We Keep Coding

Why pandas prefers float64 over Int64?

I have a pandas dataframe with a column, with 3 unique values: [0, None, 1]
When I run this line:
test_data = test_data.apply(pd.to_numeric, errors='ignore')
the above mentioned column data type is converted to float64
Why not int64? Technically integer type can handle None values, so I'm confused why it didn't pick int64?
Thanks for help,
Edit:
As I read about the difference between int64 and Int64, why pandas doesn't choose Int64 then?

The question is misleading, because you actually want to know why pandas does this, not python. As for pandas, it tries to find the smallest numerical dtype, as stated in the documentation (downcast): https://pandas.pydata.org/docs/reference/api/pandas.to_numeric.html
downcaststr, default None Can be ‘integer’, ‘signed’, ‘unsigned’, or
‘float’. If not None, and if the data has been successfully cast to a
numerical dtype (or if the data was numeric to begin with), downcast
that resulting data to the smallest numerical dtype possible according
to the following rules:
‘integer’ or ‘signed’: smallest signed int dtype (min.: np.int8)
‘unsigned’: smallest unsigned int dtype (min.: np.uint8)
‘float’: smallest float dtype (min.: np.float32)
As this behaviour is separate from the core conversion to numeric
values, any errors raised during the downcasting will be surfaced
regardless of the value of the ‘errors’ input.
In addition, downcasting will only occur if the size of the resulting
data’s dtype is strictly larger than the dtype it is to be cast to, so
if none of the dtypes checked satisfy that specification, no
downcasting will be performed on the data.
with code:
import pandas as pd
df = pd.DataFrame(
{
'int':[1,2],
'float':[1.0,2.0],
'none': [1, None]
}
)
you might notice that:
print(df.loc[1, 'none'])
# nan
returns nan, not None this is because pandas uses the numpy library.
Numpy offers a value which is treated like a number but isn't actually one and is called: not a number (nan), this value is of type float.
import numpy as np
print(type(np.nan))
# <class 'float'>
Since you used None, pandas tries to find the correct data type, finds numbers with missing values, since it can't handle that it casts it to float where it is able to inser np.nan for the missing values.

Pandas ValueError: cannot convert float NaN to integer [duplicate]

I get ValueError: cannot convert float NaN to integer for following:
df = pandas.read_csv('zoom11.csv')
df[['x']] = df[['x']].astype(int)
The "x" is a column in the csv file, I cannot spot any float NaN in the file, and I don't understand the error or why I am getting it.
When I read the column as String, then it has values like -1,0,1,...2000, all look very nice int numbers to me.
When I read the column as float, then this can be loaded. Then it shows values as -1.0,0.0 etc, still there are no any NaN-s
I tried with error_bad_lines = False and dtype parameter in read_csv to no avail. It just cancels loading with same exception.
The file is not small (10+ M rows), so cannot inspect it manually, when I extract a small header part, then there is no error, but it happens with full file. So it is something in the file, but cannot detect what.
Logically the csv should not have missing values, but even if there is some garbage then I would be ok to skip the rows. Or at least identify them, but I do not see way to scan through file and report conversion errors.
Update: Using the hints in comments/answers I got my data clean with this:
# x contained NaN
df = df[~df['x'].isnull()]
# Y contained some other garbage, so null check was not enough
df = df[df['y'].str.isnumeric()]
# final conversion now worked
df[['x']] = df[['x']].astype(int)
df[['y']] = df[['y']].astype(int)

For identifying NaN values use boolean indexing:
print(df[df['x'].isnull()])
Then for removing all non-numeric values use to_numeric with parameter errors='coerce' - to replace non-numeric values to NaNs:
df['x'] = pd.to_numeric(df['x'], errors='coerce')
And for remove all rows with NaNs in column x use dropna:
df = df.dropna(subset=['x'])
Last convert values to ints:
df['x'] = df['x'].astype(int)

ValueError: cannot convert float NaN to integer
From v0.24, you actually can. Pandas introduces Nullable Integer Data Types which allows integers to coexist with NaNs.
Given a series of whole float numbers with missing data,
s = pd.Series([1.0, 2.0, np.nan, 4.0])
s
0 1.0
1 2.0
2 NaN
3 4.0
dtype: float64
s.dtype
# dtype('float64')
You can convert it to a nullable int type (choose from one of Int16, Int32, or Int64) with,
s2 = s.astype('Int32') # note the 'I' is uppercase
s2
0 1
1 2
2 NaN
3 4
dtype: Int32
s2.dtype
# Int32Dtype()
Your column needs to have whole numbers for the cast to happen. Anything else will raise a TypeError:
s = pd.Series([1.1, 2.0, np.nan, 4.0])
s.astype('Int32')
# TypeError: cannot safely cast non-equivalent float64 to int32

Also, even at the lastest versions of pandas if the column is object type you would have to convert into float first, something like:
df['column_name'].astype(np.float).astype("Int32")
NB: You have to go through numpy float first and then to nullable Int32, for some reason.
The size of the int if it's 32 or 64 depends on your variable, be aware you may loose some precision if your numbers are to big for the format.

