[paraslash.git] / string.c

/*
 * Copyright (C) 2004-2012 Andre Noll <maan@systemlinux.org>
 *
 * Licensed under the GPL v2. For licencing details see COPYING.
 */

/** \file string.c Memory allocation and string handling functions. */

#include <sys/time.h> /* gettimeofday */
#include <pwd.h>
#include <sys/utsname.h> /* uname() */
#include <string.h>
#include <regex.h>

#include "para.h"
#include "string.h"
#include "error.h"

/**
 * Paraslash's version of realloc().
 *
 * \param p Pointer to the memory block, may be \p NULL.
 * \param size The desired new size.
 *
 * A wrapper for realloc(3). It calls \p exit(\p EXIT_FAILURE) on errors,
 * i.e. there is no need to check the return value in the caller.
 *
 * \return A pointer to  the newly allocated memory, which is suitably aligned
 * for any kind of variable and may be different from \a p.
 *
 * \sa realloc(3).
 */
__must_check __malloc void *para_realloc(void *p, size_t size)
{
        /*
         * No need to check for NULL pointers: If p is NULL, the call
         * to realloc is equivalent to malloc(size)
         */
        assert(size);
        if (!(p = realloc(p, size))) {
                PARA_EMERG_LOG("realloc failed (size = %zu), aborting\n",
                        size);
                exit(EXIT_FAILURE);
        }
        return p;
}

/**
 * Paraslash's version of malloc().
 *
 * \param size The desired new size.
 *
 * A wrapper for malloc(3) which exits on errors.
 *
 * \return A pointer to the allocated memory, which is suitably aligned for any
 * kind of variable.
 *
 * \sa malloc(3).
 */
__must_check __malloc void *para_malloc(size_t size)
{
        void *p;

        assert(size);
        p = malloc(size);
        if (!p) {
                PARA_EMERG_LOG("malloc failed (size = %zu), aborting\n",
                        size);
                exit(EXIT_FAILURE);
        }
        return p;
}

/**
 * Paraslash's version of calloc().
 *
 * \param size The desired new size.
 *
 * A wrapper for calloc(3) which exits on errors.
 *
 * \return A pointer to the allocated and zeroed-out memory, which is suitably
 * aligned for any kind of variable.
 *
 * \sa calloc(3)
 */
__must_check __malloc void *para_calloc(size_t size)
{
        void *ret = para_malloc(size);

        memset(ret, 0, size);
        return ret;
}

/**
 * Paraslash's version of strdup().
 *
 * \param s The string to be duplicated.
 *
 * A wrapper for strdup(3). It calls \p exit(EXIT_FAILURE) on errors, i.e.
 * there is no need to check the return value in the caller.
 *
 * \return A pointer to the duplicated string. If \a s was the \p NULL pointer,
 * an pointer to an empty string is returned.
 *
 * \sa strdup(3)
 */
__must_check __malloc char *para_strdup(const char *s)
{
        char *ret;

        if ((ret = strdup(s? s: "")))
                return ret;
        PARA_EMERG_LOG("strdup failed, aborting\n");
        exit(EXIT_FAILURE);
}

/**
 * Allocate a sufficiently large string and print into it.
 *
 * \param fmt A usual format string.
 *
 * Produce output according to \p fmt. No artificial bound on the length of the
 * resulting string is imposed.
 *
 * \return This function either returns a pointer to a string that must be
 * freed by the caller or aborts without returning.
 *
 * \sa printf(3).
 */
__must_check __printf_1_2 __malloc char *make_message(const char *fmt, ...)
{
        char *msg;

        PARA_VSPRINTF(fmt, msg);
        return msg;
}

/**
 * Free the content of a pointer and set it to \p NULL.
 *
 * This is equivalent to "free(*arg); *arg = NULL;".
 *
 * \param arg The pointer whose content should be freed.
 */
void freep(void *arg)
{
        void **ptr = (void **)arg;
        free(*ptr);
        *ptr = NULL;
}

