Tue, 12 Nov 2024 17:34:04 +0100
add configure option for szmul
--- title: UCX 2.1 Modules --- UCX 2.1 provided several modules for data structures and algorithms. You may choose to use specific modules by including the corresponding header file. Please note, that some modules make use of other UCX 2.1 modules. For instance, the [Allocator](#allocator) module is used by many other modules to allow flexible memory allocation. By default, the header files are placed into an `ucx` directory within your systems include directory. In this case you can use a module by including it via `#include <ucx/MODULENAME.h>`. Required modules are included automatically. <div id="modules"> ----------------------- ---------------------- -------------------------------- --------------------------- [String](#string) [Buffer](#buffer) [Allocator](#allocator) [Stack](#stack) [Memory Pool](#memory-pool) [Array](#array) [List](#list) [Map](#map) [AVL Tree](#avl-tree) [Logging](#logging) [Testing](#testing) [Utilities](#utilities) [Properties](#properties) ----------------------- ---------------------- -------------------------------- --------------------------- </div> ## Allocator *Header file:* [allocator.h](api-2.1/allocator_8h.html) *Required modules:* None. A UCX allocator consists of a pointer to the memory area / pool and four function pointers to memory management functions operating on this memory area / pool. These functions shall behave equivalent to the standard libc functions `malloc`, `calloc`, `realloc` and `free`. The signature of the memory management functions is based on the signature of the respective libc function but each of them takes the pointer to the memory area / pool as first argument. As the pointer to the memory area / pool can be arbitrarily chosen, any data can be provided to the memory management functions. One example is the [UCX Memory Pool](#memory-pool). ## Array *Header file:* [array.h](api-2.1/array_8h.html) *Required modules:* [Allocator](#allocator) The UCX Array is an implementation of a dynamic array with automatic reallocation. The array structure contains a capacity, the current size, the size of each element, the raw pointer to the memory area and an allocator. Arrays are in most cases much faster than linked list. One can decide, whether to create a new array on the heap with `ucx_array_new()` or to save one indirection by initializing a `UcxArray` structure on the stack with `ucx_array_init()`. ### Remove duplicates from an array of strings The following example shows, how a `UcxArray` can be built with a standard dynamic C array (pointer+length) as basis. ```C UcxArray* create_unique(sstr_t* array, size_t arrlen) { // worst case is no duplicates, hence the capacity is set to arrlen UcxArray* result = ucx_array_new(arrlen, sizeof(sstr_t)); // only append elements, if they are not already present in the array for (size_t i = 0 ; i < arrlen ; ++i) { if (!ucx_array_contains(result, array+i, ucx_cmp_sstr, NULL)) { ucx_array_append_from(result, array+i, 1); } } // make the array as small as possible ucx_array_shrink(result); return result; } // ... sstr_t* array = // some standard array of strings size_t arrlen = // the length of the array UcxArray* result = create_unique(array,arrlen); // Iterate over the array and print the elements sstr_t* unique = result->data; for (size_t i = 0 ; i < result->size ; i++) { printf("%" PRIsstr "\n", SFMT(unique[i])); } // Free the array. ucx_array_free(result); ``` ### Preventing out of bounds writes The functions `ucx_array_reserve()`, `ucx_array_resize()`, `ucx_array_grow()`, and `ucx_array_shrink()` allow easy management of the array capacity. Imagine you want to add `n` elements to an array. If your `n` elements are already somewhere else consecutively in memory, you can use `ucx_array_append_from()` and benefit from the autogrow facility in this family of functions. Otherwise, you can ask the array to have enough capacity for holding additional `n` elements. ```C size_t n = // ... elements to add if (ucx_array_grow(array, n)) { fprintf(stderr, "Cannot add %zu elements to the array.