Skip to content

Conversation

@chris-b1
Copy link
Contributor

@chris-b1 chris-b1 commented Feb 1, 2016

closes part of #7621

WIP, but wanted to put it up as is, since it probably interacts with #12129
cc: @kawochen

this is how I modified the schema - open to suggestions

# normal block
 {'dtype': dtype,
  'data': ExtType(0, <bytes>)}
# Categorical block
 {'dtype': 'categorical',
  'data': {
      'codes': {'dtype': 'int8', 'data': ExtType(0, <bytes>),
      'categories': {'dtype': dtype, 'data': ExtType(0, <bytes>)}
}
@pwaller
Copy link
Contributor

pwaller commented Mar 9, 2016

I have cherry picked this onto 0.18.0rc1, fixed the conflict and it worksforme™.

Can I submit a PR with it?

@jreback
Copy link
Contributor

jreback commented Mar 9, 2016

iirc last time this was looked at we were going for a slightly more general soln for extension types
I believe @kawochen made the changes in master to make this accessible

@chris-b1
@kawochen

@pwaller
Copy link
Contributor

pwaller commented Mar 9, 2016

It would be awesome if there was any chance of getting this into 0.18.0.

@jreback
Copy link
Contributor

jreback commented Mar 9, 2016

no but 0.18.1 won't be far away

@chris-b1
Copy link
Contributor Author

chris-b1 commented Mar 9, 2016

@pwaller - feel free to submit a PR on my branch or just open up a new one if it's simpler.

Looking at #12129, the Series constructor logic can be probably cleaned up, but I think the overall approach still needs to something like what I did here - extending the schema to store both categories and codes.

@pwaller
Copy link
Contributor

pwaller commented Mar 9, 2016

Submitted at #12573

@jreback
Copy link
Contributor

jreback commented Mar 9, 2016

closing in favor of #12573

@jreback jreback closed this Mar 9, 2016
@chris-b1 chris-b1 deleted the msgpack-categorical branch June 8, 2016 22:33
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Categorical Categorical Data Type Enhancement

3 participants