/**
 * Paraslash's version of strcat().
 *
 * \param a String to be appended to.
 * \param b String to append.
 *
 * Append \p b to \p a.
 *
 * \return If \a a is \p NULL, return a pointer to a copy of \a b, i.e.
 * para_strcat(NULL, b) is equivalent to para_strdup(b). If \a b is \p NULL,
 * return \a a without making a copy of \a a.  Otherwise, construct the
 * concatenation \a c, free \a a (but not \a b) and return \a c.
 *
 * \sa strcat(3)
 */
__must_check __malloc char *para_strcat(char *a, const char *b)
{
        char *tmp;

        if (!a)
                return para_strdup(b);
        if (!b)
                return a;
        tmp = make_message("%s%s", a, b);
        free(a);
        return tmp;
}

/**
 * Paraslash's version of dirname().
 *
 * \param name Pointer to the full path.
 *
 * Compute the directory component of \p name.
 *
 * \return If \a name is \p NULL or the empty string, return \p NULL.
 * Otherwise, Make a copy of \a name and return its directory component. Caller
 * is responsible to free the result.
 */
__must_check __malloc char *para_dirname(const char *name)
{
        char *p, *ret;

        if (!name || !*name)
                return NULL;
        ret = para_strdup(name);
        p = strrchr(ret, '/');
        if (!p)
                *ret = '\0';
        else
                *p = '\0';
        return ret;
}

/**
 * Paraslash's version of basename().
 *
 * \param name Pointer to the full path.
 *
 * Compute the filename component of \a name.
 *
 * \return \p NULL if (a) \a name is the empty string or \p NULL, or (b) name
 * ends with a slash.  Otherwise, a pointer within \a name is returned.  Caller
 * must not free the result.
 */
__must_check char *para_basename(const char *name)
{
        char *ret;

        if (!name || !*name)
                return NULL;
        ret = strrchr(name, '/');
        if (!ret)
                return (char *)name;
        ret++;
        return ret;
}

/**
 * Cut trailing newline.
 *
 * \param buf The string to be chopped.
 *
 * Replace the last character in \p buf by zero if it is equal to
 * the newline character.
 */
void chop(char *buf)
{
        int n = strlen(buf);

        if (!n)
                return;
        if (buf[n - 1] == '\n')
                buf[n - 1] = '\0';
}

/**
 * Get the logname of the current user.
 *
 * \return A dynamically allocated string that must be freed by the caller. On
 * errors, the string "unknown_user" is returned, i.e. this function never
 * returns \p NULL.
 *
 * \sa getpwuid(3).
 */
__must_check __malloc char *para_logname(void)
{
        struct passwd *pw = getpwuid(getuid());
        return para_strdup(pw? pw->pw_name : "unknown_user");
}

/**
 * Get the home directory of the current user.
 *
 * \return A dynamically allocated string that must be freed by the caller. If
 * the home directory could not be found, this function returns "/tmp".
 */
__must_check __malloc char *para_homedir(void)
{
        struct passwd *pw = getpwuid(getuid());
        return para_strdup(pw? pw->pw_dir : "/tmp");
}

/**
 * Get the own hostname.
 *
 * \return A dynamically allocated string containing the hostname.
 *
 * \sa uname(2).
 */
__malloc char *para_hostname(void)
{
        struct utsname u;

        uname(&u);
        return para_strdup(u.nodename);
}

/**
 * Used to distinguish between read-only and read-write mode.
 *
 * \sa for_each_line(), for_each_line_ro().
 */
enum for_each_line_modes{
        /** Activate read-only mode. */
        LINE_MODE_RO,
        /** Activate read-write mode. */
        LINE_MODE_RW
};

static int for_each_complete_line(enum for_each_line_modes mode, char *buf,
                size_t size, line_handler_t *line_handler, void *private_data)
{
        char *start = buf, *end;
        int ret, i, num_lines = 0;

