How can I remove extra whitespace from strings when parsing a csv file in Pandas?
PythonParsingPandasPython Problem Overview
I have the following file named 'data.csv':
1997,Ford,E350
1997, Ford , E350
1997,Ford,E350,"Super, luxurious truck"
1997,Ford,E350,"Super ""luxurious"" truck"
1997,Ford,E350," Super luxurious truck "
"1997",Ford,E350
1997,Ford,E350
2000,Mercury,Cougar
And I would like to parse it into a pandas DataFrame so that the DataFrame looks as follows:
Year Make Model Description
0 1997 Ford E350 None
1 1997 Ford E350 None
2 1997 Ford E350 Super, luxurious truck
3 1997 Ford E350 Super "luxurious" truck
4 1997 Ford E350 Super luxurious truck
5 1997 Ford E350 None
6 1997 Ford E350 None
7 2000 Mercury Cougar None
The best I could do was:
pd.read_table("data.csv", sep=r',', names=["Year", "Make", "Model", "Description"])
Which gets me:
Year Make Model Description
0 1997 Ford E350 None
1 1997 Ford E350 None
2 1997 Ford E350 Super, luxurious truck
3 1997 Ford E350 Super "luxurious" truck
4 1997 Ford E350 Super luxurious truck
5 1997 Ford E350 None
6 1997 Ford E350 None
7 2000 Mercury Cougar None
How can I get the DataFrame without those whitespaces?
Python Solutions
Solution 1 - Python
You could use converters:
import pandas as pd
def strip(text):
try:
return text.strip()
except AttributeError:
return text
def make_int(text):
return int(text.strip('" '))
table = pd.read_table("data.csv", sep=r',',
names=["Year", "Make", "Model", "Description"],
converters = {'Description' : strip,
'Model' : strip,
'Make' : strip,
'Year' : make_int})
print(table)
yields
Year Make Model Description
0 1997 Ford E350 None
1 1997 Ford E350 None
2 1997 Ford E350 Super, luxurious truck
3 1997 Ford E350 Super "luxurious" truck
4 1997 Ford E350 Super luxurious truck
5 1997 Ford E350 None
6 1997 Ford E350 None
7 2000 Mercury Cougar None
Solution 2 - Python
Adding parameter skipinitialspace=True to read_table worked for me.
So try:
pd.read_table("data.csv",
sep=r',',
names=["Year", "Make", "Model", "Description"],
skipinitialspace=True)
Same thing works in pd.read_csv().
Solution 3 - Python
Well, the whitespace is in your data, so you can't read in the data without reading in the whitespace. However, after you've read it in, you could strip out the whitespace by doing, e.g., df["Make"] = df["Make"].map(str.strip) (where df is your dataframe).
Solution 4 - Python
I don't have enough reputation to leave a comment, but the answer above suggesting using the map function along with strip won't work if you have NaN values, since strip only works on chars and NaN are floats.
There is a built-in pandas function to do this, which I used:
pd.core.strings.str_strip(df['Description'])
where df is your dataframe. In my case I used it on a dataframe with ~1.2 million rows and it was very fast.
Solution 5 - Python
I don't believe Pandas supported this at the time this question was posted but the the most straight forward way to do this is by using regex in the sep parameter of read_csv. So something like the following should work for this issue.
table = pd.read_table("data.csv", sep=' *, *')
Solution 6 - Python
read_table is Deprecated, Here is the message as it appears in the documentation.
> Deprecated since version 0.24.0. > > Use pandas.read_csv() instead, passing sep='\t' if necessary.
So using read_csv you can pass in a regex for the sep argument, where you can specify the separator as
sep="\s*,\s*"
any number of spaces, followed by a separator, followed by any number of space again, this will make sure all the leading and trailing spaces are also chosen as a delimiter chunk which in-turn removes the white-spaces on either side of your data.
regex details as follows:
\s -> white-space
* -> any number (zero or many)
, -> no meaning, direct character match
So, the regular expression \s*,\s* stands for white-space[any number] match a comma and white-space[any number].
if your delimiter is anything else other than a comma then replace the , in the above expression with your delimiter. Eg: \s*;\s* if ; is your delimiter.
Solution 7 - Python
Here's a function to iterate through each column and apply pd.core.strings.str_strip:
def df_strip(df):
df = df.copy()
for c in df.columns:
if df[c].dtype == np.object:
df[c] = pd.core.strings.str_strip(df[c])
df = df.rename(columns={c:c.strip()})
return df
Solution 8 - Python
The str.strip() function works really well on Series. Thus, I convert the dataframe column that contains the whitespaces into a Series, strip the whitespace using the str.strip() function and then replace the converted column back into the dataframe. Below is the example code.
import pandas as pd
data = pd.DataFrame({'values': [' ABC ', ' DEF', ' GHI ']})
new = pd.Series([])
new = data['values'].str.strip()
data['values'] = new
Solution 9 - Python
For me the best way was
def read_csv_regex(data, date_columns=[]):
df = pd.read_csv(data, quotechar='"', parse_dates=date_columns)
# remove front and ending blank spaces
df = df.replace({"^\s*|\s*$":""}, regex=True)
# if there remained only empty string "", change to Nan
df = df.replace({"":np.nan})
return df
You don't need to write converter function and set it to every column, it works for head and tail spaces and has now problem with quotas unlike regexp sep.