FAILURE MAP
← Case archive

FA-100471 / Sparse tensor canonicalization / Open access

Sparse tensor: duplicate CSR columns are retained as separate coordinates · case 01

duplicate CSR columns are retained as separate coordinates.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

CSR rows can contain duplicate column coordinates whose values contribute to one logical cell.

THE FAILURE

CSR rows can contain duplicate column coordinates whose values contribute to one logical cell.

Unsuccessful approach: Overwrite order still discards all but one duplicate contribution.

Case contract

Within each row, sum equal columns, omit zero totals, and return canonical CSR [row_ptr,columns,values].

Why this case matters

A small offline sparse-array operation has explicit coordinate and storage semantics.

1 / The failure

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(ptr,columns,values):
    return [ptr,columns,values]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('duplicates',solve([0,3],[1,1,0],[N,2,4]),[[0,2],[0,1],[4,N+2]])
check('rows isolated',solve([0,2,4],[0,0,0,0],[1,2,3,4]),[[0,1,2],[0,0],[3,7]])
check('empty',solve([0],[],[]),[[0],[],[]])
check('single',solve([0,1],[N],[6]),[[0,1],[N],[6]])
check('cancellation',solve([0,2],[0,0],[N,-N]),[[0,0],[],[]])
check('unsorted unique',solve([0,2],[2,0],[5,3]),[[0,2],[0,2],[3,5]])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
duplicates[[0, 3], [1, 1, 0], [1, 2, 4]][[0, 2], [0, 1], [4, 3]]Failed
rows isolated[[0, 2, 4], [0, 0, 0, 0], [1, 2, 3, 4]][[0, 1, 2], [0, 0], [3, 7]]Failed
empty[[0], [], []][[0], [], []]Passed
single[[0, 1], [1], [6]][[0, 1], [1], [6]]Passed
cancellation[[0, 2], [0, 0], [1, -1]][[0, 0], [], []]Failed
unsorted unique[[0, 2], [2, 0], [5, 3]][[0, 2], [0, 2], [3, 5]]Failed

SHA-256 / 0bfa57a57faf4c09ead5b54ba57a437fa0b1a10d9e0d00295380400d1f36c63a

2 / The unsuccessful fix

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(ptr,columns,values):
    outp=[0]; outc=[]; outv=[]
    for r in range(len(ptr)-1):
     pairs=sorted(zip(columns[ptr[r]:ptr[r+1]],values[ptr[r]:ptr[r+1]]))
     for c,v in pairs:
      if not outc or outp[-1]==len(outc) or outc[-1]!=c: outc.append(c); outv.append(v)
      else: outv[-1]=v
     outp.append(len(outc))
    return [outp,outc,outv]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('duplicates',solve([0,3],[1,1,0],[N,2,4]),[[0,2],[0,1],[4,N+2]])
check('rows isolated',solve([0,2,4],[0,0,0,0],[1,2,3,4]),[[0,1,2],[0,0],[3,7]])
check('empty',solve([0],[],[]),[[0],[],[]])
check('single',solve([0,1],[N],[6]),[[0,1],[N],[6]])
check('cancellation',solve([0,2],[0,0],[N,-N]),[[0,0],[],[]])
check('unsorted unique',solve([0,2],[2,0],[5,3]),[[0,2],[0,2],[3,5]])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
duplicates[[0, 2], [0, 1], [4, 2]][[0, 2], [0, 1], [4, 3]]Failed
rows isolated[[0, 1, 2], [0, 0], [2, 4]][[0, 1, 2], [0, 0], [3, 7]]Failed
empty[[0], [], []][[0], [], []]Passed
single[[0, 1], [1], [6]][[0, 1], [1], [6]]Passed
cancellation[[0, 1], [0], [1]][[0, 0], [], []]Failed
unsorted unique[[0, 2], [0, 2], [3, 5]][[0, 2], [0, 2], [3, 5]]Passed

SHA-256 / f538104b4501b921bac99fffa2bb7151b38d57c8e22f9cf4af8e3834b7c86d67

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Sign in to the archive ↗

Verification & scope

These are toy integer-valued COO or CSR contracts; they make no claim about a framework API, floating point policy, or large tensor performance. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-30T16:48:35.815395+00:00.

Case digest / b2945245e97809e6c83a209b6407485edbbd0331c1c04b9fdce23c053e31a53e