+
+/** @brief Substitutes a char for another in a string
+ *
+ * @param str the string to modify
+ * @param from char to search
+ * @param to char to put instead
+ * @param amount amount of changes to do (=0 means all)
+ */
+void xbt_str_subst(char *str, char from, char to, int occurence)
+{
+ char *p = str;
+ while (*p != '\0') {
+ if (*p == from) {
+ *p = to;
+ if (occurence == 1)
+ return;
+ occurence--;
+ }
+ p++;
+ }
+}
+
+/** @brief Replaces a set of variables by their values
+ *
+ * @param str where to apply the change
+ * @param patterns what to change
+ * @return The string modified
+ *
+ * Check xbt_strbuff_varsubst() for more details, and remember that the string may be reallocated (moved) in the process.
+ */
+
+char *xbt_str_varsubst(char *str, xbt_dict_t patterns)
+{
+ xbt_strbuff_t buff = xbt_strbuff_new_from(str);
+ char *res;
+ xbt_strbuff_varsubst(buff, patterns);
+ res = buff->data;
+ xbt_strbuff_free_container(buff);
+ return res;
+}
+
+
+/** @brief Splits a string into a dynar of strings
+ *
+ * @param s: the string to split
+ * @param sep: a string of all chars to consider as separator.
+ *
+ * By default (with sep=NULL), these characters are used as separator:
+ *
+ * - " " (ASCII 32 (0x20)) space.
+ * - "\t" (ASCII 9 (0x09)) tab.
+ * - "\n" (ASCII 10 (0x0A)) line feed.
+ * - "\r" (ASCII 13 (0x0D)) carriage return.
+ * - "\0" (ASCII 0 (0x00)) NULL.
+ * - "\x0B" (ASCII 11 (0x0B)) vertical tab.
+ */
+
+xbt_dynar_t xbt_str_split(const char *s, const char *sep)
+{
+ xbt_dynar_t res = xbt_dynar_new(sizeof(char *), xbt_free_ref);
+ const char *p, *q;
+ int done;
+ const char *sep_dflt = " \t\n\r\x0B";
+ char is_sep[256] = { 1, 0 };
+
+ /* check what are the separators */
+ memset(is_sep, 0, sizeof(is_sep));
+ if (!sep) {
+ while (*sep_dflt)
+ is_sep[(unsigned char) *sep_dflt++] = 1;
+ } else {
+ while (*sep)
+ is_sep[(unsigned char) *sep++] = 1;
+ }
+ is_sep[0] = 1; /* End of string is also separator */
+
+ /* Do the job */
+ p = q = s;
+ done = 0;
+
+ if (s[0] == '\0')
+ return res;
+
+ while (!done) {
+ char *topush;
+ while (!is_sep[(unsigned char) *q]) {
+ q++;
+ }
+ if (*q == '\0')
+ done = 1;
+
+ topush = xbt_malloc(q - p + 1);
+ memcpy(topush, p, q - p);
+ topush[q - p] = '\0';
+ xbt_dynar_push(res, &topush);
+ p = ++q;
+ }
+
+ return res;
+}
+
+/**
+ * \brief This functions splits a string after using another string as separator
+ * For example A!!B!!C splitted after !! will return the dynar {A,B,C}
+ * \return An array of dynars containing the string tokens
+ */
+xbt_dynar_t xbt_str_split_str(const char *s, const char *sep)
+{
+ xbt_dynar_t res = xbt_dynar_new(sizeof(char *), xbt_free_ref);
+ int done;
+ const char *p, *q;
+
+ p = q = s;
+ done = 0;
+
+ if (s[0] == '\0')
+ return res;
+ if (sep[0] == '\0') {
+ s = xbt_strdup(s);
+ xbt_dynar_push(res, &s);
+ return res;
+ }
+
+ while (!done) {
+ char *to_push;
+ int v = 0;
+ //get the start of the first occurence of the substring
+ q = strstr(p, sep);
+ //if substring was not found add the entire string
+ if (NULL == q) {
+ v = strlen(p);
+ to_push = malloc(v + 1);
+ memcpy(to_push, p, v);
+ to_push[v] = '\0';
+ xbt_dynar_push(res, &to_push);
+ done = 1;
+ } else {
+ //get the appearance
+ to_push = malloc(q - p + 1);
+ memcpy(to_push, p, q - p);
+ //add string terminator
+ to_push[q - p] = '\0';
+ xbt_dynar_push(res, &to_push);
+ p = q + strlen(sep);
+ }
+ }
+ return res;
+}
+
+/** @brief Splits a string into a dynar of strings, taking quotes into account
+ *
+ * It basically does the same argument separation than the shell, where white
+ * spaces can be escaped and where arguments are never splitted within a
+ * quote group.
