Indexing a vlen variable produces inconsistent results

[5tefan]

Summary

Indexing a vlen variable produces inconsistent results:

import netCDF4 as nc
import numpy as np

nc_out = nc.Dataset("ex.nc", "w", diskless=True, persist=False)
t = nc_out.createDimension("t", None)
v = nc_out.createVariable("s", str, ("t",))

v[0] = "aaa"
v[1] = "bbb"
v[2] = "ccc"
v[3] = "ddd"

print("1. ", v[:])
print("2. ", v[np.arange(4)])
print("3. ", v[np.array([0, 1, 2])])
print("4. ", v[np.array([0, 3, 2])])

Produces the following output:

1.  ['aaa' 'bbb' 'ccc' 'ddd']
2.  ['aaa' 'bbb' 'ccc' 'ddd']
3.  ['aaa' 'bbb' 'ccc']
4.  [array('aaa', dtype=object) array('ddd', dtype=object)
 array('ccc', dtype=object)]

The result of number 4 is surprising and seems to be a bug. Items 1, 2, and 3 show expected behavior. The issue occurs when indexing out of order.

Versions

1. MacOS Ventura 13.6.3 Intel based
2. Reproduced with Python v3.10.13 and 3.8.10.
3. netcdf4 Python versions 1.6.5 and 1.5.7

[jswhit]

closed by PR #1309