如何使用utf-8编码将DataFrame导出到HTML？

Question

I keep getting:

UnicodeEncodeError: 'ascii' codec can't encode characters in position 265-266: ordinal not in range(128)

when I try:

df.to_html("mypage.html")

here is a sample of how to reproduce the problem:

df = pd.DataFrame({"a": [u'Rue du Gu\xc3\xa9, 78120 Sonchamp'], "b": [u"some other thing"]})
df.to_html("mypage.html")

the list of elements in "a" are of type "unicode" .

when I want to export it to csv it works because you can do:

df.to_csv("myfile.csv", encoding="utf-8")

Answer 1

Your problem is in other code. Your sample code has a Unicode string that has been mis-decoded as latin1 , Windows-1252 , or similar, since it has UTF-8 sequences in it. Here I undo the bad decoding and redecode as UTF-8, but you'll want to find where the wrong decode is being performed:

>>> s = u'Rue du Gu\xc3\xa9, 78120 Sonchamp'
>>> s.encode('latin1').decode('utf8')
u'Rue du Gu\xe9, 78120 Sonchamp'
>>> print(s.encode('latin1').decode('utf8'))
Rue du Gué, 78120 Sonchamp

Answer 2

The issue is actually in using df.to_html("mypage.html") to save the HTML to a file directly. If instead you write the file yourself, you can avoid this encoding bug with pandas.

html = df.to_html()
with open("mypage.html", "w", encoding="utf-8") as file:
    file.write(html)

You may also need to specify the character set in the head of the HTML for it to show up properly on certain browsers (HTML5 has UTF-8 as default):

<meta charset="UTF-8">

This was the only method that worked for me out of the several I've seen.

Answer 3

The way it worked for me:

html = df.to_html()

with open("dataframe.html", "w", encoding="utf-8") as file:
    file.writelines('<meta charset="UTF-8">\n')
    file.write(html)

Answer 4

If you really need to keep the output to html, you could try cleaning the code in a numpy array before writing to_html.

df = pd.DataFrame({"a": [u'Rue du Gu\xc3\xa9, 78120 Sonchamp'], "b": [u"some other thing"]})

def clean_unicode(df):
   *#Transforms the DataFrame to Numpy array*
   df=df.as_matrix()
   *#Encode all strings with special characters* 
   for x in np.nditer(df, flags=['refs_ok'], op_flags =['copy', 'readonly']):
         df[df==x]=str(str(x).encode("latin-1", "replace").decode('utf8'))
   *#Transform the Numpy array to Dataframe again*
   df=pd.DataFrame(df)
   return df

df=clean_unicode(df)
df.to_html("Results.html') -----> Success!