TST: test_ts_plot_format_coord ValueError: Unknown format code 'f' for object of type 'str'

[yarikoptic]

Only with ubuntu 12.04 i386:

======================================================================
ERROR: test_ts_plot_format_coord (pandas.tseries.tests.test_plotting.TestTSPlot)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/tmp/buildd/pandas-0.14.0+git345-g8cd3dd6/debian/tmp/usr/lib/python2.7/dist-packages/pandas/tseries/tests/test_plotting.py", line 138, in test_ts_plot_format_coord
    check_format_of_first_point(annual.plot(), 't = 2014  y = 1.000000')
  File "/tmp/buildd/pandas-0.14.0+git345-g8cd3dd6/debian/tmp/usr/lib/python2.7/dist-packages/pandas/tseries/tests/test_plotting.py", line 135, in check_format_of_first_point
    self.assertEqual(expected_string, ax.format_coord(first_x, first_y))
  File "/tmp/buildd/pandas-0.14.0+git345-g8cd3dd6/debian/tmp/usr/lib/python2.7/dist-packages/pandas/tseries/plotting.py", line 90, in <lambda>
    y))
ValueError: Unknown format code 'f' for object of type 'str'

INSTALLED VERSIONS
------------------
commit: None
python: 2.7.3.final.0
python-bits: 32
OS: Linux
OS-release: 3.2.0-4-amd64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: C
LANG: C

pandas: 0.14.0.dev
nose: 1.1.2
Cython: 0.15.1
numpy: 1.6.1
scipy: 0.9.0
statsmodels: 0.5.0
IPython: None
sphinx: 1.1.3
patsy: 0.2.1
scikits.timeseries: None
dateutil: 1.5
pytz: 2012c
bottleneck: None
tables: 2.3.1
numexpr: 1.4.2
matplotlib: 1.1.1rc
openpyxl: 1.7.0
xlrd: 0.6.1
xlwt: 0.7.2
xlsxwriter: None
lxml: None
bs4: 4.0.2
html5lib: 0.90
httplib2: None
apiclient: None
rpy2: None
sqlalchemy: None
pymysql: None
psycopg2: None

[jreback]

cc @sinhrks and thoughts in this?

[sinhrks]

Not sure, but maybe dtype is changed in either process? Whichever, it may better to add error handling process in format_coord.

• Series construction using single value (may result in object dtype?)
• Or matplotlib internal process?

@yarikoptic Can you attach the result of the following code to confirm when dtype is changed?

import pandas as pd

print('instanciate with value')
annual = pd.Series(1, index=pd.date_range('2014-01-01', periods=3, freq='A-DEC'))
print('dtype', annual.dtypes, annual.values)
ax = annual.plot()
first_line = ax.get_lines()[0]
first_y = first_line.get_ydata()[0]
print('first_y', type(first_y), first_y)
first_y = first_line.get_ydata(orig=True)[0]
print('first_y_orig', type(first_y), first_y)
first_y = first_line.get_ydata(orig=False)[0]
print('first_y_conv', type(first_y), first_y)

print('instanciate with list')
annual = pd.Series([1, 1, 1], index=pd.date_range('2014-01-01', periods=3, freq='A-DEC'))
print('dtype', annual.dtypes, annual.values)
ax = annual.plot()
first_line = ax.get_lines()[0]
first_y = first_line.get_ydata()[0]
print('first_y', type(first_y), first_y)
first_y = first_line.get_ydata(orig=True)[0]
print('first_y_orig', type(first_y), first_y)
first_y = first_line.get_ydata(orig=False)[0]
print('first_y_conv', type(first_y), first_y)

[yarikoptic]

here you go sir

# PYTHONPATH=/tmp/buildd/pandas-0.14.0+git345-g8cd3dd6/debian/tmp/usr/lib/python2.7/dist-packages/ xvfb-run -a -s "-screen 0 1280x1024x24 -noreset" python /tmp/testcode.py 
instanciate with value
('dtype', dtype('int64'), array([1, 1, 1], dtype=int64))
('first_y', <type 'numpy.int64'>, 1)
('first_y_orig', <type 'numpy.int64'>, 1)
instanciate with list
('dtype', dtype('int64'), array([1, 1, 1], dtype=int64))
('first_y', <type 'numpy.int64'>, 1)
('first_y_orig', <type 'numpy.int64'>, 1)

[yarikoptic]

most probably it was not a type change but rather some incorrect type checking inside somewhere, and probably elderly numpy to blame (since doesn't appear with newer releases I guess was fixed)?

(Pdb) print "t = {0}  y = {1:8f}".format(Period(ordinal=int(t), freq=ax.freq),y)
*** ValueError: Unknown format code 'f' for object of type 'str'
(Pdb) print y
1
(Pdb) print type(y)
<type 'numpy.int64'>
(Pdb) print np.__version__
1.6.1

[jreback]

@yarikoptic this works for me on 1.6.1, but using matplotlib 1.1.1 (I don't have 1.1.1rc). Can you update and try with that?

[jreback]

I can just skip this on old numpy too?

[yarikoptic]

and 32bit? I guess you could indeed just skip it or (to just help yourself if users on those start chasing you) wrap into try/except and then may be spit out warning and use '%s' for that one in those cases or smth like that?

(Pdb) print "t = {0}  y = {1:8s}".format(Period(ordinal=int(t), freq=ax.freq),y)
t = 2014  y = 1