also changed print() to print raw primitives instead of generating a <string> needlessly. helped performance immensely in scripts like `examples/increment.cosmo`
The object library includes:
- __proto getter and setter tables (might make an option to lock __proto in the future)
- ischild()
cosmoB_loadDebug was renamed to cosmoB_loadVM and now only loads the VM library
You can use ':' without invoking. So you can grab a method binding to an object and later call it. Also some GC bugs were fixed.
A new preprocessor macro (VM_DEBUG) has been added. Defining this macro will print each instruction thats being executed in the VM, as
well as print each call.
a new instruction, OP_NEWARRAY has been added. this instruction will create a table out of the uint16_t values on the stack, with incremental keys [0-(u16-1)]
Tables are going to become a more generic form of data storage. While objects will be "stateful", tables will remain static.
'{}' now refers to an object declaration, '[]' will be reserved for tables.
Errors are now handled very differently, parser errors and VM errors are now treated the same.
When cosmoV_error is called, cosmoV_throw is also called, which formats the error object and sets the panic state.
state->error now points to the latest CObjError when state->panic is true. To get a nice formatted Objection message, use
cosmoV_printError() and pass the state->error. pcall() was added to the standard base library. When called, the first argument
passed is called with the subsequent arguments given. If the call completed successfully, `true`,`nil` is returned. However
when an error occurs during the call, `false`,`<error>` is returned. Simply print the `<error>` to retrieve the error string.
- added OP_MOD, which performs a modulo operation on the 2 <number> values on the stack. Pops the 2 values and pushes the result.
- also added TOKEN_PERCENT to the lexer, and extended binary() in cparse.c to support it.
cosmoO_getObject and cosmoO_setObject were renamed to cosmoV_getRawObject and cosmoV_setRawObject respectively
`__init` is now required to be defined for a proto object to be instantiated.
other minor refactoring work done in src/cvm.c
a LoopState was added to the CCompilerState struct which keeps track of breaks, start chunk index, and the start scope of the active loop.
Also, break.cosmo was added to the examples directory, 'continue' and 'break' work as expected.
cosmoV_error now pushes the error string onto the stack, as well as uses cosmoO_VFString to format strings.
cosmoV_concat is now entirely self contained, and is now stack based. Push the CValues you want to concat into strings onto the stack and call cosmoV_concat with the # of values you pushed.
ISTRING_RESERVED was added to iStrings
call & callCFunction now use memmove instead of memcpy
Additionally, added cosmoO_setUserP, cosmoO_getUserP, cosmoO_setUserI and cosmoO_getUserI to the C API.
__iter and __next are now reserved IStrings, OP_NEXT and OP_ITER have also been added. A new token (TOKEN_IN) has been added to the lexer. The parser now supports the for each loop (for i, ... in <object> do ... end).
savedPushed has been removed from the CCompilerState struct.
Objects are now separate from {} dictionaries
the . operator now only indexes fields on objects, the [] operator can only be used on objects if the __index or __newindex functions are defined
Additionally 4 new instructions have been added to the VM: OP_NEWDICT, OP_INDEX, OP_INCINDEX, and OP_NEWINDEX.
The syntax to create a dictionary is as follows { <key> : <value>, <otherkey> : <othervalue> } eg. { "hello" : "world", "foo" : 1337 }
The Lexer & Parser was extended to add the TOKEN_COLON ':' token.
Optimized the NaN box to be byte aligned, performance under the fibtest.lua script improved by ~2.5 seconds (~31 before ~28.5 after on cpunch's machine)
also cleaned up some misc. comments
