A generic graph object for your own needs to implement syntax trees and similar constructs. Provides you with a name/value/children type of graph object, where your value can be any object type you wish. The project encapsulates the C# implementation of Hyperlambda, implying its lambda node structure. To use package go to https://polterguy.github.io
magic.node - Nodes, the basic structure in Hyperlambda
magic.node is a simple name/value/children graph object, in addition to a "Hyperlambda" parser, allowing you to
create a textual string representations of graph objects easily transformed to its relational graph object syntax
and vice versa. This allows you to easily declaratively create execution trees using a format similar to YAML, for
then to access each individual node, its value, name and children from your C#, CLR code, or Hyperlambda.
For the record, Hyperlambda is much easier to understand than YAML.
Hyperlambda is perfect for creating a highly humanly readable relational configuration format, or smaller
DSL engines, especially when combined with magic.signals and magic.lambda. Below is a small
example of Hyperlambda to give you an idea of how it looks like.
3 spaces (SP) opens up the children collection of a node, and allows you to create children associated with
some other node. In the example above, the [child1] and the [child2] nodes, have [foo] as their
parent. A colon : separates the name and the value of the node - The name is to the left of the colon, and
the value to the right.
You can optionally supply a type between a node's name and its value, which you can see above where we add
the :int: parts between one of our [foo] nodes' name and value. If you don't explicitly declare a type
string will be assumed.
Parsing Hyperlambda from C#
To traverse the nodes later in for instance C#, you could do something such as the following.
var root = var result = HyperlambdaParser.Parse(hyperlambda);
foreach (var idxChild in root.Children)
{
var name = idxChild.Name;
var value = idxChild.Value;
/* ... etc ... */
}
Notice - When you parse Hyperlambda a "root node" is explicitly created wrapping all your nodes, and
this is the node that's returned to you after parsing. All nodes you declare in your Hyperlambda will be
returned to you as children of this root node.
Hyperlambda types
Although the node structure itself can hold any value type you need inside of its Value property,
Hyperlambda only supports serialising the following types by default.
string = System.String
short = System.Int16
ushort = System.UInt16
int = System.Int32
uint = System.UInt32
long = System.Int64
ulong = System.UInt64
decimal = System.Decimal
double = System.Double
single = System.Float
float = System.Float - Alias for above
bool = System.Boolean
date = System.DateTime - Always interpreted and serialized as UTC time!
time = System.TimeSpan
guid = System.Guid
char = System.Char
byte = System.Byte
sbyte = System.SByte
x = magic.node.extensions.Expression
node = magic.node.Node
The type declaration should be declared in your Hyperlambda in between the name and its value, separated by colon (:).
The default type if ommitted is string and strings do not need quotes or double quotes for this reason. An example
of declaring a couple of types associated with a node's value can be found below.
The type system is extendible, and you can easily create support for serializing your own types, by using
the Converter.AddConverter method, that can be found in the magic.node.extensions namespace.
Below is an example of how to extend the typing system, to allow for serializing and de-serializing instances
of a Foo class into Hyperlambda.
/*
* Class you want to serialize into Hyperlambda.
*/
class Foo
{
public int Value1 { get; set; }
public decimal Value2 { get; set; }
}
/*
* Adding our converter functions, and associating them
* with a type, and a Hyperlambda type name.
*/
Converter.AddConverter(
typeof(Foo),
"foo",
(obj) => {
var foo = obj as Foo;
return ("foo", $"{foo.Value1}-{foo.Value2}");
}, (obj) => {
var str = (obj as string).Split('-');
return new Foo
{
Value1 = int.Parse(str[0]),
Value2 = decimal.Parse(str[1]),
};
});
The above will allow you to serialize instances of Foo into your Hyperlambda, and
de-serialize these instances once needed. An example of adding a Foo instance into
your Hyperlambda can be found below.
.foo:foo:5-7
Later you can retrieve your Foo instances in your slots, using something
resembling the following, and all parsing and conversion will be automatically
taken care of.
var foo = node.Get<Foo>();
String literals in Hyperlambda
Hyperlambda also support strings the same way C# supports string, using any of the following string representations.
// Single quotes
.foo:'howdy world this is a string'
// Double quotes
.foo:"Howdy world, another string"
// Multiline strings
.foo:@"Notice how the new line doesn't end the string
here!"
Escape characters are supported for both single quote and double quote strings the same way they
are supported in C#, allowing you to use e.g. \r\n etc.
Lambda expressions
Lambda expressions are kind of like XPath expressions, except they will references nodes
in your Node graph object instead of XML nodes. Below is an example to give you an idea.
.foo:hello world
get-value:x:@.foo
// After invocation of the above **[get-value]**, its value will be "hello world".
Most slots in Magic can accept expressions to reference nodes, values of nodes, and children of
nodes somehow. This allows you to modify the lambda graph object, as it is currently being executed,
and hence allows you to modify "anything" from "anywhere". This resembles XPath expressions
from XML.