I know this has been answered but wanted to provide alternate solution for anyone in the future:
You can use .loc to subset the dataframe by only values that are notnull(), and then subset out the 'x' column only. Take that same vector, and apply(int) to it.
If column x is float:
df.loc[df['x'].notnull(), 'x'] = df.loc[df['x'].notnull(), 'x'].apply(int)

if you have null value then in doing mathematical operation you will get this error to resolve it use df[~df['x'].isnull()]df[['x']].astype(int) if you want your dataset to be unchangeable.

Get an error message that says "Cannot convert non-finite values (NA or inf) to integer" despite me not having any infinity or NA values [duplicate]

Is there a preferred way to keep the data type of a numpy array fixed as int (or int64 or whatever), while still having an element inside listed as numpy.NaN?
In particular, I am converting an in-house data structure to a Pandas DataFrame. In our structure, we have integer-type columns that still have NaN's (but the dtype of the column is int). It seems to recast everything as a float if we make this a DataFrame, but we'd really like to be int.
Thoughts?
Things tried:
I tried using the from_records() function under pandas.DataFrame, with coerce_float=False and this did not help. I also tried using NumPy masked arrays, with NaN fill_value, which also did not work. All of these caused the column data type to become a float.

NaN can't be stored in an integer array. This is a known limitation of pandas at the moment; I have been waiting for progress to be made with NA values in NumPy (similar to NAs in R), but it will be at least 6 months to a year before NumPy gets these features, it seems:
http://pandas.pydata.org/pandas-docs/stable/gotchas.html#support-for-integer-na
(This feature has been added beginning with version 0.24 of pandas, but note it requires the use of extension dtype Int64 (capitalized), rather than the default dtype int64 (lower case):
https://pandas.pydata.org/pandas-docs/version/0.24/whatsnew/v0.24.0.html#optional-integer-na-support
)

This capability has been added to pandas beginning with version 0.24.
At this point, it requires the use of extension dtype 'Int64' (capitalized), rather than the default dtype 'int64' (lowercase).

If performance is not the main issue, you can store strings instead.
df.col = df.col.dropna().apply(lambda x: str(int(x)) )
Then you can mix then with NaN as much as you want. If you really want to have integers, depending on your application, you can use -1, or 0, or 1234567890, or some other dedicated value to represent NaN.
You can also temporarily duplicate the columns: one as you have, with floats; the other one experimental, with ints or strings. Then inserts asserts in every reasonable place checking that the two are in sync. After enough testing you can let go of the floats.

In case you are trying to convert a float (1.143) vector to integer (1), and that vector has NAs, converting it to the new 'Int64' dtype will give you an error. In order to solve this you have to round the numbers and then do ".astype('Int64')"
s1 = pd.Series([1.434, 2.343, np.nan])
#without round() the next line returns an error
s1.astype('Int64')
#cannot safely cast non-equivalent float64 to int64
##with round() it works
s1.round().astype('Int64')
0 1
1 2
2 NaN
dtype: Int64
My use case is that I have a float series that I want to round to int, but when you do .round() still has decimals, you need to convert to int to remove decimals.

This is not a solution for all cases, but mine (genomic coordinates) I've resorted to using 0 as NaN
a3['MapInfo'] = a3['MapInfo'].fillna(0).astype(int)
This at least allows for the proper 'native' column type to be used, operations like subtraction, comparison etc work as expected

Pandas v0.24+
Functionality to support NaN in integer series will be available in v0.24 upwards. There's information on this in the v0.24 "What's New" section, and more details under Nullable Integer Data Type.
Pandas v0.23 and earlier
In general, it's best to work with float series where possible, even when the series is upcast from int to float due to inclusion of NaN values. This enables vectorised NumPy-based calculations where, otherwise, Python-level loops would be processed.
The docs do suggest : "One possibility is to use dtype=object arrays instead." For example:
s = pd.Series([1, 2, 3, np.nan])
print(s.astype(object))
0 1
1 2
2 3
3 NaN
dtype: object
For cosmetic reasons, e.g. output to a file, this may be preferable.
Pandas v0.23 and earlier: background
NaN is considered a float. The docs currently (as of v0.23) specify the reason why integer series are upcasted to float:
In the absence of high performance NA support being built into NumPy
from the ground up, the primary casualty is the ability to represent
NAs in integer arrays.
This trade-off is made largely for memory and performance reasons, and
also so that the resulting Series continues to be “numeric”.
The docs also provide rules for upcasting due to NaN inclusion:
Typeclass Promotion dtype for storing NAs
floating no change
object no change
integer cast to float64
boolean cast to object

New for Pandas v1.00 +
You do not (and can not) use numpy.nan any more.
Now you have pandas.NA.
Please read: https://pandas.pydata.org/pandas-docs/stable/user_guide/integer_na.html
IntegerArray is currently experimental. Its API or implementation may
change without warning.
Changed in version 1.0.0: Now uses pandas.NA as the missing value
rather than numpy.nan.
In Working with missing data, we saw that pandas primarily uses NaN to
represent missing data. Because NaN is a float, this forces an array
of integers with any missing values to become floating point. In some
cases, this may not matter much. But if your integer column is, say,
an identifier, casting to float can be problematic. Some integers
cannot even be represented as floating point numbers.