\n", n); return 1; } for (size_t i = 0 ; i < n ; i++) { ((int*)array->data)[array->size++] = 80; } ``` ## AVL Tree *Header file:* [avl.h](api-2.1/avl_8h.html) *Required modules:* [Allocator](#allocator) This binary search tree implementation allows average O(1) insertion and removal of elements (excluding binary search time). All common binary tree operations are implemented. Furthermore, this module provides search functions via lower and upper bounds. ### Filtering items with a time window Suppose you have a list of items which contain a `time_t` value and your task is to find all items within a time window `[t_start, t_end]`. With AVL Trees this is easy: ```C // Somewhere in a header typedef struct { time_t ts; // other important data } MyObject; // Source code UcxAVLTree* tree = ucx_avl_new(ucx_cmp_longint); // ... populate tree with objects, use '& MyObject.ts' as key ... // Now find every item, with 30 <= ts <= 70 time_t ts_start = 30; time_t ts_end = 70; printf("Values in range:\n"); for ( UcxAVLNode* node = ucx_avl_find_node( tree, (intptr_t) &ts_start, ucx_dist_longint, UCX_AVL_FIND_LOWER_BOUNDED); node && (*(time_t*)node->key) <= ts_end; node = ucx_avl_succ(node) ) { printf(" ts: %ld\n", ((MyObject*)node->value)->ts); } ucx_avl_free_content(tree, free); ucx_avl_free(tree); ``` ## Buffer *Header file:* [buffer.h](api-2.1/buffer_8h.html) *Required modules:* None. Instances of this buffer implementation can be used to read from or to write to memory like you would do with a stream. This allows the use of `ucx_stream_copy()` from the [Utilities](#utilities) module to copy contents from one buffer to another or from file or network streams to the buffer and vice-versa. More features for convenient use of the buffer can be enabled, like automatic memory management and automatic resizing of the buffer space. See the documentation of the macro constants in the header file for more information. ### Add line numbers to a file When reading a file line by line, you have three options: first, you could limit the maximum supported line length. Second, you allocate a god buffer large enough for the most lines a text file could have. And third, undoubtedly the best option, you start with a small buffer, which adjusts on demand. An `UcxBuffer` can be created to do just that for you. Just pass the `UCX_BUFFER_AUTOEXTEND` option to the initialization function. Here is a full working program, which adds line numbers to a file. ```C #include <stdio.h> #include <ucx/buffer.h> #include <ucx/utils.h> int main(int argc, char** argv) { if (argc != 2) { fprintf(stderr, "Usage: %s <file>\n", argv[0]); return 1; } FILE* input = fopen(argv[1], "r"); if (!input) { perror("Canno read input"); return 1; } const size_t chunksize = 256; UcxBuffer* linebuf = ucx_buffer_new( NULL, // the buffer should manage the memory area for us 2*chunksize, // initial size should be twice the chunk size UCX_BUFFER_AUTOEXTEND); // the buffer will grow when necessary size_t lineno = 1; do { // read line chunk size_t read = ucx_stream_ncopy( input, linebuf, fread, ucx_buffer_write, chunksize); if (read == 0) break; // handle line endings do { sstr_t bufstr = ucx_buffer_to_sstr(linebuf); sstr_t nl = sstrchr(bufstr, '\n'); if (nl.length == 0) break; size_t linelen = bufstr.length - nl.length; sstr_t linestr = sstrsubsl(bufstr, 0, linelen); printf("%zu: %" PRIsstr "\n", lineno++, SFMT(linestr)); // shift the buffer to the next line ucx_buffer_shift_left(linebuf, linelen+1); } while(1); } while(1); // print the 'noeol' line, if any sstr_t lastline = ucx_buffer_to_sstr(linebuf); if (lastline.length > 0) { printf("%zu: %" PRIsstr, lineno, SFMT(lastline)); } fclose(input); ucx_buffer_free(linebuf); return 0; } ``` ## List *Header file:* [list.h](api-2.1/list_8h.html) *Required modules:* [Allocator](#allocator) This module provides the data structure and several functions for a doubly linked list. Among the common operations like insert, remove, search and sort, we allow convenient iteration via a special `UCX_FOREACH` macro. ### Remove duplicates from an array of strings Assume you are given an array of `sstr_t` and want to create a list of these strings without duplicates. This is a similar example to the one [above](#array), but here we are using a `UcxList`. ```C #include <stdio.h> #include <ucx/list.h> #include <ucx/string.h> #include <ucx/utils.h> UcxList* remove_duplicates(sstr_t* array, size_t