BUG: inferred COO reshape dimensions lose integer precision
Maintainers usually reply within 2 days
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 25/100
Research direction
Start from the COO reshape path that infers a dimension from -1, then trace how it calculates the inferred size. Reproduce the large sparse shape case and the multiple-unknown-dimensions case described in the issue, and compare their results with NumPy. Done means both cases match the expected behavior and the four regression cases pass; an open linked pull request already addresses this work.
Written by the indexing model from the issue text.
Description
Bug
COO inferred reshapes pass the dimension calculation through floating-point division. This loses integer precision for large but extremely sparse shapes.
Reproduced on main at d63255057744331b96626a1f1c24504b3a3783cd.
import numpy as np
import sparse
rows = 2**53 + 1
size = 3 * rows
x = sparse.COO(np.array([[size - 1]], dtype=np.int64), [7], shape=(size,))
x.reshape((-1, 3))
# ValueError, although size is exactly divisible by 3.
The expected shape is (9007199254740993, 3) and the single stored coordinate should be (9007199254740992, 2). This reproducer stores only one value; it does not allocate a huge dense array.
The same inferred-dimension path also silently accepts multiple -1 entries when the array has size one, unlike NumPy.
A focused fix using exact integer division and rejecting multiple unknown dimensions is ready with regression tests. Four new cases fail before the production change.
This is an AI-assisted bug investigation with locally executed reproductions and tests. No assignment is requested.
- Dominant language
- Python
- Stars
- 670
- Forks
- 167
- Avg merge
- 3d 4h
- Merged PRs (30d)
- 7
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from pydata/sparse
-
Bug: full einsum contractions upcast the dtype and count booleansPossibly taken @RaphaelFakhri claimed this 17 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 2 days
-
BUG: ufunc outer rejects Python scalar and sequence operandsPossibly taken @nawaaaaaAaar claimed this 29 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 2 days
-
Bug: slices with a negative step and a negative stop return wrong elementsPossibly taken @JoeyTan21 claimed this 19 days ago. Open
Difficulty 3/5 1-2 days Newbie friendliness 75/100
pydata/sparse#1019 · 1 comment ·
Maintainers usually reply within 2 days
-
BUG: matmul fails with zero-length broadcast batch dimensionsPossibly taken @nawaaaaaAaar claimed this 29 days ago. Open
Difficulty 3/5 1-2 days Newbie friendliness 40/100
Maintainers usually reply within 2 days
-
BUG: zero-length tensor contractions ignore requested output typesPossibly taken @nawaaaaaAaar claimed this 29 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 35/100
Maintainers usually reply within 2 days
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
abetlen/llama-cpp-python#2424 ·
Maintainers usually reply within 1 day
-
Bug IronPython Python 3
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
pyrevitlabs/pyRevit#4184 ·
Maintainers usually reply within 1 day
-
Difficulty 1/5 Under an hour Newbie friendliness 74/100
jqueryscript/awesome-claude-code#1009 ·
-
tests
Difficulty 2/5 1-3 hours Newbie friendliness 66/100
Maintainers usually reply within 3 days
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Maintainers usually reply within 1 day