evaluate casts data types through floats messing with long int

[vaexio]

evaluate casts data types through floats messing with long int

[maartenbreddels]

That should not happen, which version are you using?

[mfouesneau]

version gives me 1.0.0-beta.4

[maartenbreddels]

Doing the following:

import vaex as vx
import numpy as np
ds = vx.from_arrays(f8=np.arange(10, dtype='f8'), i8=np.arange(10, dtype='i8'))
ds.evaluate('f8').dtype, ds.evaluate('i8').dtype, 

Gives me:

(dtype('float64'), dtype('int64'))

Can you check if you get the same?
If you don't get that, can you update to beta 6 and let me know if that fixes it?

[mfouesneau]

In [1]: import vaex as vx
...: import numpy as np
...: ds = vx.from_arrays(f8=np.arange(10, dtype='f8'), i8=np.arange(10, dtype='i8'))
...: ds.evaluate('f8').dtype, ds.evaluate('i8').dtype,
...:
Out[1]: (dtype('float64'), dtype('float64'))

[mfouesneau]

How do I install the new beta? conda-forge does not have it and pip seems to want 0.3.10 something

correction: pip install --pre vaex

[mfouesneau]

The API changed between versions.
from_arrays does not take a dataset name argument anymore.

Do you have the major breaking changes in the API somewhere?
(actually major changes like adding scatter etc, as the documentation does not tell when it is created)

[maartenbreddels]

Conda-forge has beta6 (see https://anaconda.org/conda-forge/vaex), and installing from pip you indeed should use the --pre flag.
And indeed, that was a breaking api change, that was a bad decision to take a name argument.
I plan to keep a public changelog after the 1.0 release, changes are prefixed in the commit msg'es by 'changes:' for now, new features with 'new:' and fixes with 'fix:'.

Please let me know if you've managed to get this working.
(Sorry for the late reaction, I was traveling)

[mfouesneau]

yes I looked at the source to figure the changes in some places. It works!

FYI, the healpix_plot doc needs updates or the code needs to be generalized. It currently imposes Gaia conventions that are not healpy defaults. (nested not ring for instance). I can open an issue if you need.

[maartenbreddels]

Please do, we can discuss further.