I have been meddling with using chatGPT prompts to program in C.
Using incremental prompts it is possible to get chatGPT 3.5 to do a lot more than expected.
I wonder about code quality and would enjoy feedback both on prompts and code.
thread safe code
static path
dynamic path allocation for larger path string
Standard practice for Linux / MacOs is to dynamically allocate (4095 / 1024 chart path limit), whereas Windows (260 path limit) is to statically allocate.
I incrementally build up the query, otherwise chatGPT tends to unlearn previous request changes.
In the end this looks like some crude form of meta coding where it is possible to coerce the LLM to generate the desired code.
_query LLM perplexity.ai Modify code example Portable Path Parsing and Validation Code in C use thread safe function use function pointer strtok_s() for Windows strtok_r() for linux allocate INITIAL_BUFFER_SIZE 60 char static path_copy declare pointer pfinal_path_copy initialize to path_copy when necessary dynamically allocate final_path_copy pfinal_path_copy = final_path_copy use single strcpy(pfinal_path_copy, path) function
Code is shaped one feature at a time to match the desired style
modify example replace 60 by equate
Some prompts remained sticky without being repeated. "no WSL cygwin" was added early on to prompt for Windows OS specific code, then dropped.
write code example parse path in portable way linux windows MacOS no WSL cygwin
Resulting code compiles and runs on the first go.
#include <stdio.h> #include <stdlib.h> #include <string.h> #ifdef _WIN32 #include <io.h> // For _access() on Windows #include <windows.h> // For Windows API #else #include <unistd.h> // For access() on Unix-like systems (Linux, macOS) #include <sys/stat.h> #endif // Define a constant for the initial buffer size #define INITIAL_BUFFER_SIZE 60 // Function to check if a path is valid int path_validate(const char *path) { #ifdef _WIN32 return _access(path, 0) == 0; // 0 means the file exists #else struct stat buffer; return (stat(path, &buffer) == 0); // Returns 1 if valid, 0 if invalid #endif } // Type definition for tokenization function pointer typedef char* (*tokenizer_func)(char*, const char*, char**); // Function to parse the path into components void parse_path(const char *path) { char path_copy[INITIAL_BUFFER_SIZE]; // Static buffer for initial use char *final_path_copy = NULL; // Pointer for dynamic allocation char *pfinal_path_copy = path_copy; // Pointer to point to either path_copy or final_path_copy // Check if the input path length exceeds INITIAL_BUFFER_SIZE if (strlen(path) + 1 > INITIAL_BUFFER_SIZE) { final_path_copy = malloc(strlen(path) + 1); // Dynamically allocate memory if (final_path_copy == NULL) { fprintf(stderr, "Memory allocation failed.\n"); exit(EXIT_FAILURE); } pfinal_path_copy = final_path_copy; // Point to dynamically allocated memory } // Use a single strcpy function to copy the path into the appropriate buffer strcpy(pfinal_path_copy, path); // Copy full path into the selected buffer // Determine the separator based on the platform and assign the tokenizer function const char *separator; tokenizer_func tokenizer; char *saveptr; #ifdef _WIN32 separator = "\"; tokenizer = strtok_s; // Use strtok_s for Windows #else separator = "/"; tokenizer = strtok_r; // Use strtok_r for Linux #endif // Tokenize using the assigned tokenizer function char *token = tokenizer(pfinal_path_copy, separator, &saveptr); while (token != NULL) { printf("Path component: %s\n", token); token = tokenizer(NULL, separator, &saveptr); // Continue tokenizing } free(final_path_copy); // Free dynamically allocated memory after use (if allocated) } int main(int argc, char *argv[]) { if (argc < 2) { fprintf(stderr, "Usage: %s <path>\n", argv[0]); return EXIT_FAILURE; } const char *path = argv[1]; // Validate the path using the new function name if (path_validate(path)) { printf("The path '%s' is valid.\n", path); printf("Parsing components:\n"); parse_path(path); } else { printf("The path '%s' is invalid.\n", path); } return EXIT_SUCCESS; } ## Unit test Was created but not included here ## Performance The entire exercise took 2:30 hour. I have not tested the code on Windows or MacOS. I'd like some feedback on code quality.
The above is the detailed content of chatGPT - C programming Linux Windows cross-platform - code review request. For more information, please follow other related articles on the PHP Chinese website!