I’ve just started using YAML (through pyyaml) and I was wondering if there is any way to state that the value of a key is the key name itself or the parent key.
For example
---
foo: &FOO
bar: !.
baz: !..
foo2:
<<: *FOO
…
{‘foo’: {‘bar’: ‘bar’, ‘baz’: ’foo’}, ‘foo2’:{‘bar’:’bar’, ‘baz’:’foo2’}}
(notice the dot and double dot on bar and baz respectively - those are just placeholders for getting the key name and parent key name)
I've tried using add_constructor:
def key_construct(loader, node):
# return the key here
pass
yaml.add_constructor('!.', key_construct)
but Node, doesn't hold the key (or a reference to the parent) and I couldn't find the way to get them.
EDIT:
So, here is my real use case and a solution based on Anthon's response:
I have a logger configuration file (in yaml), and I wanted to reuse some definitions there:
handlers:
base: &base_handler
(): globals.TimedRotatingFileHandlerFactory
name: ../
when: midnight
backupCount: 14
level: DEBUG
formatter: generic
syslog:
class: logging.handlers.SysLogHandler
address: ['localhost', 514]
facility: local5
level: NOTSET
formatter: syslog
access:
<<: *base_handler
error:
<<: *base_handler
loggers:
base: &base_logger
handlers: [../, syslog]
qualname: ../
access:
<<: *base_logger
error:
<<: *base_logger
handlers: [../, syslog, email]
The solution, as Anthon suggested was to traverse the configuration dictionary after is was being processed:
def expand_yaml(val, parent=None, key=None, key1=None, key2=None):
if isinstance(val, str):
if val == './':
parent[key] = key1
elif val == '../':
parent[key] = key2
elif isinstance(val, dict):
for k, v in val.items():
expand_yaml(v, val, k, k, key1)
elif isinstance(val, list):
parent[key] = val[:] # support inheritance of the list (as in *base_logger)
for index, e in enumerate(parent[key]):
expand_yaml(e, parent[key], index, key1, key2)
return val
You don't have much context when you are constructing an element, so you are not going to find your key, and certainly not the parent key, to fill in the values, without digging in the call stack for the context (the loader knows about foo, bar and baz, but not in a way you can use to determine which is the corresponding key or parent_key).
What I suggest you do is create a special node that you return with the key_construct and then after the YAML load, walk the structure that your yaml.load() returned. Unless you have other ! objects, which make it more difficult to walk the resulting combination than a pure combination of sequences/lists and mappings/dicts ¹:
import ruamel.yaml as yaml
yaml_str = """\
foo: &FOO
bar: !.
baz: !..
foo2:
<<: *FOO
"""
class Expander:
def __init__(self, tag):
self.tag = tag
def expand(self, key, parent_key):
if self.tag == '!.':
return key
elif self.tag == '!..':
return parent_key
raise NotImplementedError
def __repr__(self):
return "E({})".format(self.tag)
def expand(d, key=None, parent_key=None):
if isinstance(d, list):
for elem in d:
expand(elem, key=key, parent_key=parent_key)
elif isinstance(d, dict):
for k in d:
v = d[k]
if isinstance(v, Expander):
d[k] = v.expand(k, parent_key)
expand(d[k], key, parent_key=k)
return d
def key_construct(loader, node):
return Expander(node.tag)
yaml.add_constructor('!.', key_construct)
yaml.add_constructor('!..', key_construct)
data = yaml.load(yaml_str)
print(data)
print(expand(data))
gives you:
{'foo': {'bar': E(!.), 'baz': E(!..)}, 'foo2': {'bar': E(!.), 'baz': E(!..)}}
{'foo': {'bar': 'bar', 'baz': 'foo'}, 'foo2': {'bar': 'bar', 'baz': 'foo2'}}
¹ This was done using ruamel.yaml of which I am the author. PyYAML, of which ruamel.yaml is a functional superset, should work the same.
Related
I have a number of objects and I'd like to "pool" them, i.e., put them into lists or sets such that
every object appears in at most one list, and
every object knows which list it's in.