arrlen) { UcxList* list = NULL; for (size_t i = 0 ; i < arrlen ; ++i) { if (ucx_list_find(list, array+i, ucx_cmp_sstr, NULL) == -1) { sstr_t* s = malloc(sizeof(sstr_t)); *s = sstrdup(array[i]); list = ucx_list_append(list, s); } } return list; } // we will need this function to clean up the list contents later void free_sstr(void* ptr) { sstr_t* s = ptr; free(s->ptr); free(s); } // ... sstr_t* array = // some array of strings size_t arrlen = // the length of the array UcxList* list = remove_duplicates(array,arrlen); // Iterate over the list and print the elements UCX_FOREACH(elem, list) { sstr_t s = *((sstr_t*)elem->data); printf("%" PRIsstr "\n", SFMT(s)); } // Use our free function to free the duplicated strings. ucx_list_free_content(list, free_sstr); ucx_list_free(list); ``` ## Logging *Header file:* [logging.h](api-2.1/logging_8h.html) *Required modules:* [Map](#map), [String](#string) The logging module comes with some predefined log levels and allows some more customization. You may choose if you want to get timestamps or source file and line number logged automatically when outputting a message. The following function call initializes a debug logger with all of the above information: ```C log = ucx_logger_new(stdout, UCX_LOGGER_DEBUG, UCX_LOGGER_LEVEL | UCX_LOGGER_TIMESTAMP | UCX_LOGGER_SOURCE); ``` Afterwards you can use this logger with the predefined macros ```C ucx_logger_trace(log, "Verbose output"); ucx_logger_debug(log, "Debug message"); ucx_logger_info(log, "Information"); ucx_logger_warn(log, "Warning"); ucx_logger_error(log, "Error message"); ``` or you use ```C ucx_logger_log(log, CUSTOM_LEVEL, "Some message") ``` When you use your custom log level, don't forget to register it with ```C ucx_logger_register_level(log, CUSTOM_LEVEL, "CUSTOM") ``` where the last argument must be a string literal. ## Map *Header file:* [map.h](api-2.1/map_8h.html) *Required modules:* [Allocator](#allocator), [String](#string) This module provides a hash map implementation using murmur hash 2 and separate chaining with linked lists. Similarly to the list module, we provide a `UCX_MAP_FOREACH` macro to conveniently iterate through the key/value pairs. ### Parsing command line options Assume you want to parse command line options and record them within a map. One way to do this is shown by the following code sample: ```C UcxMap* options = ucx_map_new(16); const char *NOARG = ""; char *option = NULL; char optchar = 0; for(int i=1;i<argc;i++) { char *arg = argv[i]; size_t len = strlen(arg); if(len > 1 && arg[0] == '-') { for(int c=1;c<len;c++) { if(option) { fprintf(stderr, "Missing argument for option -%c\n", optchar); return 1; } switch(arg[c]) { default: { fprintf(stderr, "Unknown option -%c\n\n", arg[c]); return 1; } case 'v': { ucx_map_cstr_put(options, "verbose", NOARG); break; } case 'o': { option = "output"; optchar = 'o'; break; } } } } else if(option) { ucx_map_cstr_put(options, option, arg); option = NULL; } else { // ... handle argument that is not an option ... } } if(option) { fprintf(stderr, "Missing argument for option -%c\n", optchar); return 1; } ``` With the following loop, you can access the previously recorded options: ```C UcxMapIterator iter = ucx_map_iterator(options); char *arg; UCX_MAP_FOREACH(optkey, arg, iter) { char* opt = optkey.data; if (*arg) { printf("%s = %s\n", opt, arg); } else { printf("%s active\n", opt); } } ``` Don't forget to call `ucx_map_free()`, when you are done with the map. ## Memory Pool *Header file:* [mempool.h](api-2.1/mempool_8h.html) *Required modules:* [Allocator](#allocator) Here we have a concrete allocator implementation in the sense of a memory pool. This pool allows you to register destructor functions for the allocated memory, which are automatically called on the destruction of the pool. But you may also register *independent* destructor functions within a pool in case some external library allocated memory for you, which should be destroyed together with this pool. Many UCX modules support the use of an allocator. The [String Module](#string), for instance, provides the `sstrdup_a()` function, which uses the specified allocator to allocate the memory for the duplicated string. This way, you can use a `UcxMempool` to keep track of the memory occupied by duplicated strings and cleanup everything with just a single call to `ucx_mempool_destroy()`. ### Read CSV data into a structure The following code example shows some of the basic memory pool functions and how they can be used with other UCX modules. ```C #include <stdio.h> #include <ucx/mempool.h> #include <ucx/list.h> #include <ucx/string.h> #include <ucx/buffer.h> #include <ucx/utils.h> typedef struct { sstr_t column_a; sstr_t column_b; sstr_t column_c; } CSVData; int main(int argc, char** argv) { UcxMempool* pool = ucx_mempool_new(128); FILE *f = fopen("test.csv", "r"); if (!f) { perror("Cannot open file"); return 1; } // close the file automatically at pool destruction ucx_mempool_reg_destr(pool, f, (ucx_destructor) fclose); // create a buffer and register it at the memory pool for destruction UcxBuffer* content = ucx_buffer_new(NULL, 256, UCX_BUFFER_AUTOEXTEND); ucx_mempool_reg_destr(pool, content, (ucx_destructor) ucx_buffer_free); // read the file and split it by lines first ucx_stream_copy(f, content, fread, ucx_buffer_write); sstr_t contentstr = ucx_buffer_to_sstr(content); ssize_t lc = 0; sstr_t* lines = sstrsplit_a(pool->allocator, contentstr, S("\n"), &lc); // skip the header and parse the remaining data UcxList* datalist = NULL; for (size_t i = 1 ; i < lc ; i++) { if (lines[i].length == 0) continue; ssize_t fc = 3; sstr_t* fields = sstrsplit_a(pool->allocator, lines[i], S(";"), &fc); if (fc != 3) { fprintf(stderr, "Syntax error in line %zu.\n", i); ucx_mempool_destroy(pool); return 1; } CSVData* data = ucx_mempool_malloc(pool, sizeof(CSVData)); data->column_a = fields[0]; data->column_b = fields[1]; data->column_c = fields[2]; datalist = ucx_list_append_a(pool->allocator, datalist, data); } // control output UCX_FOREACH(elem, datalist) { CSVData* data = elem->data; printf("Column A: %" PRIsstr " | " "Column B: %" PRIsstr " | " "Column C: %" PRIsstr "\n", SFMT(data->column_a), SFMT(data->column_b), SFMT(data->column_c) ); } // cleanup everything, no manual free() needed ucx_mempool_destroy(pool); return 0; } ``` ### Overriding the default destructor Sometimes you need to allocate memory with `ucx_mempool_malloc()`, but the memory is not supposed to be freed with a simple call to `free()`. In this case, you can overwrite the default destructor as follows: ```C MyObject* obj = ucx_mempool_malloc(pool, sizeof(MyObject)); // some special initialization with own resource management my_object_init(obj); // register destructor function ucx_mempool_set_destr(obj, (ucx_destructor) my_object_destroy); ``` Be aware, that your destructor function should not free any memory, that is also managed by the pool. Otherwise, you might be risking a double-free. More precisely, a destructor function set with `ucx_mempool_set_destr()` MUST NOT call `free()` on the specified pointer whereas a destructor function registered with `ucx_mempool_reg_destr()` MAY (and in most cases will) call `free()`. ## Properties *Header file:* [properties.h](api-2.1/properties_8h.html) *Required modules:* [Map](#map) This module provides load and store function for `*.properties` files. The key/value pairs are stored within an UCX Map. ### Example: Loading properties from a file ```C // Open the file as usual FILE* file = fopen("myprops.properties", "r"); if (!file) { // error handling return 1; } // Load the properties from the file UcxMap* myprops = ucx_map_new(16); if (ucx_properties_load(myprops, file)) { // ... error handling ... fclose(file); ucx_map_free(myprops); return 1; } // Print out the key/value pairs char* propval; UcxMapIterator propiter = ucx_map_iterator(myprops); UCX_MAP_FOREACH(key, propval, propiter) { printf("%s = %s\n", (char*)key.data, propval); } // Don't forget to free the values before freeing the map ucx_map_free_content(myprops, NULL); ucx_map_free(myprops); fclose(file); ``` ## Stack *Header file:* [stack.h](api-2.1/stack_8h.html) *Required modules:* [Allocator](#allocator) This concrete implementation of an UCX Allocator allows you to grab some amount of memory which is then handled as a stack. Please note, that the term *stack* only refers to the behavior of this allocator. You may still choose to use either stack or heap memory for the