yaml.dump adding unwanted newlines in multiline strings

Better ruamel.yaml option for outputting multiline strings as blocks:

from ruamel.yaml.representer import RoundTripRepresenter
from ruamel.yaml import YAML

multiline_string = """\
-----BEGIN RSA PRIVATE KEY-----
MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
...
"""


def repr_str(dumper: RoundTripRepresenter, data: str):
    if '\n' in data:
        return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='|')
    return dumper.represent_scalar('tag:yaml.org,2002:str', data)


yaml = YAML()
yaml.representer.add_representer(str, repr_str)

with open('file.yaml', 'w') as fp:
    yaml.dump({'a': 1, 'b': 'hello world', 'c': multiline_string}, fp)

Output:

a: 1
b: hello world
c: |
  -----BEGIN RSA PRIVATE KEY-----
  MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
  xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
  ...

If that is the only thing going into your YAML file then you can dump with the option default_style='|' which gives you block style literal for all of your scalars (probably not what you want).

Your string, contains no special characters (that need \ escaping and double quotes), because of the newlines PyYAML decides to represented single quoted. In single quoted style a double newline is the way to represent a single newline that occurred in string that is represented. This gets "undone" on loading, but is indeed not very readable.

If you want to get the block style literals on an individual basis, you can do multiple things:

• adapt the Representer to output all strings with embedded newlines using the literal scalar block style (assuming they don't need \ escaping of special characters, which will force double quotes)

  import sys
  import yaml
  
  x = u"""\
  -----BEGIN RSA PRIVATE KEY-----
  MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
  xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
  ...
  """
  
  yaml.SafeDumper.org_represent_str = yaml.SafeDumper.represent_str
  
  def repr_str(dumper, data):
      if '\n' in data:
          return dumper.represent_scalar(u'tag:yaml.org,2002:str', data, style='|')
      return dumper.org_represent_str(data)
  
  yaml.add_representer(str, repr_str, Dumper=yaml.SafeDumper)
  
  yaml.safe_dump(dict(a=1, b='hello world', c=x), sys.stdout)
  

• make a subclass of string, that has its special representer. You should be able to take the code for that from here, here and here:

  import sys
  import yaml
  
  class PSS(str):
      pass
  
  x = PSS("""\
  -----BEGIN RSA PRIVATE KEY-----
  MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
  xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
  ...
  """)
  
  def pss_representer(dumper, data):
          style = '|'
          # if sys.versioninfo < (3,) and not isinstance(data, unicode):
          #     data = unicode(data, 'ascii')
          tag = u'tag:yaml.org,2002:str'
          return dumper.represent_scalar(tag, data, style=style)
  
  yaml.add_representer(PSS, pss_representer, Dumper=yaml.SafeDumper)
  
  yaml.safe_dump(dict(a=1, b='hello world', c=x), sys.stdout)
  

• use ruamel.yaml:

  import sys
  from ruamel.yaml import YAML
  from ruamel.yaml.scalarstring import PreservedScalarString as pss
  
  x = pss("""\
  -----BEGIN RSA PRIVATE KEY-----
  MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
  xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
  ...
  """)
  
  yaml = YAML()
  
  yaml.dump(dict(a=1, b='hello world', c=x), sys.stdout)
  

All of these give:

a: 1
b: hello world
c: |
  -----BEGIN RSA PRIVATE KEY-----
  MIIEogIBAAKCAQEA6oySC+8/N9VNpk0gJS7Gk8vn9sYN7FhjpAQnoHRqTN/Oaiyx
  xk2AleP2vXpojA/DHldT1JO+o3j56AHD+yfNFFeYvgWKDY35g49HsZZhbyCEAB45
  ...

Please note that it is not necessary to specify default_flow_style=False as the literal scalars can only appear in block style.

Adding onto Anthon's answer, I found some documentation for other options I can pass to default_style at this link here.

The best compromise I could find for representing all of my data was:

with open('foo.yaml', 'w') as f:
  yaml.safe_dump(secrets, f, explicit_start=True, default_style='\"', width=4096)

And this led to a YAML file that looks like:

---
"alex_test": "yyyyyyyy"
"alex_test_key": "-----BEGIN RSA PRIVATE KEY-----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\n+fDJX+yvxnZsY5hw6cYwC3C/axS9jq63oQ7i8FXwG/a0breCGu8=\n-----END RSA PRIVATE KEY-----"

I would have preferred to use ruamel.yaml, but this code must run in an environment where I can only use the default Python packages that I have available.