Related
I need help to convert simple_line.txt file to csv file using the pandas library. However, I am unable to categorize image file where i want to create all the values after first space in one column.
Here is the file (sample_list.txt), listed row by row:
Image Label
doc_pres223.jpg Durasal
doc_pres224.jpg Tab Cefepime
doc_pres225.jpg Tab Bleomycin
doc_pres226.jpg Budesonide is a corticosteroid,
doc_pres227.jpg prescribed for inflammatory,
I want the csv file to be like-
enter image description here
txt_file = r"./example.txt"
csv_file = r"./example.csv"
separator = "; "
with open(txt_file) as f_in, open(csv_file, "w+") as f_out:
for line in f_in:
f_out.write(separator.join(line.split(" ", maxsplit=1)))
try this:
import pandas as pd
def write_file(filename, output):
df = pd.DataFrame()
lines = open(filename, 'r').readlines()
for l in range(1, len(lines)):
line = lines[l]
arr = line.split(" ", maxsplit=1)
image_line = arr[0]
label_line = arr[1].replace('\n', '')
df = df.append({'Image': image_line, 'Label': label_line}, ignore_index=True)
df.to_csv(output)
if __name__ == '__main__':
write_file('example.txt', 'example.csv')
If the filenames in column Image is always the same length, then you could just treat is as a fixed width file. So the first column would be 15 characters, and the rest is the second column. Then just add two empty columns and write it to a new file.
# libraries
import pandas as pd
# set filename
filename = "simple_line.txt"
# read as fixed width
df = pd.read_fwf(filename, header=0, widths=[15, 100])
# add 2 empty columns
df.insert(1, 'empty1', '')
df.insert(2, 'empty2', '')
# save as a new csv file
filenew = "output.csv"
df.to_csv(filenew, sep=';', header=True, index=False)
I’m extremely new to Python & trying to figure the below out:
I have multiple CSV files (monthly files) that I’m trying to combine into a yearly file. The monthly files all have headers, so I’m trying to keep the first header & remove the rest. I used the below script which accomplished this, however there are 10 blank rows between each month.
Does anyone know what I can add to this to remove the blank rows?
import shutil
import glob
#import csv files from folder
path = r'data/US/market/merged_data'
allFiles = glob.glob(path + "/*.csv")
allFiles.sort() # glob lacks reliable ordering, so impose your own if output order matters
with open('someoutputfile.csv', 'wb') as outfile:
for i, fname in enumerate(allFiles):
with open(fname, 'rb') as infile:
if i != 0:
infile.readline() # Throw away header on all but first file
# Block copy rest of file from input to output without parsing
shutil.copyfileobj(infile, outfile)
print(fname + " has been imported.")
Thank you in advance!
assuming the dataset isn't bigger than you memory, I suggest reading each file in pandas, concatenating the dataframes and filtering from there. blank rows will probably show up as nan.
import pandas as pd
import glob
path = r'data/US/market/merged_data'
allFiles = glob.glob(path + "/*.csv")
allFiles.sort()
df = pd.Dataframe()
for i, fname in enumerate(allFiles):
#append data to existing dataframe
df = df.append(pd.read(fname), ignore_index = True)
#hopefully, this will drop blank rows
df = df.dropna(how = 'all')
#write to file
df.to_csv('someoutputfile.csv')
I've been trying to make a CSV from a big list of another CSVs and here's the deal: I want to get the names of these CSV files and put them in the CSV that I want to create, plus, I also need the row count from the CSV files that I'm getting the names of, here's what I've tried so far:
def getRegisters(file):
results = pd.read_csv(file, header = None, error_bad_lines= False, sep = '\t', low_memory = False)
print(len(results))
return len(results)
path = "C:/Users/gdldieca/Desktop/TESTSFORPANW/New folder"
dirs = os.listdir(path)
with open("C:/Users/gdldieca/Desktop/TESTSFORPANW/New folder/FilesNames.csv", 'w', newline='') as f:
writer = csv.writer(f, delimiter = '\t')
writer.writerow(("File", "Rows"))
for names in dirs:
sfile = getRegisters("C:/Users/gdldieca/Desktop/TESTSFORPANW/New folder/" + str(names))
writer.writerow((names, sfile))
However I can't seem to get the files row count even tho Pandas actually returns it. I'm getting this error:
_csv.Error: iterable expected, not int
The final result would be something like this written into the CSV
File1 90
File2 10
If you are using pandas , I think you can use also for make a csv file with all values that you need..Here an alternative
import os
import pandas as pd
directory='D:\\MY\\PATH\\ALLCSVFILE\\'
#create a list for add all
rows_list = []
for filename in os.listdir(directory):
if filename.endswith(".csv"):
file=os.path.join(directory, filename)
df=pd.read_csv(file)
#Count rows
rowcount=len(df.index)
new_row = {'namefile':filename, 'count':rowcount}
rows_list.append(new_row)
#pass list to dataframe
df1 = pd.DataFrame(rows_list)
print(df1)
df1.to_csv('test.csv', sep=',')
result :
Guys, I here have 200 separate csv files named from SH (1) to SH (200). I want to merge them into a single csv file. How can I do it?
