A handwritten Python 3.7+ implemenation of a parser for the KDL Document Language, fully compliant with KDL 2.0.0.

KDL is, as the name suggests, a document language, filling approximately the same niche as JSON/YAML/XML/etc to be a simple but powerful language for config files. It combines the best of several of these languages, while avoiding their pitfalls: more general than JSON and more powerful than XML, while avoiding the verbosity of XML or the explosive complexity of YAML.

kdl-py, in particular, is focused on ease-of-use, supporting things like date/times, ip addresses, urls, uuids, regexes, and binary data directly in your KDL document (via powerful but simple tagged values), and parsing them into native Python types automatically, or doing the reverse and letting you build KDL document trees with these values directly, and automatically and safely serializing them into KDL text for later parsing!

You can customize parsing and serialization further for your application very easily, by providing node or value "converters" to turn plain KDL values or nodes into application-specific types, and then turning them back into KDL text automatically merely by adding a .to_kdl() method to your classes.

pipx install kdl-py

When installed, a kdlreformat command-line program is also made available, which can canonicalize a KDL document. See below for options.

The kdl.parse(str, parseConfig|None) function parses, you guessed it, a string of KDL into a KDL document object:

>>> import kdl
>>> doc = kdl.parse('''
... node_name arg {
...     child_node foo=1 bar=#true
... }
... ''')
>>>
>>> doc
Document(
    nodes=[
        Node(
            name='node',
            args=['arg'],
            nodes=[
                Node(
                    name='child',
                    props=OrderedDict([
                        ('foo', 1.0),
                        ('bar', True)
                    ])
                )
            ]
        )
    ]
)

You can also create a kdl.Parser() object and call its .parse() method; Parser objects can set up parsing and printing options that'll apply by default. See below for how to configure parsing options.

Either way, you'll get back a kdl.Document object, which is fully mutable. By default, untagged KDL values are represented with native Python objects.

>>> doc.nodes[0].nodes[0].props["foo"] = 2
>>>
>>> print(doc)
node_name arg {
        child_node foo=2 bar=#true
}

Stringifying a kdl.Document object will produce a valid KDL document back. You can also call doc.print(printConfig|None) to customize the printing with a PrintConfig object, described below. See below for how to configure printing options.

kdl-py allows a number of native Python objects to be used directly in KDL documents by default, and allows you to customize your own objects for use.

kdl-py automatically recognizes and correctly serializes the following objects:

• bool: as untagged true or false
• None: as untagged null
• int, float: as untagged decimal number
• str: as untagged string
• bytes: as (base64)-tagged string
• decimal.Decimal: as (decimal)-tagged string
• datetime, date, and time: as (date-time), (date), or (time)-tagged strings
• ipaddress.IPv4Address and ipaddress.IPv6Address: as (ipv4) or (ipv6)-tagged strings
• urllib.parse.ParseResult (the result of calling urllib.parse.urlparse()): as (url)-tagged string
• uuid.UUID: as (uuid)-tagged string
• re.Pattern (the result of calling re.compile()): as (regex)-tagged raw string

All of the tags used above are reserved and predefined by the KDL specification.

In addition, any value with a .to_kdl() method can be used in a kdl-py document. The method will be called when the document is stringified, and must return one of the kdl-py types, or any of the native types defined above.

(For parsing KDL into these native types, or your own types, see the ParseConfig section, below.)

Parsing can be controlled via a kdl.ParseConfig object, which can be provided in three ways. In order of importance:

1. Passing a ParseConfig object to kdl.parse(str, ParseConfig|None) or parser.parse(str, ParseConfig|None) (if you've constructed a kdl.Parser).
2. Creating a kdl.Parser(parseConfig|None, printConfig|None), which automatically applies it to its .parse() method if not overriden.
3. Fiddling with the kdl.parsing.defaults object, which is used if nothing else provides a config.

A ParseConfig object has the following properties:

• nativeUntaggedValues: bool = True

  Controls whether the parser produces native Python objects (str, int, float, bool, None) when parsing untagged values (those without a (foo) prefix), or always produces kdl-py objects (such as kdl.String, kdl.Decimal, etc).

