Avoid references in PyYAML

Question

I use YAML with PyYAML. Is there a way to avoid the *id002 references after dumping a nested structure? For readability I'd like to see the actual (tuple) values there.

While trying to produce a mini example I noticed that it only happens when I use the same id object:

import yaml

t = ("b", "c")
x = {(1, t):1, (2, t):2, }
print(yaml.dump(x))

So I thought copy.copy() would solve the problem, however for tuples it doesn't seem to work :( Can I create a new tuple with a different id?

Answer 1

The PyYAML dumper uses an ignore_aliases method to prevent primitive types from being "anchored" and "referenced" in this way. You can override that method to always ignore_aliases independent of any object passed in. And by default the yaml.Loader class is used in yaml.load¹:

t = ("b", "c")
x = {(1, t):1, (2, t):2, }

yaml.Dumper.ignore_aliases = lambda *args : True

yaml.dump(x, sys.stdout)

will get you:

? !!python/tuple
- 1
- !!python/tuple [b, c]
: 1
? !!python/tuple
- 2
- !!python/tuple [b, c]
: 2

That way you don't have to try your best and get tuples with the same hash to look different. You might want to provide the default_flow_style parameter on yaml.load to False or True to get different layouts of the output.

The reason you could not get this to work is that the representer matches the result of id() and that is the same for two tuples generated separately as long as the elements are the same.

---

¹ _{In ruamel.yaml, of which I am the author, which is an enhanced version of PyYAML, capable of handling YAML 1.2, you can do:}

    yaml = ruamel.yaml.YAML()
    yaml.representer.ignore_aliases = lambda *args: True

Answer 2

This method works for me on Python 2 and Python 3, and does not require monkeypatching:

import yaml

class NoAliasDumper(yaml.SafeDumper):
    def ignore_aliases(self, data):
        return True

t = ("b", "c")
x = {(1, t):1, (2, t):2, }
print(yaml.dump(x, Dumper=NoAliasDumper))

which yields

? - 1
  - [b, c]
: 1
? - 2
  - [b, c]
: 2