//      PARA_NOTICE_LOG("buf: %s\n", buf);
        while (start < buf + size) {
                char *next_null;
                char *next_cr;

                next_cr = memchr(start, '\n', buf + size - start);
                next_null = memchr(start, '\0', buf + size - start);
                if (!next_cr && !next_null)
                        break;
                if (next_cr && next_null) {
                        end = next_cr < next_null? next_cr : next_null;
                } else if (next_null) {
                        end = next_null;
                } else
                        end = next_cr;
                num_lines++;
                if (!line_handler) {
                        start = ++end;
                        continue;
                }
                if (mode == LINE_MODE_RO) {
                        size_t s = end - start;
                        char *b = para_malloc(s + 1);
                        memcpy(b, start, s);
                        b[s] = '\0';
//                      PARA_NOTICE_LOG("b: %s, start: %s\n", b, start);
                        ret = line_handler(b, private_data);
                        free(b);
                } else {
                        *end = '\0';
                        ret = line_handler(start, private_data);
                }
                if (ret < 0)
                        return ret;
                start = ++end;
        }
        if (!line_handler || mode == LINE_MODE_RO)
                return num_lines;
        i = buf + size - start;
        if (i && i != size)
                memmove(buf, start, i);
        return i;
}

/**
 * Call a custom function for each complete line.
 *
 * \param buf The buffer containing data separated by newlines.
 * \param size The number of bytes in \a buf.
 * \param line_handler The custom function.
 * \param private_data Pointer passed to \a line_handler.
 *
 * If \p line_handler is \p NULL, the function returns the number of complete
 * lines in \p buf.  Otherwise, \p line_handler is called for each complete
 * line in \p buf.  The first argument to \p line_handler is the current line,
 * and \p private_data is passed as the second argument.  The function returns
 * if \p line_handler returns a negative value or no more lines are in the
 * buffer.  The rest of the buffer (last chunk containing an incomplete line)
 * is moved to the beginning of the buffer.
 *
 * \return If \p line_handler is not \p NULL, this function returns the number
 * of bytes not handled to \p line_handler on success, or the negative return
 * value of the \p line_handler on errors.
 *
 * \sa for_each_line_ro().
 */
int for_each_line(char *buf, size_t size, line_handler_t *line_handler,
                void *private_data)
{
        return for_each_complete_line(LINE_MODE_RW, buf, size, line_handler,
                private_data);
}

/**
 * Call a custom function for each complete line.
 *
 * \param buf Same meaning as in \p for_each_line().
 * \param size Same meaning as in \p for_each_line().
 * \param line_handler Same meaning as in \p for_each_line().
 * \param private_data Same meaning as in \p for_each_line().
 *
 * This function behaves like \p for_each_line(), but \a buf is left unchanged.
 *
 * \return On success, the function returns the number of complete lines in \p
 * buf, otherwise the (negative) return value of \p line_handler is returned.
 *
 * \sa for_each_line().
 */
int for_each_line_ro(char *buf, size_t size, line_handler_t *line_handler,
                void *private_data)
{
        return for_each_complete_line(LINE_MODE_RO, buf, size, line_handler,
                private_data);
}

/** Return the hex characters of the lower 4 bits. */
#define hex(a) (hexchar[(a) & 15])

static void write_size_header(char *buf, int n)
{
        static char hexchar[] = "0123456789abcdef";

        buf[0] = hex(n >> 12);
        buf[1] = hex(n >> 8);
        buf[2] = hex(n >> 4);
        buf[3] = hex(n);
        buf[4] = ' ';
}

/**
 * Read a four-byte hex-number and return its value.
 *
 * Each status item sent by para_server is prefixed with such a hex number in
 * ASCII which describes the size of the status item.
 *
 * \param buf The buffer which must be at least four bytes long.
 *
 * \return The value of the hex number on success, \p -E_SIZE_PREFIX if the
 * buffer did not contain only hex digits.
 */
int read_size_header(const char *buf)
{
        int i, len = 0;

        for (i = 0; i < 4; i++) {
                unsigned char c = buf[i];
                len <<= 4;
                if (c >= '0' && c <= '9') {
                        len += c - '0';
                        continue;
                }
                if (c >= 'a' && c <= 'f') {
                        len += c - 'a' + 10;
                        continue;
                }
                return -E_SIZE_PREFIX;
        }
        if (buf[4] != ' ')
                return -E_SIZE_PREFIX;
        return len;
}