underlying space. A typical use case is an algorithm where you need to allocate and free large amounts of memory very frequently. The following code sample shows how to initialize a stack and push and pop simple data. ```C const size_t len = 1024; char space[len]; UcxStack stack; ucx_stack_init(&stack, space, len); int i = 42; float f = 3.14f; const char* str = "Hello!"; size_t strn = 7; // push the integer ucx_stack_push(&stack, sizeof(int), &i); // push the float and rember the address float* remember = ucx_stack_push(&stack, sizeof(float), &f); // push the string with zero terminator ucx_stack_push(&stack, strn, str); // if we forget, how big an element was, we can ask the stack printf("Length of string: %zu\n", ucx_stack_topsize(&stack)-1); // retrieve the string as sstr_t, without zero terminator! sstr_t s; s.length = ucx_stack_topsize(&stack)-1; s.ptr = malloc(s.length); ucx_stack_popn(&stack, s.ptr, s.length); printf("%" PRIsstr "\n", SFMT(s)); // print the float directly from the stack and free it printf("Float: %f\n", *remember); ucx_stack_free(&stack, remember); // the last element is the integer int j; ucx_stack_pop(&stack, &j); printf("Integer: %d\n", j); ``` ## String *Header file:* [string.h](api-2.1/string_8h.html) *Required modules:* [Allocator](#allocator) This module provides a safe implementation of bounded string. Usually C strings do not carry a length. While for zero-terminated strings you can easily get the length with `strlen`, this is not generally possible for arbitrary strings. The `sstr_t` type of this module always carries the string and its length to reduce the risk of buffer overflows dramatically. ### Initialization There are several ways to create an `sstr_t`: ```C // (1) sstr() uses strlen() internally, hence cstr MUST be zero-terminated sstr_t a = sstr(cstr); // (2) cstr does not need to be zero-terminated, if length is specified sstr_t b = sstrn(cstr, len); // (3) S() macro creates sstr_t from a string using sizeof() and using sstrn(). // This version is especially useful for function arguments sstr_t c = S("hello"); // (4) SC() macro works like S(), but makes the string immutable using scstr_t. // (available since UCX 2.0) scstr_t d = SC("hello"); // (5) ST() macro creates sstr_t struct literal using sizeof() sstr_t e = ST("hello"); ``` You should not use the `S()`, `SC()`, or `ST()` macro with string of unknown origin, since the `sizeof()` call might not coincide with the string length in those cases. If you know what you are doing, it can save you some performance, because you do not need the `strlen()` call. ### Handling immutable strings *(Since: UCX 2.0)* For immutable strings (i.e. `const char*` strings), UCX provides the `scstr_t` type, which works exactly as the `sstr_t` type but with a pointer to `const char`. All UCX string functions come in two flavors: one that enforces the `scstr_t` type, and another that usually accepts both types and performs a conversion automatically, if necessary. There are some exceptions to this rule, as the return type may depend on the argument type. E.g. the `sstrchr()` function returns a substring starting at the first occurrence of the specified character. Since this substring points to the memory of the argument string, it does not accept `scstr_t` as input argument, because the return type would break the const-ness. ### Finding the position of a substring The `sstrstr()` function gives you a new `sstr_t` object starting with the requested substring. Thus determining the position comes down to a simple subtraction. ```C sstr_t haystack = ST("Here we go!"); sstr_t needle = ST("we"); sstr_t result = sstrstr(haystack, needle); if (result.ptr) printf("Found at position %zd.\n", haystack.length-result.length); else printf("Not found.