w3lib.url.safe_url_string incorrectly encode IDNA domain with port

[heroesm]

Step to reproduce:

>>> from w3lib.url import safe_url_string
>>> safe_url_string('http://新华网.中国')
'http://xn--xkrr14bows.xn--fiqs8s'
>>> safe_url_string('http://新华网.中国:80')
'http://xn--xkrr14bows.xn--:80-u68dy61b'

safe_url_string('http://新华网.中国:80')
expected result:

'http://xn--xkrr14bows.xn--fiqs8s:80'

real result:

'http://xn--xkrr14bows.xn--:80-u68dy61b'

Related code:

w3lib/w3lib/url.py

Line 80 in ef5c110

netloc = parts.netloc.encode('idna')

netloc = parts.netloc.encode('idna')

Maybe IDNA encoding should be done on hostname rather than netloc.

[wRAR]

Maybe IDNA encoding should be done on hostname rather than netloc.

Indeed, thank you for this finding.

https://stackoverflow.com/a/804380/113586 and its comments also suggest extracting user@pass from netloc but I'm not sure if we need this.