/**
 * Safely print into a buffer at a given offset.
 *
 * \param b Determines the buffer, its size, and the offset.
 * \param fmt The format string.
 *
 * This function prints into the buffer given by \a b at the offset which is
 * also given by \a b. If there is not enough space to hold the result, the
 * buffer size is doubled until the underlying call to vsnprintf() succeeds
 * or the size of the buffer exceeds the maximal size specified in \a b.
 *
 * In the latter case the unmodified \a buf and \a offset values as well as the
 * private_data pointer of \a b are passed to the \a max_size_handler of \a b.
 * If this function succeeds, i.e. returns a non-negative value, the offset of
 * \a b is reset to zero and the given data is written to the beginning of the
 * buffer. If \a max_size_handler() returns a negative value, this value is
 * returned by \a para_printf().
 *
 * Upon return, the offset of \a b is adjusted accordingly so that subsequent
 * calls to this function append data to what is already contained in the
 * buffer.
 *
 * It's OK to call this function with \p b->buf being \p NULL. In this case, an
 * initial buffer is allocated.
 *
 * \return The number of bytes printed into the buffer (not including the
 * terminating \p NULL byte) on success, negative on errors. If there is no
 * size-bound on \a b, i.e. if \p b->max_size is zero, this function never
 * fails.
 *
 * \sa make_message(), vsnprintf(3).
 */
__printf_2_3 int para_printf(struct para_buffer *b, const char *fmt, ...)
{
        int ret, sz_off = (b->flags & PBF_SIZE_PREFIX)? 5 : 0;

        if (!b->buf) {
                b->buf = para_malloc(128);
                b->size = 128;
                b->offset = 0;
        }
        while (1) {
                char *p = b->buf + b->offset;
                size_t size = b->size - b->offset;
                va_list ap;

                if (size > sz_off) {
                        va_start(ap, fmt);
                        ret = vsnprintf(p + sz_off, size - sz_off, fmt, ap);
                        va_end(ap);
                        if (ret > -1 && ret < size - sz_off) { /* success */
                                b->offset += ret + sz_off;
                                if (sz_off)
                                        write_size_header(p, ret);
                                return ret + sz_off;
                        }
                }
                /* check if we may grow the buffer */
                if (!b->max_size || 2 * b->size < b->max_size) { /* yes */
                        /* try again with more space */
                        b->size *= 2;
                        b->buf = para_realloc(b->buf, b->size);
                        continue;
                }
                /* can't grow buffer */
                if (!b->offset || !b->max_size_handler) /* message too large */
                        return -ERRNO_TO_PARA_ERROR(ENOSPC);
                ret = b->max_size_handler(b->buf, b->offset, b->private_data);
                if (ret < 0)
                        return ret;
                b->offset = 0;
        }
}

/** \cond llong_minmax */
/* LLONG_MAX and LLONG_MIN might not be defined. */
#ifndef LLONG_MAX
#define LLONG_MAX 9223372036854775807LL
#endif
#ifndef LLONG_MIN
#define LLONG_MIN (-LLONG_MAX - 1LL)
#endif
/** \endcond llong_minmax */

/**
 * Convert a string to a 64-bit signed integer value.
 *
 * \param str The string to be converted.
 * \param value Result pointer.
 *
 * \return Standard.
 *
 * \sa para_atoi32(), strtol(3), atoi(3).
 */
int para_atoi64(const char *str, int64_t *value)
{
        char *endptr;
        long long tmp;

        errno = 0; /* To distinguish success/failure after call */
        tmp = strtoll(str, &endptr, 10);
        if (errno == ERANGE && (tmp == LLONG_MAX || tmp == LLONG_MIN))
                return -E_ATOI_OVERFLOW;
        if (errno != 0 && tmp == 0) /* other error */
                return -E_STRTOLL;
        if (endptr == str)
                return -E_ATOI_NO_DIGITS;
        if (*endptr != '\0') /* Further characters after number */
                return -E_ATOI_JUNK_AT_END;
        *value = tmp;
        return 1;
}

/**
 * Convert a string to a 32-bit signed integer value.
 *
 * \param str The string to be converted.
 * \param value Result pointer.
 *
 * \return Standard.
 *
 * \sa para_atoi64().
*/
int para_atoi32(const char *str, int32_t *value)
{
        int64_t tmp;
        int ret;
        const int32_t max = 2147483647;

        ret = para_atoi64(str, &tmp);
        if (ret < 0)
                return ret;
        if (tmp > max || tmp < -max - 1)
                return -E_ATOI_OVERFLOW;
        *value = tmp;
        return 1;
}

static inline int loglevel_equal(const char *arg, const char * const ll)
{
        return !strncasecmp(arg, ll, strlen(ll));
}

