thriftpy
thriftpy copied to clipboard
Thriftpy
fix for absolute paths on windows
An absolute path on windows looks like 'C:\foo\bar.thrift' and urlparse returns 'c' as the scheme which causes a parse error to be thrown when given an absolute path.
>>> import thriftpy
>>> thriftpy.load('C:\\foo\\bar.thrift', 'Hbase_thrift')
Traceback (most recent call last):
...
File "C:\Python27\lib\site-packages\thriftpy\parser\parser.py", line 501, in parse
url_scheme))
ThriftParserError: ThriftPy does not support generating module with path in protocol 'c'
On Windows any letter 'a-zA-Z' is a valid drive letter and empty string means the current drive and so we can handle this completely with the following test.
if len(url_scheme) <= 1:
is roughly equivalent to,
if url_scheme == '' or \
url_scheme in tuple('abcdefghijklmnopqrstuvwxyz'):
In urlparse these are the valid scheme chars,
'abcdefghijklmnopqrstuvwxyz'
'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
'0123456789'
'+-.'
If any other character appears, urlparse().scheme will return an empty string. A single digit number will never be returned in the scheme since urlparse handles this as a port number and would also return an empty string. Similarly urlparse always returns the lower case version of the scheme and so uppercase letters never appear. Lastly dot '.' is a valid scheme for example .\foo\bar.thrift
This leaves only two characters '+-'
Its not really necessary to handle these two characters since the worst case is the user gets a detailed IOError Exception
>>> open('+:\\fakefile')
Traceback (most recent call last):
File "<pyshell#11>", line 1, in <module>
open('+:\\fakefile')
IOError: [Errno 2] No such file or directory: '+:\\fakefile'
>>>
which makes more sense than the "ThriftParserError: ... protocol 'c' " exception and it stops it from failing for absolute paths on windows.
Please incorporate this fix in the next release of thriftypy. Thanks
I still think you should incorporate my suggested change... sure you now handle the 'file://' scheme but with my fix you can also handle being given direct paths and it now works on linux and windows.
I can confirm I am facing the issue on the Windows platform and have had to manually update parser.py for the time being. Hopefully, a proper fix will be promoted soon.
Guys, I'm also hitting this issue when trying to use py_zipkin on windows. I'd appreciate if this 2 year old commit gets merged. Thanks!
