I posted a comment on https://github.com/ipython/ipython/issues/2379, but I realized this may be specific to Pandas. I'm using IPython 6.2.0, Python 3.5, Jupyter 4.3.0, and Pandas 0.20.1.
Pandas won't use __str__
or __repr__
(or anything I can find) to display an iterable object that defines __len__
Code Sample, a copy-pastable example if possible
class WithLen:
z = [1,2,3]
def __getitem__(self, i): return self.z[i]
def __str__(self): return "from str"
def __repr__(self): return "from repr"
def __len__(self): return len(self.z)
df = pd.DataFrame([
pd.Series({'test': WithLen()})
])
print(df) # PROBLEMATIC CASE
print(df['test'].iloc[0]) # prints fine
class WithoutLen:
z = [1,2,3]
def __getitem__(self, i): return self.z[i]
def __str__(self): return "from str"
def __repr__(self): return "from repr"
def __len__(self): return len(self.z)
print(pd.DataFrame([
pd.Series({'test': WithoutLen()})
])) # prints fine
Problem description
I've implemented __str__
and __repr__
on an iterable class with good reason, and it being iterable shouldn't override that behavior.
Expected Output
I expected and want to see this printed both with and without __len__
:
test
0 from str
Output of pd.show_versions()
INSTALLED VERSIONS
------------------
commit: None
python: 3.5.3.final.0
python-bits: 64
OS: Linux
OS-release: 4.11.6-201.fc25.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.utf8
LOCALE: en_US.UTF-8
pandas: 0.20.3
pytest: 3.0.7
pip: 9.0.1
setuptools: 36.5.0
Cython: 0.25.2
numpy: 1.13.1
scipy: 0.19.0
xarray: None
IPython: 6.2.0
sphinx: 1.5.1
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2017.2
blosc: None
bottleneck: 1.2.0
tables: 3.3.0
numexpr: 2.6.2
feather: None
matplotlib: 2.0.2
openpyxl: 2.4.1
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: 0.9.6
lxml: 3.7.3
bs4: 4.5.3
html5lib: 0.999
sqlalchemy: 1.1.6
pymysql: 0.7.11.None
psycopg2: None
jinja2: 2.9.5
s3fs: None
pandas_gbq: None
pandas_datareader: None
Comment From: TomAugspurger
You might try looking in pandas/io/formats/format.py
, specifically https://github.com/pandas-dev/pandas/blob/f9d88cd6b3543bbb678378fc4fe736f13497d21e/pandas/io/formats/format.py#L1765 and I'm assuming https://github.com/pandas-dev/pandas/blob/f9d88cd6b3543bbb678378fc4fe736f13497d21e/pandas/io/formats/format.py#L1803
Presumably we cast to np.array
somewhere in there, perhaps unnecessarily.
Comment From: jreback
to be honest this is not real well supported, nor intended. you are pretty much on your own.
putting anything else besides scalars inside a Series/DataFrame is non-idiomatic. There will be some support for containers or things (e.g. lists / arrays) in pandas2.
Comment From: TomAugspurger
I think if we can support formatting without too much effort then we should. @dmyersturnbull if you can track down the issue and suggest a fix, it'd probably be accepted.
Comment From: jorisvandenbossche
The reason is because in our 'pretty print' implementation, we have a special method for sequences that can follow the max_seq_items
option:
https://github.com/pandas-dev/pandas/blob/f9d88cd6b3543bbb678378fc4fe736f13497d21e/pandas/io/formats/printing.py#L97-L126
Which more or less hardcodes the possible results (depending on whether setitem exists, it will look like list of tuple). I am not directly sure of a way to both follow the repr of custom objects and still be able to limit the number of items showed for normal reprs.