\n"); ``` ### Spliting a string by a delimiter The `sstrsplit()` function (and its allocator based version `sstrsplit_a()`) is very powerful and might look a bit nasty at a first glance. But it is indeed very simple to use. It is even more convenient in combination with a memory pool. ```C sstr_t test = ST("here::are::some::strings"); sstr_t delim = ST("::"); ssize_t count = 0; // no limit UcxMempool* pool = ucx_mempool_new_default(); sstr_t* result = sstrsplit_a(pool->allocator, test, delim, &count); for (ssize_t i = 0 ; i < count ; i++) { // don't forget to specify the length via the %*s format specifier printf("%*s\n", result[i].length, result[i].ptr); } ucx_mempool_destroy(pool); ``` The output is: here are some strings The memory pool ensures, that all strings are freed. ### Disabling convenience macros If you are experiencing any troubles with the short convenience macros `S()`, `SC()`, or `ST()`, you can disable them by setting the macro `UCX_NO_SSTR_SHORTCUTS` before including the header (or via a compiler option). For the formatting macros `SFMT()` and `PRIsstr` you can use the macro `UCX_NO_SSTR_FORMAT_MACROS` to disable them. Please keep in mind, that after disabling the macros, you cannot use them in your code *and* foreign code that you might have included. You should only disable the macros, if you are experiencing a nasty name clash which cannot be otherwise resolved. ## Testing *Header file:* [test.h](api-2.1/test_8h.html) *Required modules:* None. This module provides a testing framework which allows you to execute test cases within test suites. To avoid code duplication within tests, we also provide the possibility to define test subroutines. You should declare test cases and subroutines in a header file per test unit and implement them as you would implement normal functions. ```C // myunit.h UCX_TEST(function_name); UCX_TEST_SUBROUTINE(subroutine_name, paramlist); // optional // myunit.c UCX_TEST_SUBROUTINE(subroutine_name, paramlist) { // ... reusable tests with UCX_TEST_ASSERT() ... } UCX_TEST(function_name) { // ... resource allocation and other test preparation ... // mandatory marker for the start of the tests UCX_TEST_BEGIN // ... verifications with UCX_TEST_ASSERT() ... // (and/or calls with UCX_TEST_CALL_SUBROUTINE()) // mandatory marker for the end of the tests UCX_TEST_END // ... resource cleanup ... // (all code after UCX_TEST_END is always executed) } ``` If you want to use the `UCX_TEST_ASSERT()` macro in a function, you are *required* to use a `UCX_TEST_SUBROUTINE`. Otherwise, the testing framework does not know where to jump, when the assertion fails. After implementing the tests, you can easily build a test suite and execute it: ```C UcxTestSuite* suite = ucx_test_suite_new(); ucx_test_register(suite, testMyTestCase01); ucx_test_register(suite, testMyTestCase02); // ... ucx_test_run(suite, stdout); // stdout, or any other FILE stream ``` ## Utilities *Header file:* [utils.h](api-2.1/utils_8h.html) *Required modules:* [Allocator](#allocator), [String](#string) In this module we provide very general utility function for copy and compare operations. We also provide several `printf` variants to conveniently print formatted data to streams or strings. ### A simple copy program The utilities package provides several stream copy functions. One of them has a very simple interface and can, for instance, be used to copy whole files in a single call. This is a minimal working example: ```C #include <stdio.h> #include <ucx/utils.h> int main(int argc, char** argv) { if (argc != 3) { fprintf(stderr, "Use %s <src> <dest>", argv[0]); return 1; } FILE *srcf = fopen(argv[1], "r"); // insert error handling on your own FILE *destf = fopen(argv[2], "w"); size_t n = ucx_stream_copy(srcf, destf, fread, fwrite); printf("%zu bytes copied.\n", n); fclose(srcf); fclose(destf); return 0; } ``` ### Automatic allocation for formatted strings The UCX utility function `ucx_asprintf()` and it's convenient shortcut `ucx_sprintf` allow easy formatting of strings, without ever having to worry about the required space. ```C sstr_t mystring = ucx_sprintf("The answer is: %d!", 42); ``` Still, you have to pass `mystring.ptr` to `free()` (or the free function of your allocator, if you use `ucx_asprintf`). If you don't have all the information ready to build your string, you can even use a [UcxBuffer](#buffer) as a target with the utility function `ucx_bprintf()`. ```C UcxBuffer* strbuffer = ucx_buffer_new(NULL, 512, UCX_BUFFER_AUTOEXTEND); for (unsigned int i = 2 ; i < 100 ; i++) { ucx_bprintf(strbuffer, "Integer %d is %s\n", i, prime(i) ? "prime" : "not prime"); } // print the result to stdout printf("%s", (char*)strbuffer->space); ucx_buffer_free(strbuffer); ```