Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
396 lines (310 sloc) 16 KB
Detailed Specification for
Various Componets
Paul Longtine (
This language will look and feel like any other normal, ordinary language.
The goal is not to innovate, but to see if _I_ can. This is a personal refuge
before my inevitable contributions to large group projects where individuals
are deemed inferior.
The syntax and semantics of the language are defined by the cryptic parser
implemented under /src/lc/, where each type of statement is built from
atomic definitions of various token types, static expressions, and dynamic
expressions. Each one of the matched statments in the parser has a method
"action" -> returns a list of other objects with an "action" method or bytes,
until all that is left is raw bytes in the list.
The following types are paraphrased to give you a breif overview:
0 VOID - Null, no data
1 ADDR - Address type (bytecode)
2 TYPE - A `type` type
3 PLIST - Parameter list
4 FUNC - Function
5 OBJBLDR - Object builder
6 OBJECT - Object/Class
7 G_PTR - Generic pointer
8 G_INT - Generic integer
9 G_FLOAT - Generic double
10 G_CHAR - Generic character
11 G_STR - Generic string
12 S_ARRAY - Static array
13 D_ARRAY - Dynamic array
14 H_TABLE - Hashtable
15 G_FIFO - Stack
The runtime architecture is based off of stack machines. If you don't know
about stack machines, go refresh yourself on stack machines along with basic
computer architecture / turing machines. Then come back here, I'm not explaining
that stuff for you.
General Architecture Overview
The runtime context keeps track of a invidual threads metadata, such as:
* The operating stack
The operating stack where current running instructions push/pop to.
* Namespace instance
Data structure that holds the references to variable containers, also provi
ing the interface for Namespace Levels.
* Arguement stack
Arguements to function calls are pushed on to this stack, flushed on call.
* Program counter
An interface around bytecode to keep track of traversing line-numbered
This context gives definition to an 'environment' where code is executed.
A key part to any operational computer language is the notion of a 'Namespace'.
This notion of a 'Namespace' refers to the ability to declare a name, along with
needed metadata, and call upon the same name to retrieve the values assosaited
with that name.
In this definition, the namespace will provide the following key mechanisms:
* Declaring a name
* Assigning a name to a value
* Retreiving a name's value
* Handle a name's scope
* Implicitly move in/out of scopes
The scope arguement is a single byte, where the format is as follows:
0000000 |0
Scopes are handled by referencing to either the Global Scope or the Local Scope.
The Local Scope is denoted by '0' in the scope arguement when refering to names,
and this scope is initialized when evaluating any new block of code. When a diff
erent block of code is called, a new scope is added as a new Namespace level.
Namespace levels act as context switches within function contexts. For example,
the local namespace must be 'returned to' if that local namespace context needs
to be preserved on return. Pushing 'Namespace levels' ensures that for every n
function calls, you can traverse n instances of previous namespaces.
For example, take this namespace level graphic, where each Level is a namespace
Level 0: Global namespace, LSB == '1'. Raw: 00000001
Level 1: Namespace level, LSB == '0'. Raw: 00000000
When a function is called, another namespace level is created and the local
level increases, like so:
Level 0: Global namespace, LSB == '1'. Raw: 00000001
<function call>
Level 1: Namespace level, where Local Level is at 1, LSB == '0'. Raw: 00000000
<function call>
Level 2: Namespace level, where Local Level is at 2, LSB == '0'. Raw: 00000000
Global scope names (LSB == 1 in the scope arguement) are persistient
through the runtime as they handle all function definitions, objects, and
names declared in the global scope. The "Local Level" is at where references
that have a scope arguement of '0' refer to when accessing names.
The Namespace arguement refers to which Namespace the variable exists in.
When the namespace arguement equals 0, the current namespace is referenced.
The global namespace is 1 by default
Variables in this definiton provide the following mechanims:
* Provide a distinguishable area of typed data
* Provide a generic container around typed data, to allow for labeling
* Declare a set of fundemental datatypes, and methods to:
* Allocate the proper space of memory for the given data type,
* Deallocate the space of memory a variables data may take up, and
* Set in place a notion of ownership
For a given variable V, V defines the following attributes
V -> Ownership
V -> Type
V -> Pointer to typed space in memory
Each variable then can be handled as a generic container.
In the previous section, the notion of Namespace levels was introduced. Much
like how names are scoped, generic variable containers must communicate their
scope in terms of location within a given set of scopes. This is what is called
'Ownership'. In a given runtime, variable containers can exist in the following
structures: A stack instance, Bytecode arguements, and Namespaces
The concept of ownership differentiates variables existing on one or more of the
structures. This is set in place to prevent accidental deallocation of variable
containers that are not copied, but instead passed as references to these
Functions in this virtual machine are a pointer to a set of instructions in a
program with metadata about parameters defined.
In this paradigm, objects are units that encapsulate a seperate namespace and
collection of methods.
Bytecode is arranged in the following order:
<opcode>, <arg 0>, <arg 1>, <arg 2>
Where the <opcode> is a single byte denoting which subroutine to call with the
following arguements when executed. Different opcodes have different arguement
lengths, some having 0 arguements, and others having 3 arguements.
Interpreting Bytecode Instructions
A bytecode instruction is a single-byte opcode, followed by at maximum 3
arguements, which can be in the following forms:
* Static (single byte)
* Name (single word)
* Address (depending on runtime state, usually a word)
* Dynamic (size terminated by NULL, followed by (size)*bytes of data)
* i.e. FF FF 00 <0xFFFF bytes of data>,
01 00 <0x1 bytes of data>,
06 00 <0x6 bytes of data>, etc
Below is the specification of all the instructions with a short description for
each instruction, and instruction category:
TOS - 'Top Of Stack' The top element
TBI - 'To be Implemented'
S<[variable]> - Static Arguement.
