A single string in single quotes with PyYAML

When I edit a YAML file in Python with PyYAML, all of my string values are saved back to the original file without quotes.

one: valueOne
two: valueTwo
three: valueThree

I wanted one of those strings to be surrounded with single quotes:

one: valueOne
two: valueTwo
three: 'valueThree'

Changing the default_style parameter in yaml_dump affects whole file, which is not desired. I thought about adding single quotes to the beginning and end of a string that I want to be surrounded with:

valueThreeVariable = "'" + valueThreeVariable + "'"

However, this ends up with a dumped YAML looking like this:

one: valueOne
two: valueTwo
three: '''valueThree'''

I have tried escaping the single quote in various ways, using unicode or raw strings, all to no avail. How can I make only one of my YAML values to be a string surrounded with single quotes?

Answer

You can graft such functionality onto PyYAML but it is not easy. The value in the mapping for three has to be some instance of a class different from a normal string, otherwise the YAML dumper doesn’t know that it has to do something special and that instance is dumped as string with quotes. On loading scalars with single quotes need to be created as instances of this class. And apart from that you probably don’t want the keys of your dict/mapping scrambled as PyYAML does by default.

I do something similar to the above in my PyYAML derivative ruamel.yaml for block style scalars:

import ruamel.yaml

yaml_str = """
one: valueOne
two: valueTwo
three: |-
  valueThree
"""

data = ruamel.yaml.round_trip_load(yaml_str)
assert ruamel.yaml.round_trip_dump(data) == yaml_str

doesn’t throw an assertion error.

---

To start with the dumper, you can “convert” the valueThree string:

import ruamel.yaml
from ruamel.yaml.scalarstring import ScalarString

yaml_str = """
one: valueOne
two: valueTwo
three: 'valueThree'
"""

class SingleQuotedScalarString(ScalarString):
    def __new__(cls, value):
        return ScalarString.__new__(cls, value)

data = ruamel.yaml.round_trip_load(yaml_str)
data['three'] = SingleQuotedScalarString(data['three'])

but this cannot be dumped, as the dumper doesn’t know about the SingleQuotedScalarString. You can solve that in different ways, the following extends ruamel.yaml‘s RoundTripRepresenter class:

from ruamel.yaml.representer import RoundTripRepresenter
import sys

def _represent_single_quoted_scalarstring(self, data):
    tag = None
    style = "'"
    if sys.version_info < (3,) and not isinstance(data, unicode):
        data = unicode(data, 'ascii')
    tag = u'tag:yaml.org,2002:str'
    return self.represent_scalar(tag, data, style=style)

RoundTripRepresenter.add_representer(
    SingleQuotedScalarString,
    _represent_single_quoted_scalarstring)

assert ruamel.yaml.round_trip_dump(data) == yaml_str

Once again doesn’t throw an error. The above can be done in PyYAML and the safe_load/safe_dump in principle, but you would need to write code to preserve the key ordering, as well as some of the base functionality. (Apart from that PyYAML only supports the older YAML 1.1 standard not the YAML 1.2 standard from 2009).

To get the loading to work without using the explicit data['three'] = SingleQuotedScalarString(data['three']) conversion, you can add the following before the call to ruamel.yaml.round_trip_load():

from ruamel.yaml.constructor import RoundTripConstructor
from ruamel.yaml.nodes import ScalarNode
from ruamel.yaml.compat import text_type

def _construct_scalar(self, node):
    if not isinstance(node, ScalarNode):
        raise ConstructorError(
            None, None,
            "expected a scalar node, but found %s" % node.id,
            node.start_mark)

    if node.style == '|' and isinstance(node.value, text_type):
        return PreservedScalarString(node.value)
    elif node.style == "'" and isinstance(node.value, text_type):
        return SingleQuotedScalarString(node.value)
    return node.value

RoundTripConstructor.construct_scalar = _construct_scalar

There are different ways to do the above, including subclassing the RoundTripConstructor class, but the actual method to change is small and can easily be patched.

---

Combining all of the above and cleaning up a bit you get:

import ruamel.yaml
from ruamel.yaml.scalarstring import ScalarString
from ruamel.yaml.representer import RoundTripRepresenter
from ruamel.yaml.constructor import RoundTripConstructor
from ruamel.yaml.nodes import ScalarNode
from ruamel.yaml.compat import text_type, PY2


class SingleQuotedScalarString(ScalarString):
    def __new__(cls, value):
        return ScalarString.__new__(cls, value)


def _construct_scalar(self, node):
    if not isinstance(node, ScalarNode):
        raise ConstructorError(
            None, None,
            "expected a scalar node, but found %s" % node.id,
            node.start_mark)

    if node.style == '|' and isinstance(node.value, text_type):
        return PreservedScalarString(node.value)
    elif node.style == "'" and isinstance(node.value, text_type):
        return SingleQuotedScalarString(node.value)
    return node.value

RoundTripConstructor.construct_scalar = _construct_scalar


def _represent_single_quoted_scalarstring(self, data):
    tag = None
    style = "'"
    if PY2 and not isinstance(data, unicode):
        data = unicode(data, 'ascii')
    tag = u'tag:yaml.org,2002:str'
    return self.represent_scalar(tag, data, style=style)

RoundTripRepresenter.add_representer(
    SingleQuotedScalarString,
    _represent_single_quoted_scalarstring)


yaml_str = """
one: valueOne
two: valueTwo
three: 'valueThree'
"""

data = ruamel.yaml.round_trip_load(yaml_str)
assert ruamel.yaml.round_trip_dump(data) == yaml_str

Which still runs without assertion error, i.e. with dump output equalling input. As indicated you can do this in PyYAML, but it requires considerably more coding.

---

With a more modern version (ruamel.yaml>0.14) you can do:

yaml = ruamel.yaml.YAML()
yaml.preserve_quotes = True

data = yaml.load(yaml_str)
yaml.dump(data, sys.stdout)

and preserve the single quotes.