In Python, I could do
# create objects
pool1 = [obj1, obj5, obj6]
pool2 = [obj3]
pool3 = [obj8, obj7]
obj1.pool = pool1
obj2.pool = None
obj3.pool = pool2
obj4.pool = None
obj5.pool = pool1
obj6.pool = pool1
obj7.pool = pool3
obj8.pool = pool3
This works, but has the disadvantage that the data structure can represent illegal states, e.g.,
pool1 = [obj1]
pool2 = []
obj1.pool = pool2
or
pool1 = [obj1]
pool2 = [obj1]
obj1.pool = pool1
Is there a more fitting data structure for this?
I don't think there is a more fitting data structure for this, as you need the association to work in two directions (from object to list, from list to object).
The best is probably to encapsulate this logic in a class and require that the caller uses only the provided methods to manipulate the data structure.
In Python that could look like this:
class Node:
def __init__(self, name):
self.name = name
self.pool = None
def __repr__(self):
return self.name
class Pools():
def __init__(self):
self._pools = {}
def assign(self, poolid, obj):
if poolid not in self._pools:
self._pools[poolid] = set()
if obj.pool is not None:
self._pools[obj.pool].discard(obj)
if poolid is not None:
self._pools[poolid].add(obj)
obj.pool = poolid
def unassign(self, obj):
self.assign(None, obj)
def content(self, poolid):
return list(self._pools[poolid])
# demo
a = Node("a")
b = Node("b")
c = Node("c")
pools = Pools()
pools.assign(0, a)
pools.assign(0, b)
pools.assign(5, c)
pools.assign(3, a)
pools.assign(3, c)
pools.unassign(b)
print(pools.content(0)) # []
print(pools.content(3)) # ['a', 'c']
print(pools.content(5)) # []
print(a.pool) # 3
print(b.pool) # None
print(c.pool) # 3
You could improve on this and make Pools a subclass of dict, but you get the idea.
I'm fairly new to Ruby and I've inherited some code which does a "deep merge" of some YAML. Here's the relevant part :-
class ::Hash
def deep_merge(second)
merger = proc { |key, v1, v2| Hash === v1 && Hash === v2 ? v1.merge(v2, &merger) : Array === v1 && Array === v2 ? v1 | v2 : [:undefined, nil, :nil].include?(v2) ? v1 : v2 }
self.merge(second.to_h, &merger)
end
end
which I found fairly unreadable TBH. It falls over when I pass it the following YAML :-
- {key: nginx.ingress.kubernetes.io/auth-type, value: basic}
- {key: nginx.ingress.kubernetes.io/auth-secret, value: basic-auth}
- {key: nginx.ingress.kubernetes.io/auth-realm, value: 'Authentication Required.'}
the "-" are all indented in the yaml input, but code formatting is messing with that here.
Here's a stripped down version of the YAML I'm trying to merge with (which also fails)
service:
container:
port: 3000
Any ideas?
OK I found the problem. I had forgotten to add a label to the YAML (annotations:) and as soon as I put that on, it started working again. Should I delete the question?
Is overriding from_yaml enough to register a tag from a class or is it necessary to use yaml.add_constructor(Class.yaml_tag, Class.from_yaml)? If I don't use te add_constructor method, my YAML tags are not recognized. Example of what I have:
import yaml
class Something(yaml.YAMLObject):
yaml_tag = u'!Something'
#classmethod
def from_yaml(cls,loader,node):
# Set attributes to None if not in file
values = loader.construct_mapping(node, deep=True)
attr = ['attr1','attr2']
result = {}
for val in attr:
try:
result[val] = values[val]
except KeyError:
result[val] = None
return cls(**result)
Is this enough for it to work? I'm confused with the use of from_yaml vs any other constructor you would register using the method I mentioned above. I suppose there's something fundamental I'm missing, since they say:
Subclassing YAMLObject is an easy way to define tags, constructors,
and representers for your classes. You only need to override the
yaml_tag attribute. If you want to define your custom constructor and
representer, redefine the from_yaml and to_yaml method
correspondingly.
