genutil.statistics.mean does not exists

[chaosphere2112]

because MV2.average does pretty much the same thing

but this would allow to use nifty axis='t' or axis='xy' or axis='(my_axis_name)'

Migrated from: CDAT/cdat#1569

[github-actions]

Marking issue as stale, since there has been no activity in 30 days.

Unless the issue is updated or the 'stale' tag is removed, this issue will be closed in 7 days.

[durack1]

@jasonb5 this is related to the fact that cdms/cdutil etc all had their origin in numeric, before numpy and as numeric didn't include all the functionality that numpy now has, basic stats function were created and are now duplicated by the numpy variants, assuming numpy is wrapped correctly to expose all the functions. The MV2 library could mostly be deprecated in favour of numpy pure functions, or alternatively MV2 just calls the numpy functions directly (preserving the cdms transient variable attributes of course)

[jypeter]

I totally agree with accessing numpy (more likely numpy.ma) with MV2, in order to preserve attributes would be very useful!

It could be useful to add a new compute_history attribute to the variable specifying which function was applied, and append this if there is a pre-existing compute_history attribute (same way the history global attribute works in nc files)

Because in some cases, the original name/units of the variable may not mean anything after applying the numpy function

[durack1]

@jypeter another good idea, so similar to how nco operates with appending to the history global attribute with a summary of what was done.

I am sure @pochedls will also have some ideas around this

[jasonb5]

The MV2 module already wraps numpy.ma functions. I see there is an average but not mean in MV2, I can add this.

As for history recording, we're planning on adding provenance support in the next-gen CDMS2.

[durack1]

@jasonb5 thanks for the update, when we talked about this a long time ago, I was suggesting that removing the duplication in MV2 by rather wrapping all the numpy.ma functions was a good idea, which means that if numpy.ma adds another function (or bugfix) we get this in CDMS for free with no maintenance overhead for you folks

[jasonb5]

Ah I see what you mean, this is inline with the next-gen CDMS where MV2 will be absorbed into the Variable class. As this would be a fairly large change for the current implementation, we'll handle it in the next.

[durack1]

Perfect, that sounds like we're on the same page here

[github-actions]

Marking issue as stale, since there has been no activity in 30 days.

Unless the issue is updated or the 'stale' tag is removed, this issue will be closed in 7 days.