summary_id is incorrectly set to -1 for wheatsheaf mean outputs if no losses

[carlfischerjba]

Issue Description

We've run a portfolio with terms such that the insured losses are all zero. This leads to aalcalc, pltcalc and wheatsheaf_aep/oep files that are empty except for the header. That's fine. But the wheatsheaf_mean_aep/oep files are populated with a number of zero rows (fine but inconsistent with the other outputs) and a summary_id which does not match the one in the summary-info file. This causes some of our scripts to fail.

We'll add special handling for this case so there's no urgency to fix at your end, especially if ktools will be rewritten soon.

Steps to Reproduce (Bugs only)

1. Create a portfolio with terms such that the insured losses will be zero.
2. Run an analysis requesting the legacy wheatsheaf_mean EP outputs.
3. Observe the mismatch in summary ids.

Version / Environment information

Oasis 2.3.10

Example data / logs

il_S1_leccalc_wheatsheaf_mean_oep.csv has summary_id set to -1.

summary_id,type,return_period,loss
-1,2,1500.000000,0.000000
-1,2,1000.000000,0.000000
-1,2,900.000000,0.000000
-1,2,800.000000,0.000000
-1,2,700.000000,0.000000
-1,2,600.000000,0.000000
-1,2,500.000000,0.000000
-1,2,400.000000,0.000000
-1,2,300.000000,0.000000
-1,2,250.000000,0.000000
-1,2,200.000000,0.000000
-1,2,150.000000,0.000000
-1,2,100.000000,0.000000
-1,2,75.000000,0.000000
-1,2,50.000000,0.000000
-1,2,25.000000,0.000000
-1,2,20.000000,0.000000
-1,2,10.000000,0.000000
-1,2,5.000000,0.000000
-1,2,2.000000,0.000000

But the real summary_id as defined in il_S1_summary-info.csv is 1.

summary_id,_not_set_,tiv
1,All-Risks,123546789.0

[SkylordA]

Just for information:
Tested this with leccalc, ordleccalc and lecpy with following script in ./OasisLMF. Bug seems to only exist in ktools.

"""Reproduction script for ktools issue #399 - run from OasisLMF root in oasis_dev conda env"""
import shutil
import struct
import subprocess
import tempfile
from pathlib import Path

ASSETS_DIR = Path("tests/assets/test_lecpy/lec_rp")
EPCALC_PERSAMPLEMEAN = "3"  # wheatsheaf mean in ORD format


def write_zero_loss_bin(path, sample_size=2):
    with open(path, "wb") as f:
        f.write(struct.pack("<iii", 50331649, sample_size, 1))  # stream header
        f.write(struct.pack("<iif", 1, 1, 0.0))                 # event_id=1, summary_id=1
        f.write(struct.pack("<if", -1, 0.0))                    # mean idx
        for sidx in range(1, sample_size + 1):
            f.write(struct.pack("<if", sidx, 0.0))              # all-zero losses
        f.write(struct.pack("<if", 0, 0.0))                     # terminator


def print_first_rows(path, label, n=5, epcalc_filter=None):
    lines = Path(path).read_text().splitlines()
    if epcalc_filter:
        lines = [l for l in lines if not l[0].isdigit() or l.split(",")[1] == epcalc_filter]
    print(f"{label} (first {n} data rows):")
    print("\n".join(lines[:n + 1]))  # +1 for header
    print()


with tempfile.TemporaryDirectory() as tmp:
    tmp = Path(tmp)
    shutil.copytree(ASSETS_DIR / "input", tmp / "input")
    (tmp / "work" / "gul").mkdir(parents=True)
    (tmp / "output").mkdir()

    write_zero_loss_bin(tmp / "work" / "gul" / "P1.bin")
    (tmp / "work" / "gul" / "P1.idx").touch()  # empty .idx triggers the bug

    # --- leccalc: legacy output ---
    subprocess.run(
        ["leccalc", "-Kgul", "-r",
         "-M", str(tmp / "output" / "lec_agg_wm.csv"),
         "-m", str(tmp / "output" / "lec_occ_wm.csv")],
        cwd=tmp, check=True, capture_output=True,
    )
    print_first_rows(tmp / "output" / "lec_agg_wm.csv", "leccalc legacy  agg wheatsheaf mean")
    print_first_rows(tmp / "output" / "lec_occ_wm.csv", "leccalc legacy  occ wheatsheaf mean")

    # --- ordleccalc: ORD output ---
    subprocess.run(
        ["ordleccalc", "-Kgul", "-r", "-W", "-M", "-w", "-m",
         "-O", str(tmp / "output" / "ord_ept.csv"),
         "-o", str(tmp / "output" / "ord_psept.csv")],
        cwd=tmp, check=True, capture_output=True,
    )
    print_first_rows(tmp / "output" / "ord_ept.csv", "ordleccalc ORD  agg+occ wheatsheaf mean",
                     epcalc_filter=EPCALC_PERSAMPLEMEAN)

    # --- lecpy: ORD output ---
    write_zero_loss_bin(tmp / "work" / "gul" / "summarypy1.bin")

    from oasislmf.pytools.lec.manager import main
    main(
        run_dir=tmp, subfolder="gul", use_return_period=True,
        agg_wheatsheaf=True, agg_wheatsheaf_mean=True,
        occ_wheatsheaf=True, occ_wheatsheaf_mean=True,
        agg_full_uncertainty=False, agg_sample_mean=False,
        occ_full_uncertainty=False, occ_sample_mean=False,
        ept=tmp / "output" / "lecpy_ept.csv",
        psept=tmp / "output" / "lecpy_psept.csv",
        ext="csv",
    )
    print_first_rows(tmp / "output" / "lecpy_ept.csv", "lecpy          agg+occ wheatsheaf mean",
                     epcalc_filter=EPCALC_PERSAMPLEMEAN)

Output shows no bug in lecpy:

leccalc legacy  agg wheatsheaf mean (first 5 data rows):
summary_id,type,return_period,loss
-1,2,1000.000000,0.000000
-1,2,500.000000,0.000000
-1,2,250.000000,0.000000
-1,2,200.000000,0.000000
-1,2,150.000000,0.000000

leccalc legacy  occ wheatsheaf mean (first 5 data rows):
summary_id,type,return_period,loss
-1,2,1000.000000,0.000000
-1,2,500.000000,0.000000
-1,2,250.000000,0.000000
-1,2,200.000000,0.000000
-1,2,150.000000,0.000000

ordleccalc ORD  agg+occ wheatsheaf mean (first 5 data rows):
SummaryId,EPCalc,EPType,ReturnPeriod,Loss
-1,3,1,1000.000000,0.000000
-1,3,1,500.000000,0.000000
-1,3,1,250.000000,0.000000
-1,3,1,200.000000,0.000000
-1,3,1,150.000000,0.000000

lecpy          agg+occ wheatsheaf mean (first 5 data rows):
SummaryId,EPCalc,EPType,ReturnPeriod,Loss
1,3,1,1000.000000,0.000000
1,3,1,500.000000,0.000000
1,3,1,250.000000,0.000000
1,3,1,200.000000,0.000000
1,3,1,150.000000,0.000000