There is indeed no need to register explicitly:
import yaml
class Something(yaml.YAMLObject):
yaml_tag = u'!Something'
def __init__(self, *args, **kw):
print('some_init', args, kw)
#classmethod
def from_yaml(cls,loader,node):
# Set attributes to None if not in file
values = loader.construct_mapping(node, deep=True)
attr = ['attr1','attr2']
result = {}
for val in attr:
try:
result[val] = values[val]
except KeyError:
result[val] = None
return cls(**result)
yaml_str = """\
test: !Something
attr1: 1
attr2: 2
"""
d = yaml.load(yaml_str)
which gives:
some_init () {'attr1': 1, 'attr2': 2}
But there is no need at all to use PyYAML's load() which is
documented to be unsafe. You can just use safe_load if you set the yaml_loader class attribute:
import yaml
class Something(yaml.YAMLObject):
yaml_tag = u'!Something'
yaml_loader = yaml.SafeLoader
def __init__(self, *args, **kw):
print('some_init', args, kw)
#classmethod
def from_yaml(cls,loader,node):
# Set attributes to None if not in file
values = loader.construct_mapping(node, deep=True)
attr = ['attr1','attr2']
result = {}
for val in attr:
try:
result[val] = values[val]
except KeyError:
result[val] = None
return cls(**result)
yaml_str = """\
test: !Something
attr1: 1
attr2: 2
"""
d = yaml.safe_load(yaml_str)
as this gives the same:
some_init () {'attr1': 1, 'attr2': 2}
(done both with Python 3.6 and Python 2.7)
The registering is done in the __init__() of the metaclass of yaml.YAMLObject:
class YAMLObjectMetaclass(type):
"""
The metaclass for YAMLObject.
"""
def __init__(cls, name, bases, kwds):
super(YAMLObjectMetaclass, cls).__init__(name, bases, kwds)
if 'yaml_tag' in kwds and kwds['yaml_tag'] is not None:
cls.yaml_loader.add_constructor(cls.yaml_tag, cls.from_yaml)
cls.yaml_dumper.add_representer(cls, cls.to_yaml)
So maybe you are somehow interfering with that initialisation in your full class definition. Try to start with a minimal implementation as I did, and add the functionality on your class that you need until things break.
I'm facing a problem that I couldn't find a working solution yet.
I have my YAML config file for the environment, let's call it development.yml.
This file is used to create the hash that should be updated:
data = YAML.load_file(File.join(Rails.root, 'config', 'environments', 'development.yml'))
What I'm trying to accomplish is something along these lines. Let's suppose we have an element of the sort
data['server']['test']['user']
data['server']['test']['password']
What I want to have is:
data['server']['test']['user'] = #{Server.Test.User}
data['server']['test']['password'] = #{Server.Test.Password}
The idea is to create a placeholder for each value that is the key mapping for that value dynamically, going until the last level of the hash and replacing the value with the mapping to this value, concatenating the keys.
Sorry, it doesn't solve my problem. The location data['server']['test']['user'] will be built dynamically, via a loop that will go through a nested Hash. The only way I found to do it was to append to the string the key for the current iteration of the Hash. At the end, I have a string like "data['server']['test']['name']", which I was thinking on converting to a variable data['server']['test']['name'] and then assigning to this variable the value #{Server.Test.Name}. Reading my question I'm not sure if this is clear, I hope this helps to clarify it.
Input sample:
api: 'active'
server:
test:
user: 'test'
password: 'passwordfortest'
prod:
user: 'nottest'
password: 'morecomplicatedthantest'
In this case, the final result should be to update this yml file in this way:
api: #{Api}
server:
test:
user: #{Server.Test.User}
password: #{Server.Test.Password}
prod:
user: #{Server.Prod.User}
password: #{Server.Prod.Password}
It sounds silly, but I couldn't figure out a way to do it.
I am posting another answer now since I realize what the question is all about.
Use Iteraptor gem:
require 'iteraptor'
require 'yaml'
# or load from file
yaml = <<-YAML.squish
api: 'active'
server:
test:
user: 'test'
password: 'passwordfortest'
prod:
user: 'nottest'
password: 'morecomplicatedthantest'
YAML
mapped =
yaml.iteraptor.map(full_parent: true) do |parent, (k, _)|
v = parent.map(&:capitalize).join('.')