• nativeTaggedValues: bool = True

  Controls whether the parser produces native Python objects when parsing tagged values, for some of KDL's predefined tags:

  • i8, i16, i32, i64, u8, u16, u32, u64 on numbers: Checks that the value is in the specified range, then converts it to an int. (It will serialize back out as an ordinary untagged number.)
  • f32, f64 on numbers: Converts it to a float. (It will serialize back out as an ordinary untagged number.)
  • decimal64, decimal128 on numbers, and decimal on strings: Converts it to a decimal.Decimal object. (Always reserializes to a (decimal)-tagged string.)
  • date-time, date, time on strings: Converts to a datetime, time, or date object.
  • ipv4, ipv6 on strings: Converts it to an ipaddress.IPv4Address or ipaddress.IPv6Address object.
  • url on strings: Converts it to a urllib.parse.ParseResult tuple.
  • uuid on strings: Converts it to a uuid.UUID object.
  • regex on strings: Converts it to a re.Pattern object. (It will serialize back out as a raw string.)
  • base64 on strings: Converts it to a bytes object.

• valueConverters: Dict[ValueKey, Callable] = {}

  A dictionary of ValueKey->converter functions, letting you parse values (like (date)"2021-01-01") into whatever types you'd like.

  Whenever the parser encounters a value, it will attempt to find an entry in this dict whose ValueKey matches the value. If it succeeds, it will call the associated converter function with two arguments: the fully-constructed kdl-py object, and a ParseFragment object giving you access to the precise characters parsed from the document (for the value; the tag, if any, won't be included, as you can get it from the object itself). Whatever you return will be inserted into the document instead of the originally-parsed value.

  You can produce KDL values (such as parsing (hex)"0x12.e5" into a kdl.Decimal, since KDL doesn't support fractional hex values), or produce any other Python type. If you return a non-KDL type, you probably want to ensure it has a .to_kdl() method (or is one of the supported built-in types), so it can be serialized back into a KDL document.

  Note that only one conversion can happen to a given value. Your converters are checked in the dictionary's iteration order, then if none of them succeeded it will attempt to run the nativeUntaggedValues or nativeTaggedValues behaviors, if they're turned on. If a converter returns NotImplemented, it will continue looking for a matching converter.

• nodeConverters: Dict[NodeKey, Callable] = {}

  Similar to valueConverters, except it uses NodeKeys, and it converts kdl.Nodes instead.

  There is no native conversion of nodes; if none of your converters successfully run, the node will be inserted as-is.

kdl.ParseFragment is passed to your custom converters, specified in kdl.ParseConfig.tags, giving you direct access to the input characters before any additional processing was done on them. This is useful, for example, to handle numeric types that might have lost precision in the normal parse.

It exposes a .fragment property, containing the raw text of the value (after the tag, if any).

It also exposes a .error(str) method, which takes a custom error message and returns a kdl.ParseError with the ParseFragment's location already built in, ready for you to raise. This should be used if your conversion fails for any reason, so your errors look the same as native parse errors.

Like parsing, printing a kdl-py Document back to a KDL string can be controlled by a kdl.PrintConfig object, which can be provided in three ways. In order of importance:

1. Passing a PrintConfig object to doc.print(PrintConfig|None).
2. Setting doc.printConfig to a PrintConfig. (This is done automatically for any documents produced by a Parser, if you pass the printConfig option to the constructor.)
3. Fiddling with the kdl.printing.defaults object, which is used if nothing else provides a config.

A PrintConfig object has the following properties:

• indent: str = "\t"

  The string used for each indent level. Defaults to tabs, but can be set to a sequence of spaces if desired (or anything else).

• semicolons: bool = False

  Whether or not nodes are ended with semicolons. (The printer always ends nodes with a newline anyway, so this is purely a stylistic choice.)

• printNullArgs: bool = True

  When False, automatically skips over any "null"/None arguments. This will corrupt documents that use the "null" keyword intentionally, but can be useful if you'd prefer to use a None value as a signal that the argument has been removed.

• printNullProps: bool = True

  Identical to printNullArgs, but applies to properties rather than arguments.

• respectStringType: bool = True

  When True, the printer will output strings as the same type they were in the input, either raw (r#"foo"#) or normal ("foo"). When False, the printer always outputs normal strings.