Iterators in lambda expressions
An expression is constructed from one or more "iterator". This makes an expression
become "dynamically chained Linq statements", where each iterator reacts upon
the results of its previous iterator. Each iterator takes as input an IEnumerable,
and returns as its result another IEnumerable, where the content of the iterator
somehow changes its given input, according to whatever the particular iterator's
implementation does. This approach just so happens to be perfect for retrieving
sub-sections of graph objects.
Each iterator ends with a "/" or a CR/LF
sequence, and before its end, its value defines what it does. For instance the above iterator in
the [get-value] invocation, starts out with a "@". This implies that the iterator will find the
first node having a name of whatever follows its "@". For the above this implies looking for the first
node who's name is ".foo". To see a slightly more advanced example, imagine the following.
It might help to transform the above expression into humanly readable language. Its
English equivalent hence becomes as follows.
Find the node with the name of '.data', then retrieve its children, and filter away everything not having a name of 'item2'
Of course, the result of the above becomes "thomas".
Below is a list of all iterators that exists in magic. Substitute "xxx" with a string,
"n" with a number, and "x" with an expression.
* Retrieves all children of its previous result
# Retrieves the value of its previous result as a node by reference
- Retrieves its previous result set's "younger sibling" (previous node)
+ Retrieves its previous result set's "elder sibling" (next node)
. Retrieves its previous result set's parent node(s)
^xxx Retrieves the first ancestor node with the specified "xxx" name. Similar to @ iterator but does not traverse siblings, only direct ancestors up in hierarchy
.. Retrieves the root node
** Retrieves its previous result set's descendant, with a "breadth first" algorithm
-- Retrieves all ancestors and older siblings upwards in object
!xxx Traverses all descendants except those matching specified "xxx" name and returns
{x} Extrapolated expression that will be evaluated assuming it yields one result, replacing itself with the value of whatever node it points to
=xxx Retrieves the node with the "xxx" value, converting to string if necessary
[n,n] Retrieves a subset of its previous result set, implying "from, to" meaning [n1,n2>
@xxx Returns the first node "before" in its hierarchy that matches the given "xxx" in its name
n (any number) Returns the n'th child of its previous result set
[x|y] Pipe separated list of names returning all nodes having a name of either "x" or "y"
Notice, you can escape iterators by using backslash "\". This allows you to look for nodes who's names
are for instance "3", without using the n'th child iterator, which would defeat the purpose. In addition,
you can quote iterators by using double quotes ", to allow for having iterators with values that are normally
not legal within an iterator, such as /, etc. If you quote an iterator you have to quote the entire expression.
Below is an example of a slightly more advanced expression.
.foo
howdy:wo/rld
jo:nothing
howdy:earth
.dyn:.foo
for-each:x:@"./*/{@.dyn}/*/""=wo/rld"""
set-value:x:@.dp/#
:thomas was here
After evaluating the above Hyperlambda, the value of all nodes having "wo/rld" as their value
inside of [.foo] will be updated to become "thomas was here". Obviously, the above expression
is a ridiculous complex example, that you will probably never encounter in your own code. However,
for reference purposes, let's break it down into its individual parts.
Get parent node
Get all children
Filter away everything not having the name of the value of {@.dyn}, which resolves to the equivalent of :x:@.dyn, being an expression, who's result becomes ".foo".
Get its children
Find all nodes who's value is "wo/rld".
98% of your expressions will have 1-3 iterators, no complex escaping, and no parameters.
And in fact, there are thousands of lines of Hyperlambda code in Magic's middleware, and
98% of these expressions are as simple as follows.
Give me the value of any [foo1] nodes, inside of the first [.arguments] node you can find upwards in the hierarchy.
Expressions can also be extrapolated, which allows you to parametrise your expressions, by nesting
expressions, substituting parts of your expression dynamically as your code is executed. Imagine
the following example.
The above expression will first evaluate the {@.arg1} parts, which results in "foo2", then evaluate the
outer expression, which now will look like this @.data/*/foo2. You can also extrapolate expressions on
values, such as illustrated below.
You can easily extend expressions in Magic, either with a "static" iterator, implying
a direct match - Or with a dynamic parametrized iterator, allowing you to create iterators that
requires "parameters". To extend the supported iterators, use any of the following two static
methods.
Iterator.AddStaticIterator - Creates a "static" iterator, implying a direct match.
Iterator.AddDynamicIterator - Creates a "dynamic iterator create function".
Below is a C# example, that creates a dynamic iterator, that will only return nodes having a value,
that once converted into a string, has exactlyn characters, not less and not more.
Iterator.AddDynamicIterator('%', (iteratorValue) => {
var no = int.Parse(iteratorValue.Substring(1));
return (identity, input) => {
return input.Where(x => x.Get<string>()?.Length == no);
};
});
var hl = @"foo
howdy1:XXXXX
howdy2:XXX
howdy3:XXXXX
";
var lambda = HyperlambdaParser.Parse(hl);
var x = new Expression("../**/%3");
var result = x.Evaluate(lambda);
Notice how the iterator we created above, uses the %3 parts of the expression, to parametrize
itself. If you exchange 3 with 5, it will only return [howdy1] and [howdy3] instead,
since it will look for values with 5 characters instead. The Iterator class can be found
in the magic.node.extensions namespace.
