Change ._data to ._parent for accessors #21906
Conversation
… BlockManager lookups
Codecov Report
@@ Coverage Diff @@
## master #21906 +/- ##
=======================================
Coverage 92.06% 92.06%
=======================================
Files 169 169
Lines 50698 50698
=======================================
Hits 46675 46675
Misses 4023 4023
Continue to review full report at Codecov.
|
|
I can see changing this as |
Not really. At the sprint I brought up the idea of changing BlockManager references from _data to something more distinctive like _mgr, but the reception was luke-warm at best. There is also #19658 but AFAIK there has been no action on that. |
| @@ -27,14 +27,14 @@ def __init__(self, data, orig): | |||
| raise TypeError("cannot convert an object of type {0} to a " | |||
| "datetimelike index".format(type(data))) | |||
|
|
|||
| self.values = data | |||
There was a problem hiding this comment.
am puzzled why this doesn't actually break anything; i think that the dateitmelike accessors do something different?
There was a problem hiding this comment.
i think that the dateitmelike accessors do something different?
CategoricalAccessor uses self.categories, is the outlier this PR leaves unchanged.
|
The CI is having unrelated errors here much more consistently than usual. |
|
@mroeschke if you're comfortable with the "Take easy stuff off jreback's plate" strategy, this may qualify. |
|
i don’t think we should do this though let me look a bit later |
Huh, here you seemed on board. My mistake.
The goal was to take "easy" stuff off your plate. This is definitely not a priority. |
|
@jbrockmendel hah, right I do remember that now. let me look again. my worry is that we are changing some things to |
|
I think using Whether we we want to go down the road of changing |
I think that's @jbrockmendel's reason for changing this in the first place, since |
|
yeah I think this change is ok. Just want to document this somewhere and quickly survey other uses for comonaility. |
|
Sounds good. http://pandas-docs.github.io/pandas-docs-travis/internals.html#
may be a decent place for documenting these kinds of relationships between
classes and their data attributes.
…On Tue, Aug 7, 2018 at 8:37 AM Jeff Reback ***@***.***> wrote:
yeah I think this change is ok. Just want to document this somewhere and
quickly survey other uses for comonaility.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#21906 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABQHIkvGI6sixHNSxFiTprLqi6DQjCATks5uOZgmgaJpZM4VPl_n>
.
|
OK, if we're going down this road, should I update |
|
yes i am ok with all non internals things changing to use _parent |
|
ok, can you add some docs in internals.rst about this |
The idea is to reduce the number of distinct meanings
._datahas. With this it is down to justIndex._dataandNDFrame._data, I think.