/**
 * Compute the loglevel number from its name.
 *
 * \param txt The name of the loglevel (debug, info, ...).
 *
 * \return The numeric representation of the loglevel name.
 */
int get_loglevel_by_name(const char *txt)
{
        if (loglevel_equal(txt, "debug"))
                return LL_DEBUG;
        if (loglevel_equal(txt, "info"))
                return LL_INFO;
        if (loglevel_equal(txt, "notice"))
                return LL_NOTICE;
        if (loglevel_equal(txt, "warning"))
                return LL_WARNING;
        if (loglevel_equal(txt, "error"))
                return LL_ERROR;
        if (loglevel_equal(txt, "crit"))
                return LL_CRIT;
        if (loglevel_equal(txt, "emerg"))
                return LL_EMERG;
        return -1;
}

static int get_next_word(const char *buf, const char *delim, char **word)
{
        enum line_state_flags {LSF_HAVE_WORD = 1, LSF_BACKSLASH = 2,
                LSF_SINGLE_QUOTE = 4, LSF_DOUBLE_QUOTE = 8};
        const char *in;
        char *out;
        int ret, state = 0;

        out = para_malloc(strlen(buf) + 1);
        *out = '\0';
        *word = out;
        for (in = buf; *in; in++) {
                const char *p;

                switch (*in) {
                case '\\':
                        if (state & LSF_BACKSLASH) /* \\ */
                                goto copy_char;
                        state |= LSF_BACKSLASH;
                        state |= LSF_HAVE_WORD;
                        continue;
                case 'n':
                case 't':
                        if (state & LSF_BACKSLASH) { /* \n or \t */
                                *out++ = (*in == 'n')? '\n' : '\t';
                                state &= ~LSF_BACKSLASH;
                                continue;
                        }
                        goto copy_char;
                case '"':
                        if (state & LSF_BACKSLASH) /* \" */
                                goto copy_char;
                        if (state & LSF_SINGLE_QUOTE) /* '" */
                                goto copy_char;
                        if (state & LSF_DOUBLE_QUOTE) {
                                state &= ~LSF_DOUBLE_QUOTE;
                                continue;
                        }
                        state |= LSF_HAVE_WORD;
                        state |= LSF_DOUBLE_QUOTE;
                        continue;
                case '\'':
                        if (state & LSF_BACKSLASH) /* \' */
                                goto copy_char;
                        if (state & LSF_DOUBLE_QUOTE) /* "' */
                                goto copy_char;
                        if (state & LSF_SINGLE_QUOTE) {
                                state &= ~LSF_SINGLE_QUOTE;
                                continue;
                        }
                        state |= LSF_HAVE_WORD;
                        state |= LSF_SINGLE_QUOTE;
                        continue;
                }
                for (p = delim; *p; p++) {
                        if (*in != *p)
                                continue;
                        if (state & LSF_BACKSLASH)
                                goto copy_char;
                        if (state & LSF_SINGLE_QUOTE)
                                goto copy_char;
                        if (state & LSF_DOUBLE_QUOTE)
                                goto copy_char;
                        if (state & LSF_HAVE_WORD)
                                goto success;
                        break;
                }
                if (*p) /* ignore delimiter at the beginning */
                        continue;
copy_char:
                state |= LSF_HAVE_WORD;
                *out++ = *in;
                state &= ~LSF_BACKSLASH;
        }
        ret = 0;
        if (!(state & LSF_HAVE_WORD))
                goto out;
        ret = -ERRNO_TO_PARA_ERROR(EINVAL);
        if (state & LSF_BACKSLASH) {
                PARA_ERROR_LOG("trailing backslash\n");
                goto out;
        }
        if ((state & LSF_SINGLE_QUOTE) || (state & LSF_DOUBLE_QUOTE)) {
                PARA_ERROR_LOG("unmatched quote character\n");
                goto out;
        }
success:
        *out = '\0';
        return in - buf;
out:
        free(*word);
        *word = NULL;
        return ret;
}