You can use the above syntax to override the default implementation of iterators, although
I wouldn't recommend it, since it would create confusion for others using your modified version.
Notice - To create an extension iterator is an exercise you will rarely if ever need to do,
but is included here for reference purposes.
Parsing Hyperlambda from C#
Magic allows you to easily parse Hyperlambda from C# if you need it, which can be done as follows.
using magic.node.extensions.hyperlambda;
var hl = GetHyperlambdaAsString();
var lambda = HyperlambdaParser.Parse(hl);
The GetHyperlambdaAsString above could for instance load Hyperlambda from a file, retrieve it
from your network, or some other way retrieve a snippet of Hyperlambda text. The HyperlambdaParser.Parse
parts above will return your Hyperlambda as its Node equivalent. The Parser class also have an
overloaded constructor for taking a Stream instead of a string.
Notice - The Node returned above will be a root node, wrapping all nodes found in your
Hyperlambda as children nodes. This is necessary in order to avoid enforcing a single "document node"
the way XML does.
Once you have a Node object, you can easily reverse the process by using the HyperlambdaGenerator
class, and its GetHyperlambda method such as the following illustrates.
using magic.node.extensions.hyperlambda;
var hl1 = GetHyperlambdaAsString();
var result = HyperlambdaParser.Parse(hl1);
var hl2 = HyperlambdaGenerator.GetHyperlambda(result.Children);
Formal specification of Hyperlambda
Notice - This part is mostly intended for developers wanting to implement their own Hyperlambda parser, and
is strictly not needed to understand neither Magic nor Hyperlambda. Feel free to skip this section if it seems
like Greek to you.
Hyperlambda contains 8 possible tokens in total, however since single line comments and multi line comments are
interchangeable, we simplify the specification by combining these into one logical token type - And the null token
isn't really an actual token, but rather a placeholder implying "absence of token". Possible logical tokens
found in Hyperlambda hence becomes as follows.
IND - Indent token consisting of exactly 3 SP characters.
COM - Comment token. Either C style (/**/) or C++ (//) style comments.
NAM - Name token declaring the name of some node.
SEP - Separator token separating the name of a node from its type, and/or value.
TYP - Type token declaring the type of value preceeding it. See possible types further up on page.
VAL - Value token, being the value of the node.
CRLF - CRLF character sequence, implying a CR, LF or CRLF. Except for inside string literals, Hyperlambda does not discriminate between and of these 3 possible combinations, and they all become interchangeable CRLF token types after parsing.
NUL - Null token, implying empty or non-existing token.
Notice, a VAL and a NAM token can be wrapped inside of quotes (') or double quotes ("), like a C# string type.
In addition to wrapping it inside a multiline C# type of string (@""). This allows you to declare VAL and NAM tokens
with CR/LF sequences as a part of their actual value.
The formal specification of Hyperlambda is derived from combining the above 7 tokens into the following. Notice, in the
following formal specification -> means "must be followed by if existing", [0..n] implies "zero to any number of repetitions",
[0..1] implies "zero to 1 repetition", [1..n] implies "at least one must exist",
and | implies "logical or". The paranthesis () implies a logical grouping of some token type(s), and the x= parts is
an assignable variable starting at 0, optionally incremented by one for each iteration through the loop. [0..x]
implies "zero to x repetitions" where x is the value of x, and [x..x+1] implies "x to x+1 number of repetitions".
The = character assigns the numbers of repetitions in its RHS value to the variable x. The default number of repetitions
if none are explicitly given is 1.
When referencing nodes in the documentation for Magic, it is common to reference them like [this], where
"this" would be the name of some node - Implying in bold characters, wrapped by square [brackets].
C# extensions
If you want to, you can easily completely exchange the underlaying file system, with your own "virtual file system", since all interaction with the physical file system is done through the IFileService and
IFolderService interfaces. This allows you to circumvent the default dependency injected service, and
binding towards some other implementation, at least in theory allowing you to (for instance) use a database
based file system, etc. If you want to do this, you'll need to supply your own bindings to the following
three interfaces, using your IoC container.
magic.node.contracts.IFileService
magic.node.contracts.IFolderService
magic.node.contracts.IStreamService
magic.node.contracts.IRootResolver
If you want to do this, you would probably want to manually declare your own implementation for these classes,
by tapping into "magic.library" somehow, or not invoking its default method that binds towards the default
implementation classes somehow.
In addition to the above file interfaces, the following interfaces are also declared in magic.node.extensions.
magic.node.contracts.IMagicConfiguration - Allows you to override (parts) of the internally used configuration object
magic.node.contracts.IServiceCreator - Helper interface allowing you to avoid the service locator pattern, yet still dynamically create services on a per need basis from within your own C# code
Project website for magic.node
The source code for this repository can be found at github.com/polterguy/magic.node, and you can provide feedback, provide bug reports, etc at the same place.
