{{{}}} & "
            "\\textbf{{{:.1f}}} & \\textbf{{{:.1f}}} & \\textbf{{{:.1f}}} & \\textbf{{{:.1f}}} & "
            "\\textbf{{{:.0f}}} & \\textbf{{{:.0f}}} & \\textbf{{{:.0f}}} & \\textbf{{{:.0f}}} \\\\\n".format(
                o["n"], o["mean"]["c1"], o["mean"]["c2"], o["mean"]["c3"], o["mean"]["c4"],
                o["min_pct"]["c1"], o["min_pct"]["c2"], o["min_pct"]["c3"], o["min_pct"]["c4"]))


NPB_FAMILIES = [
    "FRB", "brock200", "brock400", "brock800",
    "C125", "C250", "C500", "C1000",
    "c-fat200", "c-fat500", "gen200", "gen400",
    "hamming6", "hamming8", "hamming10",
    "johnson8", "johnson16", "johnson32",
    "keller4", "keller5",
    "MANN_a9", "MANN_a27", "MANN_a45", "MANN_a81",
    "p_hat300", "p_hat500", "p_hat700", "p_hat1000",
    "san200", "san400", "sanr200", "sanr400",
]
REALWORLD_FAMILIES = [
    "Bio", "Collab", "Email", "Infect/Reality", "Infra", "Rec",
    "Retweet", "SCC", "SocFB", "Social", "Tech", "Web", "Wiki",
]


def main():
    p = argparse.ArgumentParser(description=__doc__)
    p.add_argument("--frb-log")
    p.add_argument("--dimacs-log")
    p.add_argument("--network-log")
    p.add_argument("--hardest-log")
    p.add_argument("--ablation")
    p.add_argument("--out", required=True)
    args = p.parse_args()

    frb = parse_log(args.frb_log)
    dim = parse_log(args.dimacs_log)
    net = parse_log(args.network_log)
    hrd = parse_log(args.hardest_log)

    summary = None
    if args.ablation:
        with open(args.ablation) as f:
            summary = json.load(f)

    with open(args.out, "w") as out:
        out.write("% Auto-generated by make_tables.py -- do not edit by hand.\n\n")
        emit_per_instance_rows(frb, out, "FRB (Section 6.1, tab:npbench-frb)")
        out.write("\n")
        emit_per_instance_rows(dim, out, "DIMACS (Section 6.1, tab:npbench-dimacs-*)")
        out.write("\n")
        emit_per_instance_rows(net, out, "Real-world (Section 6.2, tab:realworld)")
        out.write("\n")
        emit_per_instance_rows(hrd, out, "Adversarial (Section 6.3, tab:hardest)")

        if summary is not None:
            emit_ablation_rows(summary, out, NPB_FAMILIES,
                               "NPBench per-family (Section 6.4, tab:ablation-npbench)")
            emit_ablation_rows(summary, out, REALWORLD_FAMILIES,
                               "Real-world per-family (Section 6.4, tab:ablation-realworld)")

    print("Wrote {}: FRB={} DIMACS={} Real-world={} Adversarial={}".format(
        args.out, len(frb), len(dim), len(net), len(hrd)), file=sys.stderr)


if __name__ == "__main__":
    main()