[FIX] Start fixing merging issue

This commit is contained in:
lda 2023-11-30 19:08:44 +01:00
parent 829b9bd6b4
commit e8d6f0f857
2 changed files with 643 additions and 0 deletions

300
Cytoplasm/src/HttpRouter.c Normal file
View File

@ -0,0 +1,300 @@
/*
* Copyright (C) 2022-2023 Jordan Bancino <@jordan:bancino.net>
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation files
* (the "Software"), to deal in the Software without restriction,
* including without limitation the rights to use, copy, modify, merge,
* publish, distribute, sublicense, and/or sell copies of the Software,
* and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include <HttpRouter.h>
#include <Memory.h>
#include <HashMap.h>
#include <Str.h>
#include <regex.h>
#include <string.h>
#define REG_FLAGS (REG_EXTENDED)
#define REG_MAX_SUB 8
typedef struct RouteNode
{
HttpRouteFunc *exec;
HashMap *children;
regex_t regex;
} RouteNode;
struct HttpRouter
{
RouteNode *root;
};
static RouteNode *
RouteNodeCreate(char *regex, HttpRouteFunc * exec)
{
RouteNode *node;
if (!regex)
{
return NULL;
}
node = Malloc(sizeof(RouteNode));
if (!node)
{
return NULL;
}
node->children = HashMapCreate();
if (!node->children)
{
Free(node);
return NULL;
}
/* Force the regex to match the entire path part exactly. */
regex = StrConcat(3, "^", regex, "$");
if (!regex)
{
Free(node);
return NULL;
}
if (regcomp(&node->regex, regex, REG_FLAGS) != 0)
{
HashMapFree(node->children);
Free(node);
Free(regex);
return NULL;
}
node->exec = exec;
Free(regex);
return node;
}
static void
RouteNodeFree(RouteNode * node)
{
char *key;
RouteNode *val;
if (!node)
{
return;
}
while (HashMapIterate(node->children, &key, (void **) &val))
{
RouteNodeFree(val);
}
HashMapFree(node->children);
regfree(&node->regex);
Free(node);
}
HttpRouter *
HttpRouterCreate(void)
{
HttpRouter *router = Malloc(sizeof(HttpRouter));
if (!router)
{
return NULL;
}
router->root = RouteNodeCreate("/", NULL);
return router;
}
void
HttpRouterFree(HttpRouter * router)
{
if (!router)
{
return;
}
RouteNodeFree(router->root);
Free(router);
}
int
HttpRouterAdd(HttpRouter * router, char *regPath, HttpRouteFunc * exec)
{
RouteNode *node;
char *pathPart;
char *tmp;
if (!router || !regPath || !exec)
{
return 0;
}
if (StrEquals(regPath, "/"))
{
router->root->exec = exec;
return 1;
}
regPath = StrDuplicate(regPath);
if (!regPath)
{
return 0;
}
tmp = regPath;
node = router->root;
while ((pathPart = strtok_r(tmp, "/", &tmp)))
{
RouteNode *tNode = HashMapGet(node->children, pathPart);
if (!tNode)
{
tNode = RouteNodeCreate(pathPart, NULL);
RouteNodeFree(HashMapSet(node->children, pathPart, tNode));
}
node = tNode;
}
node->exec = exec;
Free(regPath);
return 1;
}
int
HttpRouterRoute(HttpRouter * router, char *path, void *args, void **ret)
{
RouteNode *node;
char *pathPart;
char *tmp;
HttpRouteFunc *exec = NULL;
Array *matches = NULL;
size_t i;
int retval;
if (!router || !path)
{
return 0;
}
matches = ArrayCreate();
if (!matches)
{
return 0;
}
node = router->root;
if (StrEquals(path, "/"))
{
exec = node->exec;
}
else
{
path = StrDuplicate(path);
tmp = path;
while ((pathPart = strtok_r(tmp, "/", &tmp)))
{
char *key;
RouteNode *val = NULL;
regmatch_t pmatch[REG_MAX_SUB];
i = 0;
while (HashMapIterateReentrant(node->children, &key, (void **) &val, &i))
{
if (regexec(&val->regex, pathPart, REG_MAX_SUB, pmatch, 0) == 0)
{
break;
}
val = NULL;
}
if (!val)
{
exec = NULL;
break;
}
node = val;
exec = node->exec;
/* If we want to pass an arg, the match must be in parens */
if (val->regex.re_nsub)
{
/* pmatch[0] is the whole string, not the first
* subexpression */
char * substr;
regmatch_t cpmatch;
for (i = 1; i < REG_MAX_SUB; i++)
{
cpmatch = pmatch[i];
substr = StrSubstr(pathPart, cpmatch.rm_so, cpmatch.rm_eo);
if (pmatch[i].rm_so == -1)
{
break;
}
ArrayAdd(matches, substr);
}
}
}
Free(path);
}
if (!exec)
{
retval = 0;
goto finish;
}
if (ret)
{
*ret = exec(matches, args);
}
else
{
exec(matches, args);
}
retval = 1;
finish:
for (i = 0; i < ArraySize(matches); i++)
{
Free(ArrayGet(matches, i));
}
ArrayFree(matches);
return retval;
}