As ghostdog74 said, but this time with headers:
with open("out.csv", "ab") as fout:
# first file:
with open("sh1.csv", "rb") as f:
fout.writelines(f)
# now the rest:
for num in range(2, 201):
with open("sh"+str(num)+".csv", "rb") as f:
next(f) # skip the header, portably
fout.writelines(f)
Why can't you just sed 1d sh*.csv > merged.csv?
Sometimes you don't even have to use python!
Use accepted StackOverflow answer to create a list of csv files that you want to append and then run this code:
import pandas as pd
combined_csv = pd.concat( [ pd.read_csv(f) for f in filenames ] )
And if you want to export it to a single csv file, use this:
combined_csv.to_csv( "combined_csv.csv", index=False )
fout=open("out.csv","a")
for num in range(1,201):
for line in open("sh"+str(num)+".csv"):
fout.write(line)
fout.close()
I'm just going to throw another code example into the basket:
from glob import glob
with open('singleDataFile.csv', 'a') as singleFile:
for csvFile in glob('*.csv'):
for line in open(csvFile, 'r'):
singleFile.write(line)
It depends what you mean by "merging" -- do they have the same columns? Do they have headers? For example, if they all have the same columns, and no headers, simple concatenation is sufficient (open the destination file for writing, loop over the sources opening each for reading, use shutil.copyfileobj from the open-for-reading source into the open-for-writing destination, close the source, keep looping -- use the with statement to do the closing on your behalf). If they have the same columns, but also headers, you'll need a readline on each source file except the first, after you open it for reading before you copy it into the destination, to skip the headers line.
If the CSV files don't all have the same columns then you need to define in what sense you're "merging" them (like a SQL JOIN? or "horizontally" if they all have the same number of lines? etc, etc) -- it's hard for us to guess what you mean in that case.
Quite easy to combine all files in a directory and merge them
import glob
import csv
# Open result file
with open('output.txt','wb') as fout:
wout = csv.writer(fout,delimiter=',')
interesting_files = glob.glob("*.csv")
h = True
for filename in interesting_files:
print 'Processing',filename
# Open and process file
with open(filename,'rb') as fin:
if h:
h = False
else:
fin.next()#skip header
for line in csv.reader(fin,delimiter=','):
wout.writerow(line)
A slight change to the code above as it does not actually work correctly.
It should be as follows...
from glob import glob
with open('main.csv', 'a') as singleFile:
for csv in glob('*.csv'):
if csv == 'main.csv':
pass
else:
for line in open(csv, 'r'):
singleFile.write(line)
If you are working on linux/mac you can do this.
from subprocess import call
script="cat *.csv>merge.csv"
call(script,shell=True)
If the merged CSV is going to be used in Python then just use glob to get a list of the files to pass to fileinput.input() via the files argument, then use the csv module to read it all in one go.
