I have

>>> import yaml
>>> yaml.dump(u'abc')
"!!python/unicode 'abc'\n"

But I want

>>> import yaml
>>> yaml.dump(u'abc', magic="something")
'abc\n'

What magic param forces no tagging?

You can use safe_dump instead of dump. Just keep in mind that it won’t be able to represent arbitrary Python objects then. Also, when you load the YAML, you will get a str object instead of unicode.

How about this:

def unicode_representer(dumper, uni):
    node = yaml.ScalarNode(tag=u'tag:yaml.org,2002:str', value=uni)
    return node

yaml.add_representer(unicode, unicode_representer)

This seems to make dumping unicode objects work the same as dumping str objects for me (Python 2.6).

In [72]: yaml.dump(u'abc')
Out[72]: 'abc\n...\n'

In [73]: yaml.dump('abc')
Out[73]: 'abc\n...\n'

In [75]: yaml.dump(['abc'])
Out[75]: '[abc]\n'

In [76]: yaml.dump([u'abc'])
Out[76]: '[abc]\n'

You need a new dumper class that does everything the standard Dumper class does but overrides the representers for str and unicode.

from yaml.dumper import Dumper
from yaml.representer import SafeRepresenter

class KludgeDumper(Dumper):
   pass

KludgeDumper.add_representer(str,
       SafeRepresenter.represent_str)

KludgeDumper.add_representer(unicode,
        SafeRepresenter.represent_unicode)

Which leads to

>>> print yaml.dump([u'abc',u'abc\xe7'],Dumper=KludgeDumper)
[abc, "abc\xE7"]

>>> print yaml.dump([u'abc',u'abc\xe7'],Dumper=KludgeDumper,encoding=None)
[abc, "abc\xE7"]

Granted, I’m still stumped on how to keep this pretty.

>>> print u'abc\xe7'
abcç

And it breaks a later yaml.load()

>>> yy=yaml.load(yaml.dump(['abc','abc\xe7'],Dumper=KludgeDumper,encoding=None))
>>> yy
['abc', 'abc\xe7']
>>> print yy[1]
abc?
>>> print u'abc\xe7'
abcç

little addition to interjay’s excellent answer, you can keep your unicode on a reload if you take care of your file encodings.

# -*- coding: utf-8 -*-
import yaml
import codecs

data = dict(key = u"abcç\U0001F511")

fn = "test2.yaml"
with codecs.open(fn, "w", encoding="utf-8") as fo:
    yaml.safe_dump(data, fo)

with codecs.open(fn, encoding="utf-8") as fi:
    data2 = yaml.safe_load(fi)

print ("data2:", data2, "type(data.key):", type(data2.get("key")) )

print data2.get("key")

test2.yaml contents in my editor:

{key: "abc\xE7\uD83D\uDD11"}

print outputs:

('data2:', {'key': u'abc\xe7\U0001f511'}, 'type(data.key):', <type 'unicode'>)
abcç?

Plus, after reading http://nedbatchelder.com/blog/201302/war_is_peace.html I am pretty sure that safe_load/safe_dump is where I want to be anyway.

I’ve just started with Python and YAML, but probably this may also help. Just compare outputs:

def test_dump(self):
    print yaml.dump([{'name': 'value'}, {'name2': 1}], explicit_start=True)
    print yaml.dump_all([{'name': 'value'}, {'name2': 1}])