[k, "\#{#{v}}"]
end
puts YAML.dump(mapped)
#⇒ ---
# api: "#{Api}"
# server:
# test:
# user: "#{Server.Test.User}"
# password: "#{Server.Test.Password}"
# prod:
# user: "#{Server.Prod.User}"
# password: "#{Server.Prod.Password}"
puts YAML.dump(mapped).delete('"')
#⇒ ---
# api: #{Api}
# server:
# test:
# user: #{Server.Test.User}
# password: #{Server.Test.Password}
# prod:
# user: #{Server.Prod.User}
# password: #{Server.Prod.Password}
Use String#%:
input = %|
data['server']['host']['name'] = %{server_host}
data['server']['host']['user'] = %{server_host_user}
data['server']['host']['password'] = %{server_host_password}
|
puts (
input % {server_host: "Foo",
server_host_user: "Bar",
server_host_password: "Baz"})
#⇒ data['server']['host']['name'] = Foo
# data['server']['host']['user'] = Bar
# data['server']['host']['password'] = Baz
You can not add key-value pair to a string.
data['server']['host'] # => which results in a string
Option 1:
You can either save Server.Host as host name in the hash
data['server']['host']['name'] = "#{Server.Host}"
data['server']['host']['user'] = "#{Server.Host.User}"
data['server']['host']['password'] = "#{Server.Host.Password}"
Option 2:
You can construct the hash in a single step with Host as key.
data['server']['host'] = { "#{Server.Host}" => {
'user' => "#{Server.Host.User}",
'password' => "#{Server.Host.Password}"
}
}
I have an ABC BaseAbstract class with several getter/setter properties defined.
I want to require that the value to be set is an int and from 0 - 15.
#luminance.setter
#abstractproperty
#ValidateProperty(Exception, types=(int,), valid=lambda x: True if 0 <= x <= 15 else False)
def luminance(self, value):
"""
Set a value that indicate the level of light emitted from the block
:param value: (int): 0 (darkest) - 15 (brightest)
:return:
"""
pass
Can someone help me figure out what my ValidateProperty class/method should look like. I started with a class and called the accepts method but this is causing an error:
function object has no attribute 'func_code'
current source:
class ValidateProperty(object):
#staticmethod
def accepts(exception, *types, **kwargs):
def check_accepts(f, **kwargs):
assert len(types) == f.func_code.co_argcount
def new_f(*args, **kwds):
for i, v in enumerate(args):
if f.func_code.co_varnames[i] in types and\
not isinstance(v, types[f.func_code.co_varnames[i]]):
arg = f.func_code.co_varnames[i]
exp = types[f.func_code.co_varnames[i]]
raise exception("arg '{arg}'={r} does not match {exp}".format(arg=arg,
r=v,
exp=exp))
# del exp (unreachable)
for k,v in kwds.__iter__():
if k in types and not isinstance(v, types[k]):
raise exception("arg '{arg}'={r} does not match {exp}".format(arg=k,
r=v,
exp=types[k]))
return f(*args, **kwds)
new_f.func_name = f.func_name
return new_f
return check_accepts
One of us is confused about how decorators, descriptors (e.g. properties), and abstracts work -- I hope it's not me. ;)
Here is a rough working example:
from abc import ABCMeta, abstractproperty
class ValidateProperty:
def __init__(inst, exception, arg_type, valid):
# called on the #ValidateProperty(...) line
#
# save the exception to raise, the expected argument type, and
# the validator code for later use
inst.exception = exception
inst.arg_type = arg_type
inst.validator = valid
def __call__(inst, func):
# called after the def has finished, but before it is stored
#
# func is the def'd function, save it for later to be called
# after validating the argument
def check_accepts(self, value):
if not inst.validator(value):
raise inst.exception('value %s is not valid' % value)
func(self, value)
return check_accepts
class AbstractTestClass(metaclass=ABCMeta):
#abstractproperty
def luminance(self):
# abstract property
return
#luminance.setter
#ValidateProperty(Exception, int, lambda x: 0 <= x <= 15)
def luminance(self, value):
# abstract property with validator
return
class TestClass(AbstractTestClass):
# concrete class
val = 7
#property
def luminance(self):
# concrete property
return self.val
#luminance.setter
def luminance(self, value):
# concrete property setter
# call base class first to activate the validator
AbstractTestClass.__dict__['luminance'].__set__(self, value)