OR, you could just do
cat sh*.csv > merged.csv
You can simply use the in-built csv library. This solution will work even if some of your CSV files have slightly different column names or headers, unlike the other top-voted answers.
import csv
import glob
filenames = [i for i in glob.glob("SH*.csv")]
header_keys = []
merged_rows = []
for filename in filenames:
with open(filename) as f:
reader = csv.DictReader(f)
merged_rows.extend(list(reader))
header_keys.extend([key for key in reader.fieldnames if key not in header_keys])
with open("combined.csv", "w") as f:
w = csv.DictWriter(f, fieldnames=header_keys)
w.writeheader()
w.writerows(merged_rows)
The merged file will contain all possible columns (header_keys) that can be found in the files. Any absent columns in a file would be rendered as blank / empty (but preserving rest of the file's data).
Note:
This won't work if your CSV files have no headers. In that case you can still use the csv library, but instead of using DictReader & DictWriter, you'll have to work with the basic reader & writer.
This may run into issues when you are dealing with massive data since the entirety of the content is being store in memory (merged_rows list).
Over the solution that made #Adders and later on improved by #varun, I implemented some little improvement too leave the whole merged CSV with only the main header:
from glob import glob
filename = 'main.csv'
with open(filename, 'a') as singleFile:
first_csv = True
for csv in glob('*.csv'):
if csv == filename:
pass
else:
header = True
for line in open(csv, 'r'):
if first_csv and header:
singleFile.write(line)
first_csv = False
header = False
elif header:
header = False
else:
singleFile.write(line)
singleFile.close()
Best regards!!!
You could import csv then loop through all the CSV files reading them into a list. Then write the list back out to disk.
import csv
rows = []
for f in (file1, file2, ...):
reader = csv.reader(open("f", "rb"))
for row in reader:
rows.append(row)
writer = csv.writer(open("some.csv", "wb"))
writer.writerows("\n".join(rows))
The above is not very robust as it has no error handling nor does it close any open files.
This should work whether or not the the individual files have one or more rows of CSV data in them. Also I did not run this code, but it should give you an idea of what to do.
I modified what #wisty said to be worked with python 3.x, for those of you that have encoding problem, also I use os module to avoid of hard coding
import os
def merge_all():
dir = os.chdir('C:\python\data\\')
fout = open("merged_files.csv", "ab")
# first file:
for line in open("file_1.csv",'rb'):
fout.write(line)
# now the rest:
list = os.listdir(dir)
number_files = len(list)
for num in range(2, number_files):
f = open("file_" + str(num) + ".csv", 'rb')
f.__next__() # skip the header
for line in f:
fout.write(line)
f.close() # not really needed
fout.close()
Here is a script:
Concatenating csv files named SH1.csv to SH200.csv
Keeping the headers
import glob
import re
# Looking for filenames like 'SH1.csv' ... 'SH200.csv'
pattern = re.compile("^SH([1-9]|[1-9][0-9]|1[0-9][0-9]|200).csv$")
file_parts = [name for name in glob.glob('*.csv') if pattern.match(name)]
with open("file_merged.csv","wb") as file_merged:
for (i, name) in enumerate(file_parts):
with open(name, "rb") as file_part:
if i != 0:
next(file_part) # skip headers if not first file
file_merged.write(file_part.read())
Updating wisty's answer for python3
fout=open("out.csv","a")
# first file:
for line in open("sh1.csv"):
fout.write(line)
# now the rest:
for num in range(2,201):
f = open("sh"+str(num)+".csv")
next(f) # skip the header
for line in f:
fout.write(line)
f.close() # not really needed
fout.close()
Let's say you have 2 csv files like these:
csv1.csv:
id,name
1,Armin
2,Sven
csv2.csv:
id,place,year
1,Reykjavik,2017
2,Amsterdam,2018
3,Berlin,2019
and you want the result to be like this csv3.csv:
id,name,place,year
1,Armin,Reykjavik,2017
2,Sven,Amsterdam,2018
3,,Berlin,2019
Then you can use the following snippet to do that:
import csv
import pandas as pd
# the file names
f1 = "csv1.csv"
f2 = "csv2.csv"
out_f = "csv3.csv"
# read the files
df1 = pd.read_csv(f1)
df2 = pd.read_csv(f2)
# get the keys
keys1 = list(df1)
keys2 = list(df2)
# merge both files
for idx, row in df2.iterrows():
data = df1[df1['id'] == row['id']]
# if row with such id does not exist, add the whole row
if data.empty:
next_idx = len(df1)
for key in keys2:
df1.at[next_idx, key] = df2.at[idx, key]
# if row with such id exists, add only the missing keys with their values
else:
i = int(data.index[0])
for key in keys2:
if key not in keys1:
df1.at[i, key] = df2.at[idx, key]
# save the merged files
df1.to_csv(out_f, index=False, encoding='utf-8', quotechar="", quoting=csv.QUOTE_NONE)
With the help of a loop you can achieve the same result for multiple files as it is in your case (200 csv files).