  Note that this only has an effect on kdl.String and kdl.RawString objects; if the document contains Python str objects, they will always output as normal strings.

• respectRadix: bool = True

  Similar to respectStringType, when True the printer will output numbers as the radix they were in the input, like 0x1b for hex numbers. When False, the printer always outputs decimal numbers.

  Again, this only has an effect on kdl-py objects; native Python numbers are printed as normal for Python.

• exponent: str = "e"

  What character to use for the exponent part of decimal numbers, when printed with scientific notation. Should only be set to "e" or "E".

  Like the previous options, this only has an effect on kdl-py objects; native Python numbers are printed as normal for Python.

• kdl.parse(str, config: kdl.ParseConfig|None) -> kdl.Document

• kdl.Parser(parseConfig: kdl.ParseConfig|None, printConfig: kdl.PrintConfig|None)

  • parser.parse(str, config: kdl.ParseConfig|None) -> kdl.Document
  • parser.print(config: kdl.PrintConfig|None) -> str

• kdl.Document(nodes: list[kdl.Node]?, printConfig: kdl.PrintConfig|None)

  • doc.print(PrintConfig|None) -> str
  • doc[NodeKey] -> Node returns the first child node matching the NodeKey. Raises a KeyError if nothing matches the NodeKey, similar to a dict.
  • doc.get(NodeKey, default: T = None) -> kdl.Node | T returns the first child node matching the NodeKey. Returns the default value if nothing matches.
  • doc.getAll(NodeKey) -> Iterable[kdl.Node] returns all child nodes matching the NodeKey

• kdl.Node(name: str, tag: str|None, args: list[Any]?, props: dict[str, Any]?, nodes: list[kdl.Node]?)

  • node[NodeKey] -> Node returns the first child node matching the NodeKey. Raises a KeyError if nothing matches the NodeKey, similar to a dict.
  • node.get(NodeKey, default: T = None) -> kdl.Node | T returns the first child node matching the NodeKey. Returns the default value if nothing matches.
  • node.getAll(NodeKey) -> Iterable[kdl.Node] returns all child nodes matching the NodeKey
  • node.getProps(ValueKey) -> Iterable[tuple[str, Any]] returns an iterator of (name, value) pairs for the properties whose value matches the ValueKey (Note: for this purpose, non-KDL values, such as int, have no tag, and can only meaningfully be tested via a TypeKey.)
  • node.getArgs(ValueKey) -> Iterable[Any] returns an iterator of the arguments that match the ValueKey (Same disclaimer as .getProps().)
  • node.matchesKey(NodeKey) -> bool returns whether the node matches the NodeKey

• kdl.Value ‡

  • val.matchesKey(ValueKey) -> bool returns whether the value matches the ValueKey

• kdl.Binary(value: int, tag: str|None)

• kdl.Octal(value: int, tag: str|None)

• kdl.Decimal(mantissa: int|float, exponent: int|None, tag: str|None)

  • dec.value: readonly, mantissa * (10**exponent)

• kdl.Hex(value: int, tag: str|None)

• kdl.Bool(value: bool, tag: str|None)

• kdl.Null(tag: str|None)

  • null.value: readonly, always None

• kdl.RawString(value: str, tag: str|None)

• kdl.String(value: str, tag: str|None)

• kdl.ExactValue(chars: str, tag: str|None) †

• kdl.Numberish, kdl.Stringish ‡

• kdl.ParseConfig(...) see above for options

  • kdl.parsing.defaults: default ParseConfig

• kdl.PrintConfig(...) see above for options

  • kdl.printing.defaults: default PrintConfig

• kdl.ParseError: thrown for all parsing errors

  • error.msg: str: hopefully informative
  • error.line: int: 1-indexed
  • error.col: int: 1-indexed

• kdl.ParseFragment: passed to converter functions

  • pf.fragment: slice from the source string
  • pf.error(msg: str) returns a kdl.ParseError with error location set properly already

• kdl.nodeMatchesKey(val: Any, key: kdl.NodeKey) -> bool

• kdl.valueMatchesKey(val: Any, key: kdl.ValueKey) -> bool

• kdl.tagMatchesKey(val: str|None, key: kdl.TagKey) -> bool

• kdl.nameMatchesKey(val: str|None, key: kdl.NameKey) -> bool

• kdl.typeMatchesKey(val: str|None, key: kdl.TypeKey) -> bool

  • Functions implementing the tag/name/type matching used by the node.matchesKey() and value.matchesKey() methods, in case you want to use the same filtering yourself.

