[reportlab-users] encoding issue in utf8str() in pdfdoc.py

[Henning von Bargen]

...

>

>      When I tried to generate a pdf file whose path name included some Chinese characters, under Windows XP, some error occured:

>

> Traceback (most recent call last):

>    File "D:\python\byj1018\main.py", line 1278, in MenuPrint

>      pdf.build(story)

>    File "C:\Python26\Lib\site-packages\reportlab\platypus\doctemplate.py", line 1110, in build

>      BaseDocTemplate.build(self,flowables, canvasmaker=canvasmaker)

>    File "C:\Python26\Lib\site-packages\reportlab\platypus\doctemplate.py", line 903, in build

>      self._endBuild()

>    File "C:\Python26\Lib\site-packages\reportlab\platypus\doctemplate.py", line 845, in _endBuild

>      if getattr(self,'_doSave',1): self.canv.save()

>    File "C:\Python26\Lib\site-packages\reportlab\pdfgen\canvas.py", line 1091, in save

>      self._doc.SaveToFile(self._filename, self)

>    File "C:\Python26\Lib\site-packages\reportlab\pdfbase\pdfdoc.py", line 234, in SaveToFile

>      f = open(filename, "wb")

> IOError: [Errno 2] No such file or directory: 'D:\\python\\byj1018\\data\\IQC\xe8\xaf\x95\xe9\xaa\x8c3\\IQC\xe8\xaf\x95\xe9\xaa\x8c3.pdf'

>

>     But the same code went well under Linux.

>     I thought it was caused by the wrong encoding by the function utf8str() in pdfdoc.py. In that file, utf8 was always

> used to encode unicode string. But Windows XP does not recognise the utf8 string by default. Then I modified that function a bit,

> things went well.

>     So I suggest to get system default encoding from sys.stdin instead of the fixed utf8, such as:

>

> import sys

> def utf8str(x):

>      if isinstance(x,unicode):

>          return x.encode(sys.stdin.encoding)

>      else:

>          return str(x)

>

> Best regards

> Tubo

> --


Probably not a good idea.
Instead, you better remove the 2 calls to utf8str in saveToFile
entirely and use the filename as is.
Then, when you specify the filename argument, use a unicode
object instead of a byte-string (the built-in open works fine with unicode).

HTH
Henning