Python Pandas, write DataFrame to fixed-width file (to_fwf?)
Question:
I see that Pandas has read_fwf, but does it have something like DataFrame.to_fwf? I’m looking for support for field width, numerical precision, and string justification. It seems that DataFrame.to_csv doesn’t do this. numpy.savetxt does, but I wouldn’t want to do:
numpy.savetxt('myfile.txt', mydataframe.to_records(), fmt='some format')
That just seems wrong. Your ideas are much appreciated.
Answers:
I’m sure you found a workaround for this issue but for anyone else who is curious…
If you write the DF into a list, you can write it out to a file by giving the ‘format as a string’.format(list indices)
eg:
df=df.fillna('')
outF = 'output.txt'
dbOut = open(temp, 'w')
v = df.values.T.tolist()
for i in range(0,dfRows):
dbOut.write((
'{:7.2f}{:>6.2f}{:>2.0f}{:>4.0f}{:>5.0f}{:6.2f}{:6.2f}{:6.2f}{:6.1f {:>15}{:>60}'
.format(v[0][i],v[1][i],v[2][i],v[3][i],v[4][i],v[5][i],v[6][i],v[7][i],v[8][i],
v[9][i],v[10][i]) ))
dbOut.write("n")
dbOut.close
Just make sure to match up each index with the correct format 🙂
Hope that helps!
Until someone implements this in pandas, you can use the tabulate package:
import pandas as pd
from tabulate import tabulate
def to_fwf(df, fname):
content = tabulate(df.values.tolist(), list(df.columns), tablefmt="plain")
open(fname, "w").write(content)
pd.DataFrame.to_fwf = to_fwf
Python, Pandas : write content of DataFrame into text File
The question aboves answer helped me. It is not the best, but until to_fwf exists this will do the trick for me…
np.savetxt(r'c:datanp.txt', df.values, fmt='%d')
or
np.savetxt(r'c:datanp.txt', df.values, fmt='%10.5f')
For custom format for each column you can set format for whole line.
fmt param provides formatting for each line
with open('output.dat') as ofile:
fmt = '%.0f %02.0f %4.1f %3.0f %4.0f %4.1f %4.0f %4.1f %4.0f'
np.savetxt(ofile, df.values, fmt=fmt)
found a very simple solution! (Python). In the code snapped I am trying to write a DataFrame to a positional File. “finalDataFrame.values.tolist()” will return u a list in which each row of the DataFrame is turn into an another list just a [[‘Camry’,2019,’Toyota’],[‘Mustang’,’2016′,’Ford’]]. after that with the help of for loop and if statement i am trying to set its fix length. rest is obvious!
with open (FilePath,'w') as f:
for i in finalDataFrame.values.tolist():
widths=(0,0,0,0,0,0,0)
if i[2] == 'nan':
i[2]=''
for h in range(7):
i[2]= i[2] + ' '
else:
x=7-len(str(i[2]))
a=''
for k in range(x):
a=a+' '
i[2]=str(i[2])+a
if i[3] == '':
i[3]=''
for h in range(25):
i[3]=i[3]+' '
else:
x = 25 - len(i[3])
print(x)
a = ''
for k in range(x):
a = a + ' '
print(a)
i[3] = i[3] + a
i[4] = str(i[4])[:10]
q="".join("%*s" % i for i in zip(widths, i))
f.write(q+'n')
pandas.DataFrame.to_string() is all you need. The only trick is how to manage the index.
# Write
# df.reset_index(inplace=True) # uncomment if the index matters
df.to_string(filepath, index=False)
# Read
df = pd.read_fwf(filepath)
# df.set_index(index_names, inplace=True) # uncomment if the index matters
If the index is a pandas.Index that has no name, reset_index() should assign it to column "index". If it is a pandas.MultiIndex that has no names, it should be assigned to columns ["level_0", "level_1", ...].
Based on others’ answer, here is the snippet I wrote, not the best in coding and performance:
import pandas as pd
import pickle
import numpy as np
from tabulate import tabulate
left_align_gen = lambda length, value: eval(r"'{:<<<length>>}'.format('''<<value>>'''[0:<<length>>])".replace('<<length>>', str(length)).replace('<<value>>', str(value)))
right_align_gen = lambda length, value: eval(r"'{:><<length>>}'.format('''<<value>>'''[0:<<length>>])".replace('<<length>>', str(length)).replace('<<value>>', str(value)))
# df = pd.read_pickle("dummy.pkl")
with open("df.pkl", 'rb') as f:
df = pickle.load(f)
# field width defines here, width of each field
widths=(22, 255, 14, 255, 14, 255, 255, 255, 255, 255, 255, 22, 255, 22, 255, 255, 255, 22, 14, 14, 255, 255, 255, 2, )
# format datetime
df['CREATED_DATE'] = df['CREATED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['LAST_MODIFIED_DATE'] = df['LAST_MODIFIED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['TERMS_ACCEPTED_DATE'] = df['TERMS_ACCEPTED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['PRIVACY_ACCEPTED_DATE'] = df['PRIVACY_ACCEPTED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
# print(type(df.iloc[0]['CREATED_DATE']))
# print(df.iloc[0])
record_line_list = []
# for row in df.iloc[:10].itertuples():
for row in [tuple(x) for x in df.to_records(index=False)]:
record_line_list.append("".join(left_align_gen(length, value) for length, value in zip(widths, row)))
with open('output.txt', 'w') as f:
f.write('n'.join(record_line_list))
Try using hollerith. pypi github. It’s a (very) new python library that deals with fixed width formatting – and open to contribution. Unfortunately trying to use built-in python string formatting (or the C printf for that matter!) for fixed width doesn’t work well for some widths when there are large integers and certain double precision floats.
