Newbie question: zend_hash_find View as Markdown

hi,

Have you tried looking at how the parse_ini_file function works? that
should answer your questions on this.

• tul

Arnold Daniels wrote:

Hi,

I'm new at PHP internals and have read the 'Extension Writing' tutorial
(part 1-3) of devzone.zend.com. I'm writing a set of parsers, to parse
different kind of configuration formats. The first is an .ini parser.
The code almost works correctly. The only problem I have is that
sections get overwritten if appear multiple times in the .ini content.

[abc]
test = 1
[abc]
xyz = 2

Should result in: array('abc'=>array('test'=>1, 'xyz'=>2))
but instead returns: array('abc'=>array('xyz'=>2))

The logical conclusion is that my zend_hash_find call doesn't work
correctly and is always returning FAILURE. Though I've got no clue
what's wrong. Perhaps someone could have a quick look at the code.

Thanks a lot,
Arnold

---

/**

• Parse .ini content into a zend array
  */
  PHP_FUNCTION(qconf_ini_parse)
  {
  char *content;
  int len;

  char *c;
  char *ln_start;
  char key[ASCIILINESZ+1];
  char val[ASCIILINESZ+1];
  zval *cur_out;
  zval **fnd_sec;

  char t; // boolean: trim value yes/no

  if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s", &content,
  &len) == FAILURE) {
  RETURN_NULL();
  }

  array_init(return_value);

  cur_out = return_value;
  ln_start = content;

  for (c=content; c <= content+len; c++) {
  if (*c == '\n' || *c == '\0' || c == content+len) {
  *c = '\0';

       t = 0;
       ln_start = strskp(ln_start); /* Skip leading spaces */
       if (*ln_start!=';' && *ln_start!='#' && *ln_start!=0) { /* 
  

Skip comment lines /
if (sscanf(ln_start, "[%[^]]", key)==1) {
/ Valid section name /
if (zend_hash_find(Z_ARRVAL_P(return_value), key,
strlen(key), (void*)&fnd_sec) == FAILURE) {
ALLOC_INIT_ZVAL(cur_out);
array_init(cur_out);
add_assoc_zval(return_value, key, cur_out);
} else {
cur_out = fnd_sec;
}
} else if (sscanf (ln_start, "%[^=] = "%[^"]"", key,
val) == 2
|| sscanf (ln_start, "%[^=] = '%[^']'", key,
val) == 2
|| (t=1 && sscanf (ln_start, "%[^=] = %[^;#]",
key, val) == 2)) {
strcpy(key, strlwc(strcrop(key)));
/
* sscanf cannot handle "" or '' as empty value,
* this is done here
*/
if (!strcmp(val, """") || !strcmp(val, "''")) {
val[0] = (char)0;
} else if (t) {
strcpy(val, strcrop(val));
}
add_assoc_string(cur_out, key, val, 1);
}
}

       ln_start = c+1;
   }

}
}

Hi Arnold,

zend_hash_find() is always returning FAILURE because its "key_length"
parameter includes the terminating '\0'. :-) So you need strlen(key) + 1.

BTW, are any of your keys going to have numeric values ([123] sections
etc.)? Then you'll also have a problem with hash_find() as it doesn't
handle numeric string keys since PHP 5 (converting them to a true number
index like when you use $arr['123']), but the add_assoc_* functions do. If
that's an issue, you have to use zend_symtable_find() since PHP 5, which
works exactly like hash_find() in PHP 4.

Hope that helps.

Matt

----- Original Message -----
From: "Arnold Daniels"
Sent: Wednesday, November 08, 2006

Hi,

I'm new at PHP internals and have read the 'Extension Writing' tutorial
(part 1-3) of devzone.zend.com. I'm writing a set of parsers, to parse
different kind of configuration formats. The first is an .ini parser.
The code almost works correctly. The only problem I have is that
sections get overwritten if appear multiple times in the .ini content.

[abc]
test = 1
[abc]
xyz = 2

Should result in: array('abc'=>array('test'=>1, 'xyz'=>2))
but instead returns: array('abc'=>array('xyz'=>2))

The logical conclusion is that my zend_hash_find call doesn't work
correctly and is always returning FAILURE. Though I've got no clue
what's wrong. Perhaps someone could have a quick look at the code.

Thanks a lot,
Arnold

---

/**

• Parse .ini content into a zend array
  */
  PHP_FUNCTION(qconf_ini_parse)
  {
  char *content;
  int len;

  char *c;
  char *ln_start;
  char key[ASCIILINESZ+1];
  char val[ASCIILINESZ+1];
  zval *cur_out;
  zval **fnd_sec;

  char t; // boolean: trim value yes/no

  if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s", &content,
  &len) == FAILURE) {
  RETURN_NULL();
  }

  array_init(return_value);

  cur_out = return_value;
  ln_start = content;

  for (c=content; c <= content+len; c++) {
  if (*c == '\n' || *c == '\0' || c == content+len) {
  *c = '\0';

       t = 0;
       ln_start = strskp(ln_start); /* Skip leading spaces */
       if (*ln_start!=';' && *ln_start!='#' && *ln_start!=0) { /*
  

Skip comment lines /
if (sscanf(ln_start, "[%[^]]", key)==1) {
/ Valid section name /
if (zend_hash_find(Z_ARRVAL_P(return_value), key,
strlen(key), (void*)&fnd_sec) == FAILURE) {
ALLOC_INIT_ZVAL(cur_out);
array_init(cur_out);
add_assoc_zval(return_value, key, cur_out);
} else {
cur_out = fnd_sec;
}
} else if (sscanf (ln_start, "%[^=] = "%[^"]"", key,
val) == 2
|| sscanf (ln_start, "%[^=] = '%[^']'", key,
val) == 2
|| (t=1 && sscanf (ln_start, "%[^=] = %[^;#]",
key, val) == 2)) {
strcpy(key, strlwc(strcrop(key)));
/
* sscanf cannot handle "" or '' as empty value,
* this is done here
*/
if (!strcmp(val, """") || !strcmp(val, "''")) {
val[0] = (char)0;
} else if (t) {
strcpy(val, strcrop(val));
}
add_assoc_string(cur_out, key, val, 1);
}
}

        ln_start = c+1;
    }
}

}