If the files aren't numbered in order, take the hassle-free approach below:
Python 3.6 on windows machine:
import pandas as pd
from glob import glob
interesting_files = glob("C:/temp/*.csv") # it grabs all the csv files from the directory you mention here
df_list = []
for filename in sorted(interesting_files):
df_list.append(pd.read_csv(filename))
full_df = pd.concat(df_list)
# save the final file in same/different directory:
full_df.to_csv("C:/temp/merged_pandas.csv", index=False)
An easy-to-use function:
def csv_merge(destination_path, *source_paths):
'''
Merges all csv files on source_paths to destination_path.
:param destination_path: Path of a single csv file, doesn't need to exist
:param source_paths: Paths of csv files to be merged into, needs to exist
:return: None
'''
with open(destination_path,"a") as dest_file:
with open(source_paths[0]) as src_file:
for src_line in src_file.read():
dest_file.write(src_line)
source_paths.pop(0)
for i in range(len(source_paths)):
with open(source_paths[i]) as src_file:
src_file.next()
for src_line in src_file:
dest_file.write(src_line)
import pandas as pd
import os
df = pd.read_csv("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data\\Sales_April_2019.csv")
files = [file for file in os.listdir("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data")
for file in files:
print(file)
all_data = pd.DataFrame()
for file in files:
df=pd.read_csv("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data\\"+file)
all_data = pd.concat([all_data,df])
all_data.head()
I have done it by implementing a function that expect output file and paths of the input files.
The function copy the file content of the first file into the output file and then does the same for the rest of input files but without the header line.
def concat_files_with_header(output_file, *paths):
for i, path in enumerate(paths):
with open(path) as input_file:
if i > 0:
next(input_file) # Skip header
output_file.writelines(input_file)
Usage example of the function:
if __name__ == "__main__":
paths = [f"sh{i}.csv" for i in range(1, 201)]
with open("output.csv", "w") as output_file:
concat_files_with_header(output_file, *paths)
Guys, I here have 200 separate csv files named from SH (1) to SH (200). I want to merge them into a single csv file. How can I do it?
As ghostdog74 said, but this time with headers:
with open("out.csv", "ab") as fout:
# first file:
with open("sh1.csv", "rb") as f:
fout.writelines(f)
# now the rest:
for num in range(2, 201):
with open("sh"+str(num)+".csv", "rb") as f:
next(f) # skip the header, portably
fout.writelines(f)
Why can't you just sed 1d sh*.csv > merged.csv?
Sometimes you don't even have to use python!
Use accepted StackOverflow answer to create a list of csv files that you want to append and then run this code:
import pandas as pd
combined_csv = pd.concat( [ pd.read_csv(f) for f in filenames ] )
And if you want to export it to a single csv file, use this:
combined_csv.to_csv( "combined_csv.csv", index=False )
fout=open("out.csv","a")
for num in range(1,201):
for line in open("sh"+str(num)+".csv"):
fout.write(line)
fout.close()
I'm just going to throw another code example into the basket:
from glob import glob
with open('singleDataFile.csv', 'a') as singleFile:
for csvFile in glob('*.csv'):
for line in open(csvFile, 'r'):
singleFile.write(line)
It depends what you mean by "merging" -- do they have the same columns? Do they have headers? For example, if they all have the same columns, and no headers, simple concatenation is sufficient (open the destination file for writing, loop over the sources opening each for reading, use shutil.copyfileobj from the open-for-reading source into the open-for-writing destination, close the source, keep looping -- use the with statement to do the closing on your behalf). If they have the same columns, but also headers, you'll need a readline on each source file except the first, after you open it for reading before you copy it into the destination, to skip the headers line.