+ * Several subsequent spaces are ignored (unless within quotes, of course).
+ *
+ */
+
+xbt_dynar_t xbt_str_split_quoted(const char *s)
+{
+ xbt_dynar_t res = xbt_dynar_new(sizeof(char *), xbt_free_ref);
+ char *str_to_free; /* we have to copy the string before, to handle backslashes */
+ char *beg, *end; /* pointers around the parsed chunk */
+ int in_simple_quote = 0, in_double_quote = 0;
+ int done = 0;
+ int ctn = 0; /* Got something in this block */
+
+ if (s[0] == '\0')
+ return res;
+ beg = str_to_free = xbt_strdup(s);
+
+ /* trim leading spaces */
+ xbt_str_ltrim(beg, " ");
+ end = beg;
+
+ while (!done) {
+
+
+ switch (*end) {
+ case '\\':
+ ctn = 1;
+ /* Protected char; move it closer */
+ memmove(end, end + 1, strlen(end));
+ if (*end == '\0')
+ THROW0(arg_error, 0, "String ends with \\");
+ end++; /* Pass the protected char */
+ break;
+
+ case '\'':
+ ctn = 1;
+ if (!in_double_quote) {
+ in_simple_quote = !in_simple_quote;
+ memmove(end, end + 1, strlen(end));
+ } else {
+ /* simple quote protected by double ones */
+ end++;
+ }
+ break;
+ case '"':
+ ctn = 1;
+ if (!in_simple_quote) {
+ in_double_quote = !in_double_quote;
+ memmove(end, end + 1, strlen(end));
+ } else {
+ /* double quote protected by simple ones */
+ end++;
+ }
+ break;
+
+ case ' ':
+ case '\t':
+ case '\n':
+ case '\0':
+ if (*end == '\0' && (in_simple_quote || in_double_quote)) {
+ THROW2(arg_error, 0,
+ "End of string found while searching for %c in %s",
+ (in_simple_quote ? '\'' : '"'), s);
+ }
+ if (in_simple_quote || in_double_quote) {
+ end++;
+ } else {
+ if (ctn) {
+ /* Found a separator. Push the string if contains something */
+ char *topush = xbt_malloc(end - beg + 1);
+ memcpy(topush, beg, end - beg);
+ topush[end - beg] = '\0';
+ xbt_dynar_push(res, &topush);
+ }
+ ctn = 0;
+
+ if (*end == '\0') {
+ done = 1;
+ break;
+ }
+
+ beg = ++end;
+ xbt_str_ltrim(beg, " ");
+ end = beg;
+ }
+ break;
+
+ default:
+ ctn = 1;
+ end++;
+ }
+ }
+ free(str_to_free);
+ return res;
+}
+
+#ifdef SIMGRID_TEST
+#include "xbt/str.h"
+
+#define mytest(name, input, expected) \
+ xbt_test_add0(name); \
+ d=xbt_str_split_quoted(input); \
+ s=xbt_str_join(d,"XXX"); \
+ xbt_test_assert3(!strcmp(s,expected),\
+ "Input (%s) leads to (%s) instead of (%s)", \
+ input,s,expected);\
+ free(s); \
+ xbt_dynar_free(&d);
+
+XBT_TEST_SUITE("xbt_str", "String Handling");
+XBT_TEST_UNIT("xbt_str_split_quoted", test_split_quoted,"test the function xbt_str_split_quoted")
+{
+ xbt_dynar_t d;
+ char *s;
+
+ mytest("Empty", "", "");
+ mytest("Basic test", "toto tutu", "totoXXXtutu");
+ mytest("Useless backslashes", "\\t\\o\\t\\o \\t\\u\\t\\u", "totoXXXtutu");
+ mytest("Protected space", "toto\\ tutu", "toto tutu");
+ mytest("Several spaces", "toto tutu", "totoXXXtutu");
+ mytest("LTriming", " toto tatu", "totoXXXtatu");
+ mytest("Triming", " toto tutu ", "totoXXXtutu");
+ mytest("Single quotes", "'toto tutu' tata", "toto tutuXXXtata");
+ mytest("Double quotes", "\"toto tutu\" tata", "toto tutuXXXtata");
+ mytest("Mixed quotes", "\"toto' 'tutu\" tata", "toto' 'tutuXXXtata");