N<[variable]> - Name.
A<[variable]> - Address Arguement.
D<[variable]> - Dynamic bytecode arguement.
Hex | Memnonic | arguments - description
1 - Stack manipulation
These subroutines operate on the current-working stack(1).
10 POP S<n> - pops the stack n times.
11 ROT - rotates top of stack
12 DUP - duplicates the top of the stack
13 ROT_THREE - rotates top three elements of stack
2 - Variable management
20 DEC S<scope> S<type> N<ref> - declare variable of type
21 LOV S<scope> N<ref> - loads reference variable on to stack
22 STV S<scope> N<ref> - stores TOS to reference variable
23 CTV S<scope> N<ref> D<data> - loads constant into variable
24 CTS D<data> - loads constant into stack
3 - Type management
Types are in the air at this moment. I'll detail what types there are when
the time comes
30 TYPEOF - pushes type of TOS on to the stack TBI
31 CAST S<type> - Tries to cast TOS to <type> TBI
4 - Binary Ops
OPS take the two top elements of the stack, preform an operation and push
the result on the stack.
40 ADD - adds
41 SUB - subtracts
42 MULT - multiplies
43 DIV - divides
44 POW - power, TOS^TOS1 TBI
45 BRT - base root, TOS root TOS1 TBI
46 SIN - sine TBI
47 COS - cosine TBI
48 TAN - tangent TBI
49 ISIN - inverse sine TBI
4A ICOS - inverse consine TBI
4B ITAN - inverse tangent TBI
4C MOD - modulus TBI
4D OR - or's TBI
4E XOR - xor's TBI
4F NAND - and's TBI
5 - Conditional Expressions
Things for comparison, < > = ! and so on and so forth.
Behaves like Arithmetic instructions, besides NOT instruction. Pushes boolean
to TOS
50 GTHAN - Greater than
51 LTHAN - Less than
52 GTHAN_EQ - Greater than or equal to
53 LTHAN_EQ - Less than or equal to
54 EQ - Equal to
55 NEQ - Not equal to
56 NOT - Inverts TOS if TOS is boolean
57 OR - Boolean OR
58 AND - Boolean AND
6 - Loops
60 STARTL - Start of loop
61 CLOOP - Conditional loop. If TOS is true, continue looping, else break
6E BREAK - Breaks out of loop
6F ENDL - End of loop
7 - Code flow
These instructions dictate code flow.
70 GOTO A<addr> - Goes to address
71 JUMPF A<n> - Goes forward <n> lines
72 IFDO - If TOS is TRUE, do until done, if not, jump to done
73 ELSE - Chained with an IFDO statement, if IFDO fails, execute ELSE
block until DONE is reached.
74 JTR - jump-to-return. TBI
75 JTE - jump-to-error. Error object on TOS TBI
7D ERR - Start error block, uses TOS to evaluate error TBI
7E DONE - End of block
7F CALL N<ref> - Calls function, pushes return value on to STACK.
8 - Generic object interface. Expects object on TOS
80 GETN N<name> - Returns variable assosiated with name in object
81 SETN N<name> - Sets the variable assosiated with name in object
Object on TOS, Variable on TOS1
82 CALLM N<name> - Calls method in object
83 INDEXO - Index an object, uses arguement stack
84 MODO S<OP> - Modify an object based on op. [+, -, *, /, %, ^ .. etc]
F - Functions/classes
FF DEFUN N<ref> S<type> D<args> - Un-funs everything. no, no- it defines a
function. D<ref> is its name, S<type> is
the return value, D<args> is the args.
FE DECLASS N<ref> D<args> - Defines a class.
FD DENS S<ref> - Declares namespace
F2 ENDCLASS - End of class block
F1 NEW S<scope> N<ref> - Instantiates class
F0 RETURN - Returns from function
00 NULL - No-op
01 LC N<name> - Calls OS function library, i.e. I/O, opening files, etc TBI
02 PRINT - Prints whatever is on the TOS.
03 DEBUG - Toggle debug mode
0E ARGB - Builds arguement stack
0F PC S<ref> - Primitive call, calls a subroutine A<ref>. A list of TBI
primitive subroutines providing methods to tweak
objects this bytecode set cannot touch. Uses argstack.
Going from code to bytecode is what this section is all about. First off an
abstract notation for the code will be broken down into a binary tree as so:
/ \
/ \
<arg> <next>
<node> can be an argument of its parent node, or the next instruction.
Instruction nodes are nodes that will produce an instruction, or multiple based
on the bytecode interpretation of its instruction. For example, this line of
int x = 3
would translate into:
/ \
/ \
/ \
/ \
int set
/\ /\
/ \ / \
null 'x' 'x' null
/ \
null 3
Functions are expressed as individual binary trees. The root of any file is
treated as an individual binary tree, as this is also a function.
The various instruction nodes are as follows:
* def <type> <name>
- Define a named space in memory with the type specified
- See the 'TYPES' section under 'OVERVIEW'
* set <name> <value>
- Set a named space in memory with value specified
Going from Binary Trees to Bytecode
The various instruction nodes within the tree will call specific functions
that will take arguemets specified and lookahead and lookbehind to formulate the
correct bytecode equivilent.