If the CSV files don't all have the same columns then you need to define in what sense you're "merging" them (like a SQL JOIN? or "horizontally" if they all have the same number of lines? etc, etc) -- it's hard for us to guess what you mean in that case.
Quite easy to combine all files in a directory and merge them
import glob
import csv
# Open result file
with open('output.txt','wb') as fout:
wout = csv.writer(fout,delimiter=',')
interesting_files = glob.glob("*.csv")
h = True
for filename in interesting_files:
print 'Processing',filename
# Open and process file
with open(filename,'rb') as fin:
if h:
h = False
else:
fin.next()#skip header
for line in csv.reader(fin,delimiter=','):
wout.writerow(line)
A slight change to the code above as it does not actually work correctly.
It should be as follows...
from glob import glob
with open('main.csv', 'a') as singleFile:
for csv in glob('*.csv'):
if csv == 'main.csv':
pass
else:
for line in open(csv, 'r'):
singleFile.write(line)
If you are working on linux/mac you can do this.
from subprocess import call
script="cat *.csv>merge.csv"
call(script,shell=True)
If the merged CSV is going to be used in Python then just use glob to get a list of the files to pass to fileinput.input() via the files argument, then use the csv module to read it all in one go.
OR, you could just do
cat sh*.csv > merged.csv
You can simply use the in-built csv library. This solution will work even if some of your CSV files have slightly different column names or headers, unlike the other top-voted answers.
import csv
import glob
filenames = [i for i in glob.glob("SH*.csv")]
header_keys = []
merged_rows = []
for filename in filenames:
with open(filename) as f:
reader = csv.DictReader(f)
merged_rows.extend(list(reader))
header_keys.extend([key for key in reader.fieldnames if key not in header_keys])
with open("combined.csv", "w") as f:
w = csv.DictWriter(f, fieldnames=header_keys)
w.writeheader()
w.writerows(merged_rows)
The merged file will contain all possible columns (header_keys) that can be found in the files. Any absent columns in a file would be rendered as blank / empty (but preserving rest of the file's data).
Note:
This won't work if your CSV files have no headers. In that case you can still use the csv library, but instead of using DictReader & DictWriter, you'll have to work with the basic reader & writer.
This may run into issues when you are dealing with massive data since the entirety of the content is being store in memory (merged_rows list).
Over the solution that made #Adders and later on improved by #varun, I implemented some little improvement too leave the whole merged CSV with only the main header:
from glob import glob
filename = 'main.csv'
with open(filename, 'a') as singleFile:
first_csv = True
for csv in glob('*.csv'):
if csv == filename:
pass
else:
header = True
for line in open(csv, 'r'):
if first_csv and header:
singleFile.write(line)
first_csv = False
header = False
elif header:
header = False
else:
singleFile.write(line)
singleFile.close()
Best regards!!!
You could import csv then loop through all the CSV files reading them into a list. Then write the list back out to disk.
import csv
rows = []
for f in (file1, file2, ...):
reader = csv.reader(open("f", "rb"))
for row in reader:
rows.append(row)
writer = csv.writer(open("some.csv", "wb"))
writer.writerows("\n".join(rows))
The above is not very robust as it has no error handling nor does it close any open files.
This should work whether or not the the individual files have one or more rows of CSV data in them. Also I did not run this code, but it should give you an idea of what to do.