I see that Pandas has read_fwf, but does it have something like DataFrame.to_fwf? I’m looking for support for field width, numerical precision, and string justification. It seems that DataFrame.to_csv doesn’t do this. numpy.savetxt does, but I wouldn’t want to do:
numpy.savetxt('myfile.txt', mydataframe.to_records(), fmt='some format')
That just seems wrong. Your ideas are much appreciated.
I’m sure you found a workaround for this issue but for anyone else who is curious…
If you write the DF into a list, you can write it out to a file by giving the ‘format as a string’.format(list indices)
eg:
df=df.fillna('')
outF = 'output.txt'
dbOut = open(temp, 'w')
v = df.values.T.tolist()
for i in range(0,dfRows):
dbOut.write((
'{:7.2f}{:>6.2f}{:>2.0f}{:>4.0f}{:>5.0f}{:6.2f}{:6.2f}{:6.2f}{:6.1f {:>15}{:>60}'
.format(v[0][i],v[1][i],v[2][i],v[3][i],v[4][i],v[5][i],v[6][i],v[7][i],v[8][i],
v[9][i],v[10][i]) ))
dbOut.write("n")
dbOut.close
Just make sure to match up each index with the correct format 🙂
Hope that helps!
Until someone implements this in pandas, you can use the tabulate package:
import pandas as pd
from tabulate import tabulate
def to_fwf(df, fname):
content = tabulate(df.values.tolist(), list(df.columns), tablefmt="plain")
open(fname, "w").write(content)
pd.DataFrame.to_fwf = to_fwf
Python, Pandas : write content of DataFrame into text File
The question aboves answer helped me. It is not the best, but until to_fwf exists this will do the trick for me…
np.savetxt(r'c:datanp.txt', df.values, fmt='%d')
or
np.savetxt(r'c:datanp.txt', df.values, fmt='%10.5f')
For custom format for each column you can set format for whole line.
fmt param provides formatting for each line
with open('output.dat') as ofile:
fmt = '%.0f %02.0f %4.1f %3.0f %4.0f %4.1f %4.0f %4.1f %4.0f'
np.savetxt(ofile, df.values, fmt=fmt)
found a very simple solution! (Python). In the code snapped I am trying to write a DataFrame to a positional File. “finalDataFrame.values.tolist()” will return u a list in which each row of the DataFrame is turn into an another list just a [[‘Camry’,2019,’Toyota’],[‘Mustang’,’2016′,’Ford’]]. after that with the help of for loop and if statement i am trying to set its fix length. rest is obvious!
with open (FilePath,'w') as f:
for i in finalDataFrame.values.tolist():
widths=(0,0,0,0,0,0,0)
if i[2] == 'nan':
i[2]=''
for h in range(7):
i[2]= i[2] + ' '
else:
x=7-len(str(i[2]))
a=''
for k in range(x):
a=a+' '
i[2]=str(i[2])+a
if i[3] == '':
i[3]=''
for h in range(25):
i[3]=i[3]+' '
else:
x = 25 - len(i[3])
print(x)
a = ''
for k in range(x):
a = a + ' '
print(a)
i[3] = i[3] + a
i[4] = str(i[4])[:10]
q="".join("%*s" % i for i in zip(widths, i))
f.write(q+'n')
pandas.DataFrame.to_string() is all you need. The only trick is how to manage the index.
# Write
# df.reset_index(inplace=True) # uncomment if the index matters
df.to_string(filepath, index=False)
# Read
df = pd.read_fwf(filepath)
# df.set_index(index_names, inplace=True) # uncomment if the index matters
If the index is a pandas.Index that has no name, reset_index() should assign it to column "index". If it is a pandas.MultiIndex that has no names, it should be assigned to columns ["level_0", "level_1", ...].
Based on others’ answer, here is the snippet I wrote, not the best in coding and performance:
import pandas as pd
import pickle
import numpy as np
from tabulate import tabulate
left_align_gen = lambda length, value: eval(r"'{:<<<length>>}'.format('''<<value>>'''[0:<<length>>])".replace('<<length>>', str(length)).replace('<<value>>', str(value)))
right_align_gen = lambda length, value: eval(r"'{:><<length>>}'.format('''<<value>>'''[0:<<length>>])".replace('<<length>>', str(length)).replace('<<value>>', str(value)))
# df = pd.read_pickle("dummy.pkl")
with open("df.pkl", 'rb') as f:
df = pickle.load(f)
# field width defines here, width of each field
widths=(22, 255, 14, 255, 14, 255, 255, 255, 255, 255, 255, 22, 255, 22, 255, 255, 255, 22, 14, 14, 255, 255, 255, 2, )
# format datetime
df['CREATED_DATE'] = df['CREATED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['LAST_MODIFIED_DATE'] = df['LAST_MODIFIED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['TERMS_ACCEPTED_DATE'] = df['TERMS_ACCEPTED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
df['PRIVACY_ACCEPTED_DATE'] = df['PRIVACY_ACCEPTED_DATE'].apply(lambda x: x.to_pydatetime().strftime('%Y%m%d%H%M%S'))
# print(type(df.iloc[0]['CREATED_DATE']))
# print(df.iloc[0])
record_line_list = []
# for row in df.iloc[:10].itertuples():
for row in [tuple(x) for x in df.to_records(index=False)]:
record_line_list.append("".join(left_align_gen(length, value) for length, value in zip(widths, row)))
with open('output.txt', 'w') as f:
f.write('n'.join(record_line_list))
Try using hollerith. pypi github. It’s a (very) new python library that deals with fixed width formatting – and open to contribution. Unfortunately trying to use built-in python string formatting (or the C printf for that matter!) for fixed width doesn’t work well for some widths when there are large integers and certain double precision floats.