self.val = value
tc = TestClass()
print(tc.luminance)
tc.luminance = 10
print(tc.luminance)
tc.luminance = 25
print(tc.luminance)
Which results in:
7
10
Traceback (most recent call last):
File "abstract.py", line 47, in <module>
tc.luminance = 25
File "abstract.py", line 40, in luminance
AbstractTestClass.__dict__['luminance'].__set__(self, value)
File "abstract.py", line 14, in check_accepts
raise inst.exception('value %s is not valid' % value)
Exception: value 25 is not valid
A few points to think about:
The ValidateProperty is much simpler because a property setter only takes two parameters: self and the new_value
When using a class for a decorator, and the decorator takes arguments, then you will need __init__ to save the parameters, and __call__ to actually deal with the defd function
Calling a base class property setter is ugly, but you could hide that in a helper function
you might want to use a custom metaclass to ensure the validation code is run (which would also avoid the ugly base-class property call)
I suggested a metaclass above to eliminate the need for a direct call to the base class's abstractproperty, and here is an example of such:
from abc import ABCMeta, abstractproperty
class AbstractTestClassMeta(ABCMeta):
def __new__(metacls, cls, bases, clsdict):
# create new class
new_cls = super().__new__(metacls, cls, bases, clsdict)
# collect all base class dictionaries
base_dicts = [b.__dict__ for b in bases]
if not base_dicts:
return new_cls
# iterate through clsdict looking for properties
for name, obj in clsdict.items():
if not isinstance(obj, (property)):
continue
prop_set = getattr(obj, 'fset')
# found one, now look in bases for validation code
validators = []
for d in base_dicts:
b_obj = d.get(name)
if (
b_obj is not None and
isinstance(b_obj.fset, ValidateProperty)
):
validators.append(b_obj.fset)
if validators:
def check_validators(self, new_val):
for func in validators:
func(new_val)
prop_set(self, new_val)
new_prop = obj.setter(check_validators)
setattr(new_cls, name, new_prop)
return new_cls
This subclasses ABCMeta, and has ABCMeta do all of its work first, then does some additional processing. Namely:
go through the created class and look for properties
check the base classes to see if they have a matching abstractproperty
check the abstractproperty's fset code to see if it is an instance of ValidateProperty
if so, save it in a list of validators
if the list of validators is not empty
make a wrapper that will call each validator before calling the actual property's fset code
replace the found property with a new one that uses the wrapper as the setter code
ValidateProperty is a little different as well:
class ValidateProperty:
def __init__(self, exception, arg_type):
# called on the #ValidateProperty(...) line
#
# save the exception to raise and the expected argument type
self.exception = exception
self.arg_type = arg_type
self.validator = None
def __call__(self, func_or_value):
# on the first call, func_or_value is the function to use
# as the validator
if self.validator is None:
self.validator = func_or_value
return self
# every subsequent call will be to do the validation
if (
not isinstance(func_or_value, self.arg_type) or
not self.validator(None, func_or_value)
):
raise self.exception(
'%r is either not a type of %r or is outside '
'argument range' %
(func_or_value, type(func_or_value))
)
The base AbstractTestClass now uses the new AbstractTestClassMeta, and has the validator code directly in the abstractproperty:
class AbstractTestClass(metaclass=AbstractTestClassMeta):
#abstractproperty
def luminance(self):
# abstract property
pass
#luminance.setter
#ValidateProperty(Exception, int)
def luminance(self, value):
# abstract property validator
return 0 <= value <= 15
The final class is the same:
class TestClass(AbstractTestClass):
# concrete class
val = 7
#property
def luminance(self):
# concrete property
return self.val
#luminance.setter
def luminance(self, value):
# concrete property setter
# call base class first to activate the validator
# AbstractTestClass.__dict__['luminance'].__set__(self, value)
self.val = value