Bug #73650 [Com]: why lose sql-results their type in PHP?

From: Date: Fri, 17 Feb 2017 12:39:25 +0000
Subject: Bug #73650 [Com]: why lose sql-results their type in PHP?
References: 1  Groups: php.bugs 
Request: Send a blank email to php-bugs+get-207428@lists.php.net to get a copy of this message
Edit report at https://bugs.php.net/bug.php?id=73650&edit=1 ID: 73650 Comment by: giunta dot gaetano at gmail dot com Reported by: spam2 at rhsoft dot net Summary: why lose sql-results their type in PHP? Status: Open Type: Bug Package: *Database Functions PHP Version: 7.0.14 Block user comment: N Private report: N New Comment: As far as I am concerned, +1 for default type conversion with data from databases, with an optional switch to enable/disable it. I have read the RFC mentioned above, and tbh I am slightly surprised to see that it mentions conversion of data gotten from HTTP requests as primary usecase, instead of DB APIs. In my mind, HTTP is a text-based protocol: I always expect data coming in to be an untrusted piece of text, that I have to cast, sanitize, validate. I would be too scared of enabling automatic type casting for that, and would expect a lot of existing code to break. Otoh, I would expect the correct php types to be gotten from databases as: DBs definitely do have typed data; the int32/64 problem is something that, I think, would pop up more often at design/configuration time rather than runtime; the casting of data gotten from DB is boilerplate code boring enough that I do not think I have ever seen a single dev add to it the check for over/under-flows any way... Previous Comments: ------------------------------------------------------------------------ [2016-12-08 21:55:27] yohgaki@php.net We cannot ignore 32 bit architecture machines due to raise of IoT. Cheap IoT uses 32 bit CPU and I don't think we have to deal with such architecture while. Even with 64 bit CPU, there is signed/unsigned int, numeric data types with db. Converting "int" like value to signed "long" could be wrong. $myint = (int)$int_like_value_from_somethere; is not absolutely safe. Programmer has to make sure it's safe, but PHP cannot. (Unless there is way PHP to know it is safe) In the future, we may have to deal with 128 bit int also. Chances are high we are using 64 bit int at that time. I don't oppose to have conversion features. In fact, I'm the one proposing type affinity, but type conversions have to be controlled by users(programmers). Otherwise, it cannot work correctly. So reasonable choice would be having both "(semi) automatic conversions" and "no conversions" , "no conversion" by default. ------------------------------------------------------------------------ [2016-12-08 19:49:39] spam2 at rhsoft dot net anyways, at that point in time i would be happy if only run-tests.php would respect the environment and not load random configurations not matching the build which is running the test and the reason for run it directly instead of "make test" is that "make test" has nothing better to do then copy /etc/php.ini instead enforcing a empty default config https://bugs.php.net/bug.php?id=73609 ------------------------------------------------------------------------ [2016-12-08 19:37:47] spam2 at rhsoft dot net > PHP's int is only signed 32 bits int under 32 bit architecture show me the machines running in 2016 PHP70/PHP71 under a 32 bit architecture *and* then using declare(strict_types=1); - frankly linux distributions are stopping to build any 32bit stuff at all - there is no RHEL7 even while it ships PHP5.4 how many decades should we accept a doezen potential 32 bit machines stop envolving things? when i develop in strict mode and store a integer to mysql i excpect to get that integer back and not a random string which i have to cast manually or the whole strict_types is pointless because it only leads to fatal errors for no good reasons or you have to use (int) everywhere when something might have came from a database which makes any benefit of typing pointless ------------------------------------------------------------------------ [2016-12-08 19:26:18] cmb@php.net If there is any chance, that an integer retrieved from a database won't fit into a PHP int, the proper way to deal with this would be something like: <?php if ($row['id'] >= PHP_INT_MIN && $row['id'] <= PHP_INT_MAX) { foo((int) $row['id']); } else { // uhm, we can't cast $row['id'] to int … } ------------------------------------------------------------------------ [2016-12-08 19:06:04] spam2 at rhsoft dot net > Integers in database may not match PHP internal > date types. e.g. PHP's int is only signed 32 bits > int under 32 bit architecture but how is that different to $row['id'] = (int)$row['id'] which you now must do manually becaus eotherwise sooner or later a function in a class with declare(strict_types=1); will end in a fatal error? __________________________ inlcude('functions.php'); $row = mysqli_fetch_result($result); foo($row['id']); is a fatal error currently without foo((int)$row['id']) __________________________ functions.php: <?php declare(strict_types=1); function foo(int $id) ( ) ?> ------------------------------------------------------------------------ The remainder of the comments for this report are too long. To view the rest of the comments, please view the bug report online at https://bugs.php.net/bug.php?id=73650 -- Edit this bug report at https://bugs.php.net/bug.php?id=73650&edit=1

« previous php.bugs (#207428) next »