是否应使用引号将来自数据库的JSON编码的浮点数括在引号中？

Question

I have the following code:

$stmt = $db->prepare("SELECT LATITUDE, LONGITUDE FROM NODEGEOLOCATION WHERE NODEID = $i");
$stmt->execute();
$path[] = $stmt->fetch(PDO::FETCH_ASSOC);

and I use json_encode to convert this to a JSON format. However, I am getting the following JSON:

[{"route":[{"LATITUDE":"32.224519","LONGITUDE":"-110.947325"},{"LATITUDE":"32.227820","LONGITUDE":"-110.947293"},{"LATITUDE":"32.227843","LONGITUDE":"-110.943865"},{"LATITUDE":"32.230618","LONGITUDE":"-110.943919"},{"LATITUDE":"32.231755","LONGITUDE":"-110.943927"},{"LATITUDE":"32.233836","LONGITUDE":"-110.943963"},{"LATITUDE":"32.233850","LONGITUDE":"-110.946061"},{"LATITUDE":"32.236035","LONGITUDE":"-110.946061"},{"LATITUDE":"32.235993","LONGITUDE":"-110.948083"},{"LATITUDE":"32.235977","LONGITUDE":"-110.952433"}]

I don't want the value of LATITUDE and LONGITUDE to be enclosed in quotes. Is this even possible?

Answer 1

It depends on their type in the database I believe.

If they are varchars or string in the database they are treated as strings.

If they are FLOAT 's DOUBLE 's, they shouldn't have quotes.

Answer 2

It looks like PDO always returns database results as strings, even when they are numbers or floats.

PHP's JSON encoder doesn't perform any variety of type sniffing : if the value is held in a PHP string, it gets put in quotes, even if it could be represented as an integer or float.

You can deal with this in two ways.

1. Expressly cast the value to a float before using json_encode .
2. Alter the consumer of the JSON to cast the value to a float before working with it.

---

Here's an example of automatic casting based on getColumnMeta . First let's create a dummy database and insert some data.

php > $pdo = new PDO('sqlite::memory:');
php > $pdo->exec('create table foo(a integer, b float, c text)');
php > $sh = $pdo->prepare('insert into foo(a, b, c) values(?, ?, ?)');
php > $sh->execute(array(1, 2.2, 'Three'));
php > $sh->execute(array(4, 5.55, 'Six'));
php > $sh->execute(array(7, 88.888, 'Nine'));

Now we'll prove that the data is there, for our demo:

php > $sh = $pdo->prepare('select * from foo where a = 4');
php > $sh->execute();
php > $row = $sh->fetch(PDO::FETCH_ASSOC);
php > var_dump($row);
array(3) {
  ["a"]=>
  string(1) "4"
  ["b"]=>
  string(4) "5.55"
  ["c"]=>
  string(3) "Six"
}

Let's see what PDO will tell us about the first column.

php > print_r($sh->getColumnMeta(0));
Array
(
    [native_type] => integer
    [sqlite:decl_type] => integer
    [flags] => Array
        (
        )

    [name] => a
    [len] => 4294967295
    [precision] => 0
    [pdo_type] => 2
)

Beautiful. I did this for columns 1 and 2 as well, which returned "double" and "string" respectively. Now let's build a map of column name to type:

php > $cols = count($row);
php > $col_types = array(); 
php > while($cols-- > 0) { 
          $col_info = $sh->getColumnMeta($cols);
          $col_types[ $col_info['name'] ] = $col_info['native_type']; 
      }
php > print_r($col_types);
Array
(
    [c] => string
    [b] => double
    [a] => integer
)

Yeah, it's in reverse, doesn't matter! Now let's convert our row.

php > foreach($row as $k => $v) {
          if($col_types[$k] == 'double')
              $row[$k] = (float)$v;
          if($col_types[$k] == 'integer')
              $row[$k] = (int)$v;
      }
php > var_dump($row);
array(3) {
  ["a"]=>
  int(4)
  ["b"]=>
  float(5.55)
  ["c"]=>
  string(3) "Six"
}
php > echo json_encode($row);
{"a":4,"b":5.55,"c":"Six"}

Tada! You're probably going to want to wrap this in a convenience method, it's a bit hairy to do every single time by hand. I could also have used columnCount to get the number of columns in the result set rather than counting $row .

---

Overkill much? Here's an untested example that targets the specific column names and builds the array of hashes by hand in the process.

$results = array();
$convert_columns = array( 'b' );
$sh = $pdo->prepare('select * from foo where a = 1');
$sh->execute();
while($row = $sh->fetch(PDO::FETCH_ASSOC)) {
    foreach($convert_columns as $colname)
        $row[$colname] = (float)$row[$colname];
    $results[] = $row;
}