Shared memory and numpy arrays

[PaulRudin]

Greetings once again!

>>> a = c.get("o000066a0a7c6fb1c")
>>> a
ndarray([      14,      226,      276, ...,   131315,  3220012, 11171032],
        dtype=int32)
>>> b = c.get("o000066a0a7c6fb1c")
>>> a is b
False
>>> a.data is b.data
False
>>> a.data
<memory at 0x7f943d167340>
>>> b.data
<memory at 0x7f943d167700>

I think I was expecting the underlying data to be the same bit of memory. I guess that's not so?

[PaulRudin]

Ah - maybe it is?

>>> a.base.data
<memory at 0x7f943d167340>
>>> b.base.data
<memory at 0x7f943d167640>

[PaulRudin]

Ah, those are not the same... just close.

[sighingnow]

Ah, those are not the same... just close.

a.base.data and b.base.data is not the memory address.

[sighingnow]

Try run a.base.data multiple times, you'll see different output for each run.

[sighingnow]

@PaulRudin numpy.ndarray.base.data is not the memory address of the array payload. With vineyard, get the same object twice with the same client instance, the memory address should be the same.

In [3]: import vineyard
import
In [4]: import numpy as np

In [5]: client = vineyard.connect("/tmp/vineyard.sock")

In [6]: object_id = client.put(np.ones((10000, 1000)))

In [7]: r1 = client.get(object_id)

In [8]: r2 = client.get(object_id)

In [9]: r1 is r2
Out[9]: False

In [10]: r1.__array_interface__["data"]
Out[10]: (140492393889856, True)

In [11]: r2.__array_interface__["data"]
Out[11]: (140492393889856, True)

In [12]: r1.__array_interface__["data"] == r2.__array_interface__["data"]
Out[12]: True

See https://stackoverflow.com/questions/11264838/how-to-get-the-memory-address-of-a-numpy-array-for-c for more discussion about the memory address of numpy ndarray.