I modified what #wisty said to be worked with python 3.x, for those of you that have encoding problem, also I use os module to avoid of hard coding
import os
def merge_all():
dir = os.chdir('C:\python\data\\')
fout = open("merged_files.csv", "ab")
# first file:
for line in open("file_1.csv",'rb'):
fout.write(line)
# now the rest:
list = os.listdir(dir)
number_files = len(list)
for num in range(2, number_files):
f = open("file_" + str(num) + ".csv", 'rb')
f.__next__() # skip the header
for line in f:
fout.write(line)
f.close() # not really needed
fout.close()
Here is a script:
Concatenating csv files named SH1.csv to SH200.csv
Keeping the headers
import glob
import re
# Looking for filenames like 'SH1.csv' ... 'SH200.csv'
pattern = re.compile("^SH([1-9]|[1-9][0-9]|1[0-9][0-9]|200).csv$")
file_parts = [name for name in glob.glob('*.csv') if pattern.match(name)]
with open("file_merged.csv","wb") as file_merged:
for (i, name) in enumerate(file_parts):
with open(name, "rb") as file_part:
if i != 0:
next(file_part) # skip headers if not first file
file_merged.write(file_part.read())
Updating wisty's answer for python3
fout=open("out.csv","a")
# first file:
for line in open("sh1.csv"):
fout.write(line)
# now the rest:
for num in range(2,201):
f = open("sh"+str(num)+".csv")
next(f) # skip the header
for line in f:
fout.write(line)
f.close() # not really needed
fout.close()
Let's say you have 2 csv files like these:
csv1.csv:
id,name
1,Armin
2,Sven
csv2.csv:
id,place,year
1,Reykjavik,2017
2,Amsterdam,2018
3,Berlin,2019
and you want the result to be like this csv3.csv:
id,name,place,year
1,Armin,Reykjavik,2017
2,Sven,Amsterdam,2018
3,,Berlin,2019
Then you can use the following snippet to do that:
import csv
import pandas as pd
# the file names
f1 = "csv1.csv"
f2 = "csv2.csv"
out_f = "csv3.csv"
# read the files
df1 = pd.read_csv(f1)
df2 = pd.read_csv(f2)
# get the keys
keys1 = list(df1)
keys2 = list(df2)
# merge both files
for idx, row in df2.iterrows():
data = df1[df1['id'] == row['id']]
# if row with such id does not exist, add the whole row
if data.empty:
next_idx = len(df1)
for key in keys2:
df1.at[next_idx, key] = df2.at[idx, key]
# if row with such id exists, add only the missing keys with their values
else:
i = int(data.index[0])
for key in keys2:
if key not in keys1:
df1.at[i, key] = df2.at[idx, key]
# save the merged files
df1.to_csv(out_f, index=False, encoding='utf-8', quotechar="", quoting=csv.QUOTE_NONE)
With the help of a loop you can achieve the same result for multiple files as it is in your case (200 csv files).
If the files aren't numbered in order, take the hassle-free approach below:
Python 3.6 on windows machine:
import pandas as pd
from glob import glob
interesting_files = glob("C:/temp/*.csv") # it grabs all the csv files from the directory you mention here
df_list = []
for filename in sorted(interesting_files):
df_list.append(pd.read_csv(filename))
full_df = pd.concat(df_list)
# save the final file in same/different directory:
full_df.to_csv("C:/temp/merged_pandas.csv", index=False)
An easy-to-use function:
def csv_merge(destination_path, *source_paths):
'''
Merges all csv files on source_paths to destination_path.
:param destination_path: Path of a single csv file, doesn't need to exist
:param source_paths: Paths of csv files to be merged into, needs to exist
:return: None
'''
with open(destination_path,"a") as dest_file:
with open(source_paths[0]) as src_file:
for src_line in src_file.read():
dest_file.write(src_line)
source_paths.pop(0)
for i in range(len(source_paths)):
with open(source_paths[i]) as src_file:
src_file.next()
for src_line in src_file:
dest_file.write(src_line)
import pandas as pd
import os
df = pd.read_csv("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data\\Sales_April_2019.csv")
files = [file for file in os.listdir("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data")
for file in files:
print(file)
all_data = pd.DataFrame()
for file in files:
df=pd.read_csv("e:\\data science\\kaggle assign\\monthly sales\\Pandas-Data-Science-Tasks-master\\SalesAnalysis\\Sales_Data\\"+file)
all_data = pd.concat([all_data,df])
all_data.head()
I have done it by implementing a function that expect output file and paths of the input files.
The function copy the file content of the first file into the output file and then does the same for the rest of input files but without the header line.
def concat_files_with_header(output_file, *paths):
for i, path in enumerate(paths):
with open(path) as input_file:
if i > 0:
next(input_file) # Skip header
output_file.writelines(input_file)
Usage example of the function:
if __name__ == "__main__":
paths = [f"sh{i}.csv" for i in range(1, 201)]
with open("output.csv", "w") as output_file:
concat_files_with_header(output_file, *paths)