Minimal systems programming language
- C 99.9%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| docs | ||
| include/tau | ||
| samples | ||
| source | ||
| test | ||
| .gitignore | ||
| build.sh | ||
| compile_flags.txt | ||
| LICENSE | ||
| main.tau | ||
| README.md | ||
tau-lang
Minimal systems programming language
Building
Dependencies:
- C compiler (like clang or gcc)
- LLVM library (llvm-c to be exact)
The current method of compiling the compiler is by running the build.sh script.
TODO
-
Priority 1
- Implement Diagnostics system
- Implement Parser #notcl4y
- Add other token types (-, *, /, +=, -=, ==, !=, !, &, ||, comma, colon, etc.)
- Implement AST printer
- Implement diagnostics into parser and semantics
- Implement Semantics Analysis #notcl4y
-
Priority 2
- Implement source ranges for tokens and nodes
- Implement Source into Compiler (and allow the systems to read the source's filename)
- Implement lexer
- Fix a "typo" when printing diagnostics
- Two-character operator tokens
- Add Diagnostics-related functions for Compiler
- Add functions for printing diagnostics result
- Implement CLI (at least file reading)
- Create functions for printing source positions, source ranges, token types, AST node types
-
Priority 3
- Parse semicolons by ignoring them
- Implement IR Builder
- IR Build binary expressions (more operators: -, * and /)
- IR Build resolve identifiers
- Assign symbols to identifiers during Semantics Analysis (AST identifier terms AND AST variable statements)
- Parse functions
- Analyse functions
- Analyse blocks
- IR Build functions and blocks
- Implement function parameters
- Implement block statements
- Implement sub-scopes for functions and block statements
- Remove useless AST semantics decorators
-
Priority 4
- Typeless and valueless variable/const statements
- Implement more IR canonical types
- Implement auto-casting (for IR)
- Allow void types only in function return types
- Implement certain functions for "convenience" to prevent some "long messy member accesses" (like
node->varStat.ident->identTerm.ident->AstNodeGetIdentValue(node->varStat.ident)) - Implement special binary operations for certain types (like float) (That also includes cast operations)
- Implement floats
- Implement return statements
- Build return statements
- Keep track of functions and check for type matching with return statements using that
-
Priority 5
- Parse, analyse and build call expressions
- Parse, analyse and build struct statements
- Make structs as types
- Parse, analyse and build member expressions
- Have struct fields symbols in a different scope or stored in a way such that they won't conflict with their neighbor symbols
- Build comparative binary operations
- Implement if statements
- Analyse terms and determine their type from their surrounding
- Analyse binexpr-s to assign cast expr types
- Create a helper function for handling implicit casts in Semantics Analysis
- Implement while statements
- Parse, analyse and build assignment expressions
-
Priority 6
- Parenthesised expressions
- Parse, analyse and build pointers (and dereference)
- Parse, analyse and build addresses
- Untangle the confusion between functions SemanticsAnalysisConstructTypeSymbol, SemanticsAnalysisResolveIdent and SemanticsAnalysisResolveType
- Parse unary expressions
- Build logical binary and unary operations
-
Priority 7
- CLI print stage results
- Bodyless function declarations
- LLVM output ll
- Analyse strings
- Build strings
-
Priority 8
- Figure out why TypeTable doesn't work the way it's expected to work (semantics analysis has to fallback to searching canonical types in the symbol table)
- Clean up
- Remove unnecessary comments (ir.c, parseAst.c)
- irVisit.c: IrBuilderVisitAssignExpr
- Rename AST PARAM_ARR node types to PARAMS
- Align spaces
- ast.h
- irVisit.c: ___IrBuilderResolveCanonicalType
- token.c: TokenTypeAsString
- Document SemanticsAnalysisHandleImplicitCast
- Whatever at those places are
- irVisit.c: IrBuilderResolveTypeCast
- Separate functions' parts into helper functions
- irVisit.c: IrBuilderVisitFnStat
- irVisit.c: IrBuilderVisitStructStat
- irVisit.c: IrBuilderVisitCallExpr
- analysisVisit.c: SemanticsAnalysisVisitMemberExpr
- analysisVisit.c: SemanticsAnalysisVisitStructStat
- analysisVisit.c: SemanticsAnalysisVisitFnStat
- Remove unnecessary comments (ir.c, parseAst.c)
- Improve token array printer
- Implement symbol table printer
- Implement string buffer builder
- Integrate string arena
-
Priority 9
- Bodyless structs
- Analyse for function parameter count to compare it with argument count
- Add missing errors
- analysisVisit.c: SemanticsAnalysisVisitReturnStat (check for function return type and void)
- Lexerize split ("spaced") integers (
100_000_000,20_000) - Lexerize HEX integers (
0x1234ABCD,0xffff) - Detect attempts to modify constants
- Global variables
- Bit/bitshift operations and operators
- Detect identifier redefinitions
- Constant values such as null, true and false
- Comments
- Enums
- Reconceptualize member expression relations and pointer-value relations
- Custom enum types and auto determine enum's type
- Custom enum values
- Implement IR boolean expression resolvers
- Bools sometimes don't get IR built into i1-s for some reason or something (they become i8 instead and refuse to even be casted)
- Implement HIR (High Level IR)
- Integrate HIR. Make the Semantics Analysis generate it and LLVM IR Builder take it
-
Priority 10
- Arrays
- Unions
-
Priority Unplanned
- This list is for those TODO entries that have an unspecified priority or a yet-to-be-planned implementation.
- Document functions, structures, modules
- Implement proper error system (not the diagnostics, those are only for the programming language itself and not the compiler)
- Support implicit types
- Allocate variables for parameters when required (taking an address from a parameter or changing a parameter's value) and replace them
- Warn potential null pointer dereferences
- Error for doing logic stuff outside of functions
-
Priority unprioritized
- This list is for those TODO entries that need to be done but have yet to be prioritized
- Implement a CLI option to disable some stages (as in for example "up until to parsing", which disables semantics analysis and IR builder)
- Search for memory leaks and free them (might require setting up some free systems like arenas)
- Require semicolons at the end of the lines
- Lexerize strings with backslash characters
- Lexerize literals with explicit types (like
25.0f,4.5f64,255u8) - Lexerize zero-less float tokens (
.0,.25f) - Lexerize binary integers (
0b101,0b10000011) - Multiple fields with one type (instead of
x: i32, y: i32you can dox, y: i32) - Character literals
- Analyse and build overload functions
- Parse, analyse and build methods
- Tuples
- Parse, analyse, build arrays
- Make IR Builder independent of the declaration placement (if you declare a global variable after a function, IR gets confused)
- And after that check whether or not setting global variables to function-returnd values work
- Fix order of parse precedence (like binary operations)
- Store computed enum values as constants instead of recomputing each IR visit
- Implement modules (not imports. imports import modules from files)
- Implement imports (do modules before imports)
- Struct and array initializers (compound and array literals)
Example syntax
import "std.io";
import "std.string";
fn main (): void {
const x: i32 = 10;
const y: i32 = 20;
const z: i32 = x + y;
const string: String = ("{1} + {2} = {3}").format(x, y, z);
std.io.println(string); // "10 + 20 = 30"
}