† Not produced by the parser. Can be returned by a user's .to_kdl() method if they want to produce a value precisely in a particular syntax, in a way that the built-in kdl-py classes don't.

‡ Not produced by the parser. These are abstract base classes to help in type testing: Value matches all eight value classes, Numberish matches all four numeric value classes, and Stringish matches both string value classes.

A few type aliases also exist, used by the module in a few places and potentially useful for your code:

• kdl.KDLAny: a Document, Node, or any of the Value subtypes
• kdl.KDLValue: any of the Value subtypes
• kdl.KDLishValue: a KDLValue or one of the supported Python native types see "Inserting Native Types"
• kdl.ValueKey, kdl.NodeKey, kdl.TagKey, kdl.NameKey, kdl.TypeKey: the types for ValueKeys and NodeKeys (and their individual pieces)

These aliases only exist when typing.TYPE_CHECKING is true, so they're only useful for writing types; they won't be visible at runtime.

A few data structures and functions take a NodeKey to match against a node, based on its name and/or tag.

A NodeKey is either a NameKey or a (TagKey, NameKey) tuple, matching against the node's name and/or tag.

NameKeys and TagKeys can be:

• None: When matched against a tag, only succeeds against a None tag (aka a tagless node like foo). When matched against a name, automatically succeeds, since nodes are guaranteed to have a name.
• ...: Always succeeds. Use this when, say, you want to match against a particular tag, regardless of the nodename, like ("my-tag", ...).
• a str: Succeeds if the tag/name is that exact string.
• a re.Pattern (a regex, such as re.compile(r".*foo"): Succeeds if the regex matches the tag/name. Note that this uses .match() semantics, automatically anchoring against the start of the string; prepend your regex with .*? if you want it to match anywhere.
• a function, taking a str | None and returning a bool: Succeeds if the function, when called with the tag/name, returns True.

So, for example, doc.getAll("foo") would return all nodes whose name is "foo", regardless of tag. doc.getAll(("my-tag", ...)) would return all nodes whose tag is "my-tag", regardless of the node name. doc.getAll(re.compile(r"-")) would return all nodes whose name starts with "-". Etc.

Similarly to NodeKey, a few things take a ValueKey to match against values, based on its tag and/or type.

A ValueKey is either a TagKey or a (TagKey, TypeKey) tuple, matching against the value's tag and/or type.

TagKey works identically to how it appears in NodeKey. TypeKey is either ..., which matches any type, or the same sort of argument you'd pass as the second argument to isinstance().

For example, a "foo" key would match any values with the tag "foo", like (foo)1 or (foo)"bar". A (..., kdl.Numberish) will match any value that's a "number" type: a Hex, Octal, Binary, or Decimal, regardless of its tag. Etc.

The kdlreformat command-line program is installed by default when you install this module from pypi. It can also be run manually from the kdlreformat.py file at the root of this repository (or from the kdl.cli.cli() function)

usage: kdlreformat [-h] [--indent INDENT] [--semicolons] [--radix]
                   [--no-radix] [--raw-strings] [--no-raw-strings]
                   [--exponent EXPONENT]
                   [infile] [outfile]

KDL parser/printer, letting you easily reformat KDL files into a canonical
representation.

positional arguments:
  infile
  outfile

optional arguments:
  -h, --help           show this help message and exit
  --indent INDENT      How many spaces for each level of indent. -1 indicates
                       to indent with tabs.
  --semicolons         Whether to end nodes with semicolons or not.
  --radix              Output numeric values in the radix used by the input.
                       (0x1a outputs as 0x1a)
  --no-radix           Convert all numeric arguments to decimal. (0x1a outputs
                       as 26)
  --raw-strings        Output string values in the string type used by the
                       input.
  --no-raw-strings     Convert all string arguments into plain strings.
  --exponent EXPONENT  What character to use ('e' or 'E') for indicating
                       exponents on scinot numbers.