8000 BUG: `savez()` is not deterministic · Issue #9439 · numpy/numpy · GitHub
[go: up one dir, main page]

Skip to content
BUG: savez() is not deterministic #9439
Open
@meawoppl

Description

@meawoppl

The results of savez() depend on the current time, because the underlying implementation uses tempdir to cache result about to be written. From there the internal instance of ZipFile() auto-detects the time/date and encodes them into the zipfile headers somewhere.

Here is a simple demo:

In [1]: import numpy as np

In [2]: import io

In [3]: bio1 = io.BytesIO()

In<
7C20
/span> [4]: bio2 = io.BytesIO()

In [5]: np.savez_compressed(bio1, foo="bar")

In [6]: import time

In [7]: time.sleep(1)

In [8]: np.savez_compressed(bio2, foo="bar")

In [9]: bio1.getvalue()
Out[9]: b'PK\x03\x04\x14\x00\x00\x00\x08\x00\x84s\xf2J\xd6i\x8cTK\x00\x00\x00\\\x00\x00\x00\x07\x00\x00\x00foo.npy\x9b\xec\x17\xea\x1b\x10\xc9\xc8\xe0\xc6P\xad\x9e\x92Z\x9c\\\xa4n\xa5\xa0n\x13j\xac\xae\xa3\xa0\x9e\x96_TR\x94\x98\x17\x9f_\x94\x92\n\x12wK\xcc)N\x05\x8a\x17g$\x16\xa4\x02\xf9\x1a\x9a:\n\xb5\n(\x80+\x89\x81\x81!\x11\x88\x8b\x80\x18\x00PK\x01\x02\x14\x03\x14\x00\x00\x00\x08\x00\x84s\xf2J\xd6i\x8cTK\x00\x00\x00\\\x00\x00\x00\x07\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x80\x81\x00\x00\x00\x00foo.npyPK\x05\x06\x00\x00\x00\x00\x01\x00\x01\x005\x00\x00\x00p\x00\x00\x00\x00\x00'
                                               ^ HERE
In [10]: bio2.getvalue()
Out[10]: b'PK\x03\x04\x14\x00\x00\x00\x08\x00\x8cs\xf2J\xd6i\x8cTK\x00\x00\x00\\\x00\x00\x00\x07\x00\x00\x00foo.npy\x9b\xec\x17\xea\x1b\x10\xc9\xc8\xe0\xc6P\xad\x9e\x92Z\x9c\\\xa4n\xa5\xa0n\x13j\xac\xae\xa3\xa0\x9e\x96_TR\x94\x98\x17\x9f_\x94\x92\n\x12wK\xcc)N\x05\x8a\x17g$\x16\xa4\x02\xf9\x1a\x9a:\n\xb5\n(\x80+\x89\x81\x81!\x11\x88\x8b\x80\x18\x00PK\x01\x02\x14\x03\x14\x00\x00\x00\x08\x00\x8cs\xf2J\xd6i\x8cTK\x00\x00\x00\\\x00\x00\x00\x07\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x80\x81\x00\x00\x00\x00foo.npyPK\x05\x06\x00\x00\x00\x00\x01\x00\x01\x005\x00\x00\x00p\x00\x00\x00\x00\x00'
                                               ^ HERE
In [11]: bio1.getvalue() == bio2.getvalue()
Out[11]: False

A minimum of 1 second difference is necessary for byte-wise similarity to fail.

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0