Hacker News new | ask | show | jobs
by coderdude 4888 days ago
One thing that bugs me is MySQL will truncate a value if the column can't hold it all and will only issue a warning (by default?). Although in most instances I've already done the foot work of checking a value's length, it would be nice to be able to throw an exception when it occurs. Of course that could be the choice of my framework and not MySQL's fault.
1 comments
AaronBBrown 4888 days ago 

  SET sql_mode = 'STRICT_ALL_TABLES';
http://dev.mysql.com/doc/refman/5.5/en/server-sql-mode.html#...

Yes, it's not default. Yes, it can be overridden. Yes, that bothers a lot of people.

link
rosser 4888 days ago
Yes, that bothers a lot of people.

With good reason. This is your database we're talking about. If it's going to be sloppy with your data by default, it's the wrong tool.

link
AaronBBrown 4888 days ago
I suppose that depends on your definition of "database." This argument is very close to the No True Scotsman fallacy. Nowhere in the definition of "database" does it say that it won't default to truncating your data when you try to stuff it into a field that is too small. MongoDB, Riak, VoltDB, PostgreSQL, and CouchBase are all "databases" and they all have varying levels of strictness.

"it's the wrong tool" is an opinion and it is unsupported by history. It many cases, it's exactly the right tool and users are given the power to control the level of strictness they they require. Want to allow GROUP BYs that don't include all non-aggregate columns? Sure, you can do that. Or not. Want MySQL to accept what you give it and throw warnings instead of errors? Yep, you can do that, too. Or not.

The defaults are bad. No one will defend that. They are also easy to change. A privilege level allowing/denying the ability to change the sql_mode would be a welcome addition, also. Yet, MySQL is a capable tool that does some things really well, something things not so well, and some things terribly. Just like every other product out there. It is a tool. It works well for many, but not all cases. Apparently, you find it inadequate for your particular needs. I get that. You shouldn't use it.

link
rosser 4888 days ago
If no-one will defend that the defaults are bad defaults, then why are they the defaults?

As for the "No True Scotsman" notion, fair enough; point conceded. I just can't imagine anyone who actually, you know, cares about their data being happy about finding out — by virtue of the fact that the sloppy behaviors are the default — that they've lost data, or had it corrupted somehow, because of those defaults.

Yes, it's their responsibility, in choosing the tool, to make sure they're using it properly. But the folks making the tool should also have some degree of responsibility to their users not to fuck up their data. Maybe I'm just spoiled, having lived and worked in the PostgreSQL world as long as I have. There, breaking users' data is tolerated about as well as Linus tolerates kernels breaking userspace...

link