If there are blanks in the text data, columns that would normally be integers will be cast to floats as float64 dtype because int64 dtype cannot handle nulls. This can cause inconsistent schema if you are loading multiple files some with blanks (which will end up as float64 and others without which will end up as int64
This code will attempt to convert any number type columns to Int64 (as opposed to int64) since Int64 can handle nulls
import pandas as pd
import numpy as np
#show datatypes before transformation
mydf.dtypes
for c in mydf.select_dtypes(np.number).columns:
try:
mydf[c] = mydf[c].astype('Int64')
print('casted {} as Int64'.format(c))
except:
print('could not cast {} to Int64'.format(c))
#show datatypes after transformation
mydf.dtypes

This is now possible, since pandas v 0.24.0
pandas 0.24.x release notes
Quote: "Pandas has gained the ability to hold integer dtypes with missing values.

I know that OP has asked for NumPy or Pandas only, but I think it is worth mentioning polars as an alternative that supports the requested feature.
In Polars any missing values in an integer column are simply null values and the column remains an integer column.
See Polars - User Guide > Coming from Pandas for more info.

Pandas (Python) reading and working on Java BigInteger/ large numbers

I have a data file (csv) with Nilsimsa hash values. Some of them would have as long as 80 characters. I wish to read them in Python for data analysis tasks. Is there a way to import the data in python without information loss?
EDIT: I have tried the implementations proposed in the comments but that does not work for me.
Example data in csv file would be: 77241756221441762028881402092817125017724447303212139981668021711613168152184106

Start with a simple text file to read in, just one variable and one row.
%more foo.txt
x
77241756221441762028881402092817125017724447303212139981668021711613168152184106
In [268]: df=pd.read_csv('foo.txt')
Pandas will read it in as a string because it's too big to store as a core number type like int64 or float64. But the info is there, you didn't lose anything.
In [269]: df.x
Out[269]:
0 7724175622144176202888140209281712501772444730...
Name: x, dtype: object
In [270]: type(df.x[0])
Out[270]: str
And you can use plain python to treat it as a number. Recall the caveats from the links in the comments, this isn't going to be as fast as stuff in numpy and pandas where you have stored a whole column as int64. This is using the more flexible but slower object mode to handle things.
You can change a column to be stored as longs (long integers) like this. (But note that the dtype is still object because everything except the core numpy types (int32, int64, float64, etc.) are stored as objects.)
In [271]: df.x = df.x.map(int)
And then can more or less treat it like a number.
In [272]: df.x * 2
Out[272]:
0 1544835124428835240577628041856342500354488946...
Name: x, dtype: object
You'll have to do some formatting to see the whole number. Or go the numpy route which will default to showing the whole number.
In [273]: df.x.values * 2
Out[273]: array([ 154483512442883524057762804185634250035448894606424279963336043423226336304368212L], dtype=object)

As explained by #JohnE in his answer that we do not lose any information while reading big numbers using Pandas. They are stored as dtype=object, to make numerical computation on them we need to transform this data into numerical type.
For series:
We have to apply the map(func) to the series in the dataframe:
df['columnName'].map(int)
Whole dataframe:
If for some reason, our entire dataframe is composed of columns with dtype=object, we look at applymap(func)
from the documentation of Pandas:
DataFrame.applymap(func): Apply a function to a DataFrame that is intended to operate elementwise, i.e. like doing map(func, series) for each series in the DataFrame
so to transform all columns in dataframe:
df.applymap(int)

unwanted type conversion in pandas.DataFrame.update

Is there any reason why pandas changes the type of columns from int to float in update, and can I prevent it from doing it? Here is some example code of the problem
import pandas as pd
import numpy as np
df = pd.DataFrame({'int': [1, 2], 'float': [np.nan, np.nan]})
print('Integer column:')
print(df['int'])
for _, df_sub in df.groupby('int'):
df_sub['float'] = float(df_sub['int'])
df.update(df_sub)
print('NO integer column:')
print(df['int'])

here's the reason for this: since you are effectively masking certain values on a column and replace them (with your updates), some values could become `nan
in an integer array this is impossible, so numeric dtypes are apriori converted to float (for efficiency), as checking first is more expensive that doing this
a change of dtype back is possible...just not in the code right now, therefor this a bug (a bit non-trivial to fix though): github.com/pydata/pandas/issues/4094

This causes data precision loss if you have big values in your int64 column, when update converts them to float. So going back with what Jeff suggests: df['int'].astype(int)
is not always possible.
My workaround for cases like this is:
df_sub['int'] = df_sub['int'].astype('Int64') # Int64 with capital I, supports NA values
df.update(df_sub)
df_sub['int'] = df_sub['int'].astype('int')
The above avoids the conversion to float type. The reason I am converting back to int type (instead of leaving it as Int64) is that pandas seems to lack support for that type in several operations (e.g. concat gives an error about missing .view).
Maybe they could incorporate the above fix in issue 4094