diff --git a/README.md b/README.md index 6d212b7..19ca98e 100644 --- a/README.md +++ b/README.md @@ -1,38 +1,129 @@ -# ISSUES -When a memory bug is detected in one of the treads it won't report a non-zero -exit code while running the tests. +# what is the goal? +- A standard library inspired by languages like zig, odin and rust. +- Single header, multiple source files, and no build step. -# A collection of libraries inspired by zig, and other stuff. +# current and coming features -I like zig, but I really like writing C. I am starting with implementing -different allocator types. I don't know how the other parts of the library will -fit in, but they will certainly use the allocators. +- [x] allocators +- [x] dynamic arrays +- [x] simple cli arg parsing +- [-] functions to build parsers +- [-] string builder +- [ ] hashmap +- [ ] easing functions +- [ ] simple gui engine -I want a series of standard collection types with nice interfaces. ArrayLists, -LinkedLists, value-based HashMaps, reference-based HashMaps. +# example -I also want to provide template header files that use the X-macro pattern so -that users can easily make tagged unions, option types, automatic bit-flags, -and maybe even an ECS inspierd struct filled with optional fields. +## allocators and dynamic arrays +```c +#include +#include -I probably want a ui library. It will rely on a graphical drawing interface. It -will be a little different though. It is inspired from my experience with using -raygui. instead of making some sort of layout engine, I will provide a set of -useful functions to manipulate rectangles. I think this produces programs that -are easier to reason about, while still reducing needless duplication of code. -The workflow relies heavily on function-static variables, which are a c -superpower in this scenario. You will essentially start with some 'fixed' -rectangle (probably derived from the window size) and then perform splitting -operations on that rectangle. Even a dynamic split that creates a draggable -border, updating the local function-static variable (which is a normalized -float) to let you resize sections. Also, the interface will be like immediate -mode guis, with one exception from the way raygui works. The drawing operations -are deferred, and called in the reverse order. This way the first gui function -that captrues input can signal to the other function that input is captured, -and the drawing of the gui elements can reflect the priority of the gui -functions. Overlapping gui elements in raygui is the main painpoint imo. +#define CIG_IMPL +#include "cig/cig.h" -# TODO +int main(void) { + + with_borrow(allocator) { + // initial capacity is optional + int *ns = make_arr(int, allocator, .initial_capacity=10); + + for (int i = 0; i < 100; i++) { + // the array grows as the capacity is reached + arr_append(ns, i); + } + + for (int i = 0; i < 10; i++) { + int last = arr_pop(ns); + printf("last value was %d\n", last); + } + + for (int i = 0; i < arr_len(ns); i++) { + printf("value of ns[%d] is %d\n", i, ns[i]); + } + + // you can break out of a with_borrow scope + break; + + printf("this is never printed\n"); + + // WARNING: returning from within a with_borrow + // scope will cause a memory leak + return; + } + // everything allocated by the borrow allocator is + // freed here, and the borrow allocator is also out of + // scope. + + + // you can also make the borrow allocator like this: + allocator_t backing = borrow_allocator_create(); + + // this arena allocator is slightly clever. You assign + // an initial size of the memory block, but uses a + // backing arena to allow you to allocate more than + // that. Whenever allocator_reset is called it figures + // out how much was allocated outside the big chunk of + // memory, frees everything and allocates a bigger + // block. So it's size will move towards what you + // happen to use in the game loop. It assumes that it + // is the owner of the given backing allocator, and it + // should probably be a borrow_allocator. + allocator_t arena_allocator = arena_allocator_create(backing, 100 * MB); + + // Imagine a game loop + while (true) { + // every time this resets, assuming the backing + // memory chunk doesn't need to grow, this + // operation is very fast. + allocator_reset(arena_allocator); + + float *fs = make_arr(float, arena_allocator); + arr_append(fs, 1.1); + arr_append(fs, 1.2); + arr_append(fs, 1.3); + + break; + } + + allocator_reset(backing); + return 0; +} +``` + +## cli +```c +#include +#include +#include + +#define CIG_IMPL +#include "cig/cig.h" + +int main(int argc, const char **argv) { + args_t args = cli_make_args(argc, argv); + + if (cli_command(&args, "say-foo")) { + printf("foo\n"); + } else + if (cli_command(&args, "say-bar")) { + printf("bar\n"); + } else + if (cli_command(&args), "greet") { + bool do_nothing = cli_bool(args, "--do-nothing"); + const char *name = cli_req_str(args, "--name"); + if (args.help) exit(0); + + if (do_nothing) { + return; + } + + printf("Hello %s!\n", name); + } else { + printf("use the -h or --help flag to see available commands\n"); + } +} + +``` -If there is an allocation error it does not actually cause a crash... -FIX!!!