343
Cytoplasm/src/Str.c Normal file
View File

@ -0,0 +1,343 @@
/*
* Copyright (C) 2022-2023 Jordan Bancino <@jordan:bancino.net>
*
* Permission is hereby granted, free of charge, to any person
* obtaining a copy of this software and associated documentation files
* (the "Software"), to deal in the Software without restriction,
* including without limitation the rights to use, copy, modify, merge,
* publish, distribute, sublicense, and/or sell copies of the Software,
* and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include <Str.h>
#include <Memory.h>
#include <Util.h>
#include <Rand.h>
#include <Int.h>
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include <stdarg.h>
#include <pthread.h>
#include <unistd.h>
UInt32
StrUtf16Decode(UInt16 high, UInt16 low)
{
if (high <= 0xD7FF)
{
return high;
}
else if (high <= 0xDBFF)
{
unsigned short hS = (high - 0xD800) * 0x400;
unsigned short lS = low - 0xDC00;
return (lS | hS) + 0x10000;
}
else
{
return 0;
}
}
char *
StrUtf8Encode(UInt32 codepoint)
{
char *str;
str = Malloc(5 * sizeof(char));
if (!str)
{
return NULL;
}
if (codepoint <= 0x7F && codepoint != 0) /* Plain ASCII */
{
str[0] = (char) codepoint;
str[1] = '\0';
}
else if (codepoint <= 0x07FF) /* 2-byte */
{
str[0] = (char) (((codepoint >> 6) & 0x1F) | 0xC0);
str[1] = (char) (((codepoint >> 0) & 0x3F) | 0x80);
str[2] = '\0';
}
else if (codepoint <= 0xFFFF) /* 3-byte */
{
str[0] = (char) (((codepoint >> 12) & 0x0F) | 0xE0);
str[1] = (char) (((codepoint >> 6) & 0x3F) | 0x80);
str[2] = (char) (((codepoint >> 0) & 0x3F) | 0x80);
str[3] = '\0';
}
else if (codepoint <= 0x10FFFF)/* 4-byte */
{
str[0] = (char) (((codepoint >> 18) & 0x07) | 0xF0);
str[1] = (char) (((codepoint >> 12) & 0x3F) | 0x80);
str[2] = (char) (((codepoint >> 6) & 0x3F) | 0x80);
str[3] = (char) (((codepoint >> 0) & 0x3F) | 0x80);
str[4] = '\0';
}
else
{
/* Send replacement character */
str[0] = (char) 0xEF;
str[1] = (char) 0xBF;
str[2] = (char) 0xBD;
str[3] = '\0';
}
return str;
}
char *
StrDuplicate(const char *inStr)
{
size_t len;
char *outStr;
if (!inStr)
{
return NULL;
}
len = strlen(inStr);
outStr = Malloc(len + 1); /* For the null terminator */
if (!outStr)
{
return NULL;
}
strncpy(outStr, inStr, len + 1);
return outStr;
}
char *
StrSubstr(const char *inStr, size_t start, size_t end)
{
size_t len;
size_t i;
size_t j;
char *outStr;
if (!inStr)
{
return NULL;
}
if (start >= end)
{
return NULL;
}
len = end - start;
outStr = Malloc(len + 1);
if (!outStr)
{
return NULL;
}
j = 0;
for (i = start; i < end; i++)
{
if (inStr[i] == '\0')
{
break;
}
outStr[j] = inStr[i];
j++;
}
outStr[j] = '\0';
return outStr;
}
char *
StrConcat(size_t nStr,...)
{
va_list argp;
char *str;
char *strp;
size_t strLen = 0;
size_t i;
va_start(argp, nStr);
for (i = 0; i < nStr; i++)
{
char *argStr = va_arg(argp, char *);
if (argStr)
{
strLen += strlen(argStr);
}
}
va_end(argp);
str = Malloc(strLen + 1);
strp = str;
va_start(argp, nStr);
for (i = 0; i < nStr; i++)
{
/* Manually copy chars instead of using strcopy() so we don't
* have to call strlen() on the strings again, and we aren't
* writing useless null chars. */
char *argStr = va_arg(argp, char *);
if (argStr)
{
while (*argStr)
{
*strp = *argStr;
strp++;
argStr++;
}
}
}
va_end(argp);
str[strLen] = '\0';
return str;
}
int
StrBlank(const char *str)
{
int blank = 1;
size_t i = 0;
while (str[i])
{
blank &= isspace((unsigned char) str[i]);
/* No need to continue if we don't have a blank */
if (!blank)
{
break;
}
i++;
}
return blank;
}
char *
StrRandom(size_t len)
{
static const char charset[] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ";
char *str;
int *nums;
size_t i;
if (!len)
{
return NULL;
}
str = Malloc(len + 1);
if (!str)
{
return NULL;
}
nums = Malloc(len * sizeof(int));
if (!nums)
{
Free(str);
return NULL;
}
/* TODO: This seems slow. */
RandIntN(nums, len, sizeof(charset) - 1);
for (i = 0; i < len; i++)
{
str[i] = charset[nums[i]];
}
Free(nums);
str[len] = '\0';
return str;
}
char *
StrInt(long i)
{
char *str;
int len;
len = snprintf(NULL, 0, "%ld", i);
str = Malloc(len + 1);
if (!str)
{
return NULL;
}
snprintf(str, len + 1, "%ld", i);
return str;
}
char *
StrLower(char *str)
{
char *ret;
size_t len;
size_t i;
if (!str)
{
return NULL;
}
len = strlen(str);
ret = Malloc(len + 1);
for (i = 0; i < len; i++)
{
ret[i] = tolower(str[i]);
}
ret[len] = '\0';
return ret;
}
int
StrEquals(const char *str1, const char *str2)
{
/* Both strings are NULL, they're equal */
if (!str1 && !str2)
{
return 1;
}
/* One or the other is NULL, they're not equal */
if (!str1 || !str2)
{
return 0;
}
/* Neither are NULL, do a regular string comparison */
return strcmp(str1, str2) == 0;
}