/**
 * Get the number of the word the cursor is on.
 *
 * \param buf The zero-terminated line buffer.
 * \param delim Characters that separate words.
 * \param point The cursor position.
 *
 * \return Zero-based word number.
 */
int compute_word_num(const char *buf, const char *delim, int point)
{
        int ret, num_words;
        const char *p;
        char *word;

        for (p = buf, num_words = 0; ; p += ret, num_words++) {
                ret = get_next_word(p, delim, &word);
                if (ret <= 0)
                        break;
                free(word);
                if (p + ret >= buf + point)
                        break;
        }
        return num_words;
}

/**
 * Free an array of words created by create_argv().
 *
 * \param argv A pointer previously obtained by \ref create_argv().
 */
void free_argv(char **argv)
{
        int i;

        if (!argv)
                return;
        for (i = 0; argv[i]; i++)
                free(argv[i]);
        free(argv);
}

/**
 * Split a buffer into words.
 *
 * This parser honors single and double quotes, backslash-escaped characters
 * and special characters like \p \\n. The result contains pointers to copies
 * of the words contained in \a buf and has to be freed by using \ref
 * free_argv().
 *
 * \param buf The buffer to be split.
 * \param delim Each character in this string is treated as a separator.
 * \param result The array of words is returned here.
 *
 * \return Number of words in \a buf, negative on errors.
 */
int create_argv(const char *buf, const char *delim, char ***result)
{
        char *word, **argv = para_malloc(2 * sizeof(char *));
        const char *p;
        int ret, num_words;

        for (p = buf, num_words = 0; ; p += ret, num_words++) {
                ret = get_next_word(p, delim, &word);
                if (ret < 0)
                        goto err;
                if (!ret)
                        break;
                argv = para_realloc(argv, (num_words + 2) * sizeof(char*));
                argv[num_words] = word;
        }
        argv[num_words] = NULL;
        *result = argv;
        return num_words;
err:
        while (num_words > 0)
                free(argv[--num_words]);
        free(argv);
        *result = NULL;
        return ret;
}

/**
 * Compile a regular expression.
 *
 * This simple wrapper calls regcomp() and logs a message on errors.
 *
 * \param preg See regcomp(3).
 * \param regex See regcomp(3).
 * \param cflags See regcomp(3).
 *
 * \return Standard.
 */
int para_regcomp(regex_t *preg, const char *regex, int cflags)
{
        char *buf;
        size_t size;
        int ret = regcomp(preg, regex, cflags);

        if (ret == 0)
                return 1;
        size = regerror(ret, preg, NULL, 0);
        buf = para_malloc(size);
        regerror(ret, preg, buf, size);
        PARA_ERROR_LOG("%s\n", buf);
        free(buf);
        return -E_REGEX;
}

/**
 * strdup() for not necessarily zero-terminated strings.
 *
 * \param src The source buffer.
 * \param len The number of bytes to be copied.
 *
 * \return A 0-terminated buffer of length \a len + 1.
 *
 * This is similar to strndup(), which is a GNU extension. However, one
 * difference is that strndup() returns \p NULL if insufficient memory was
 * available while this function aborts in this case.
 *
 * \sa strdup(), \ref para_strdup().
 */
char *safe_strdup(const char *src, size_t len)
{
        char *p;

        assert(len < (size_t)-1);
        p = para_malloc(len + 1);
        if (len > 0)
                memcpy(p, src, len);
        p[len] = '\0';
        return p;
}

/**
 * Copy the value of a key=value pair.
 *
 * This checks whether the given buffer starts with "key=", ignoring case. If
 * yes, a copy of the value is returned. The source buffer may not be
 * zero-terminated.
 *
 * \param src The source buffer.
 * \param len The number of bytes of the tag.
 * \param key Only copy if it is the value of this key.
 *
 * \return A zero-terminated buffer, or \p NULL if the key was
 * not of the given type.
 */
char *key_value_copy(const char *src, size_t len, const char *key)
{
        int keylen = strlen(key);

        if (len <= keylen)
                return NULL;
        if (strncasecmp(src, key, keylen))
                return NULL;
        if (src[keylen] != '=')
                return NULL;
        return safe_strdup(src + keylen + 1, len - keylen - 1);
}