+ mytest("Backslashed quotes", "\\'toto tutu\\' tata",
+ "'totoXXXtutu'XXXtata");
+ mytest("Backslashed quotes + quotes", "'toto \\'tutu' tata",
+ "toto 'tutuXXXtata");
+
+}
+
+#define mytest_str(name, input, separator, expected) \
+ xbt_test_add0(name); \
+ d=xbt_str_split_str(input, separator); \
+ s=xbt_str_join(d,"XXX"); \
+ xbt_test_assert3(!strcmp(s,expected),\
+ "Input (%s) leads to (%s) instead of (%s)", \
+ input,s,expected);\
+ free(s); \
+ xbt_dynar_free(&d);
+
+XBT_TEST_UNIT("xbt_str_split_str", test_split_str,"test the function xbt_str_split_str")
+{
+ xbt_dynar_t d;
+ char *s;
+
+ mytest_str("Empty string and separator", "", "", "");
+ mytest_str("Empty string", "", "##", "");
+ mytest_str("Empty separator", "toto", "", "toto");
+ mytest_str("String with no separator in it", "toto", "##", "toto");
+ mytest_str("Basic test", "toto##tutu", "##", "totoXXXtutu");
+}
+#endif /* SIMGRID_TEST */
+
+/** @brief Join a set of strings as a single string */
+
+char *xbt_str_join(xbt_dynar_t dyn, const char *sep)
+{
+ int len = 1, dyn_len = xbt_dynar_length(dyn);
+ unsigned int cpt;
+ char *cursor;
+ char *res, *p;
+
+ if (!dyn_len)
+ return xbt_strdup("");
+
+ /* compute the length */
+ xbt_dynar_foreach(dyn, cpt, cursor) {
+ len += strlen(cursor);
+ }
+ len += strlen(sep) * dyn_len;
+ /* Do the job */
+ res = xbt_malloc(len);
+ p = res;
+ xbt_dynar_foreach(dyn, cpt, cursor) {
+ if ((int) cpt < dyn_len - 1)
+ p += sprintf(p, "%s%s", cursor, sep);
+ else
+ p += sprintf(p, "%s", cursor);
+ }
+ return res;
+}
+
+#if !defined(HAVE_GETLINE) || defined(DOXYGEN)
+/* prototype here, just in case */
+long getline(char **buf, size_t * n, FILE * stream);
+
+/** @brief Get a single line from the stream (reimplementation of the GNU getline)
+ *
+ * This is a redefinition of the GNU getline function, used on platforms where it does not exists.
+ *
+ * getline() reads an entire line from stream, storing the address of the buffer
+ * containing the text into *buf. The buffer is null-terminated and includes
+ * the newline character, if one was found.
+ *
+ * If *buf is NULL, then getline() will allocate a buffer for storing the line,
+ * which should be freed by the user program. Alternatively, before calling getline(),
+ * *buf can contain a pointer to a malloc()-allocated buffer *n bytes in size. If the buffer
+ * is not large enough to hold the line, getline() resizes it with realloc(), updating *buf and *n
+ * as necessary. In either case, on a successful call, *buf and *n will be updated to
+ * reflect the buffer address and allocated size respectively.
+ */
+long getline(char **buf, size_t * n, FILE * stream)
+{
+
+ size_t i;
+ int ch;
+
+ if (!*buf) {
+ *buf = xbt_malloc(512);
+ *n = 512;
+ }
+
+ if (feof(stream))
+ return (ssize_t) - 1;
+
+ for (i = 0; (ch = fgetc(stream)) != EOF; i++) {
+
+ if (i >= (*n) + 1)
+ *buf = xbt_realloc(*buf, *n += 512);
+
+ (*buf)[i] = ch;
+
+ if ((*buf)[i] == '\n') {
+ i++;
+ (*buf)[i] = '\0';
+ break;
+ }
+ }
+
+ if (i == *n)
+ *buf = xbt_realloc(*buf, *n += 1);
+
+ (*buf)[i] = '\0';
+
+ return (ssize_t) i;