KeyError when generating scatter plot of DataFrame columns #4493
Comments
|
@fonnesbeck sorry to hear about all your plotting troubles! i'll check this one out |
ghost
assigned
|
if i do I get the following plot: Here are my versions |
|
Try going back to my example, and doing the following: |
|
whoa i get a totally different error than the one you showed above... you can't plot converters = {'active': lambda x: bool(int(x))}
df = read_csv('processed_data.csv', index_col=0, parse_dates=[-2, -1], converters=converters)
df = df.convert_objects(convert_numeric=True)
scatter(df.pdgt10, df.loa)You should get: |
|
I wonder why |
|
in to see it do |
|
Also some that can't just evaluate to 0 |
|
Sorry, that was a bad choice of column. The one in my real model is the result of a merge on 2 tables like this and use modified pdgt10 columns, without these unconvertable values. I will get a better example up when I get a chance. |
|
Try the and see if you do not get the KeyError. |
|
i don't get the error |
|
The issue is probably related to this then: |
|
i'm guessing that's dev let me try it out |
|
Yes, current master. |
|
hm nope works with that too... |
|
Okay, back to the drawing board for me then. |
|
can u try with |
|
Will try. I wonder if your recent fix for my other issue was related? Getting on a plane; will update later. Thanks. |
|
don't think so since these series have unique monotonic indices |
|
Can you try importing the data with |
|
able to plot on 1.3.1 trying 1.4.x |
|
NICE, GOT IT with 1.4.x |
|
the issue is that matplotlib calls a little unsatisfying, but i'm not sure what else you can do... |
|
i'll look at a diff of @fonnesbeck thanks for the report! |
|
BTW, I get similar behavior with |
|
recent change to that that |
|
@fonnesbeck can u try with latest matplotlib master? it's working on my end |
|
here'st he diff for the offending chunk from latest matplotlib diff --git a/lib/matplotlib/units.py b/lib/matplotlib/units.py
index 8f08a64..73d363a 100644
--- a/lib/matplotlib/units.py
+++ b/lib/matplotlib/units.py
@@ -134,8 +134,19 @@ class Registry(dict):
converter = self.get(classx)
if isinstance(x, np.ndarray) and x.size:
- converter = self.get_converter(x.ravel()[0])
- return converter
+ xravel = x.ravel()
+ try:
+ # pass the first value of x that is not masked back to
+ # get_converter
+ if not np.all(xravel.mask):
+ # some elements are not masked
+ converter = self.get_converter(
+ xravel[np.argmin(xravel.mask)])
+ return converter
+ except AttributeError:
+ # not a masked_array
+ converter = self.get_converter(xravel[0])
+ return converter
if converter is None and iterable(x):
for thisx in x: |
|
success! |
|
excellent! |
|
hm i'm not sure this doing the right thing tho.... in pandas pre series->ndframe refactor in post refactor seriess world the |
|
OTOH maybe good enough is good enough |
|
pushing to 0.14 |
|
closing as stale. @fonnesbeck ping us if this comes up again |
|
Never mind, figured it out =). Was similar to issue #6127. Just passing @cpcloud, I'm getting what I think is the same error but when trying to create a histogram with matplotlib. Edit: After looking at the above errors a bit more, its not quite the same. Let me know if I should create a new issue for this. I have Pandas 0.14, but still running matplotlib 1.3.1. This function was working previously on the same data (not sure what versions, it was a few months ago) but isn't now. I was on Pandas 0.13.x earlier and it wasn't working, tried upgrading and it still isn't working =(. Any ideas? The histogram works on with some sub-sets of the data not others (example below). I haven't quite been able to find the differences in what is working or not. Update: This issue has something to do with series which have indices that don't start at 0. I can get it working if create a new index from 0 -> Length with the same series that wasn't working before. One of the sub-sets it is not working on: And here are the errors: |
This commit prevents the KeyError raised when DataFrame.plot() is called with xerr or yerr being a Series or DataFrame whose index doesn't include 0. The error comes from matplotlib code which tries to access xerr[0] or yerr[0], so to solve the problem, we convert xerr and yerr from Pandas objects to NumPy ndarrays before sending them through to matplotlib. This is a different instance of the same type of problem in Github issues pandas-dev#4493 and pandas-dev#6127 (and perhaps others).
I have two columns of a
DataFramethat I am trying to plot usingscatterin Matplotlib. Here are the two series:(Actually the latter is the result of a comparison operation on the column to see if values are greater or less than some threshold value)
Unfortunately, calling
scatteron these two axes raises aKeyError: 0exception for some reason.This worked like a charm prior to updating pandas and matplotlib a week or so ago.
Running Python 2.7.2 on OS X 10.8.4.
The text was updated successfully, but these errors were encountered: