On 2015/12/03 3:04 AM, Scott Hess wrote: > I discourage this kind of usage because it means that in some distant > future when someone has to make things work with a different database > engine, they have to grind through and check every weirdo VARCHAR(73) and > MEDIUMBIGINT declaration someone put in, because none of them have ever > been tested with range enforcement enabled. So where someone meant > "VARCHAR(256)" but actually typed "VARCHAR(6)", will your code suddenly > start throwing clean errors which immediately suggest where to look, or > will it just muddle through corrupting your data? There can certainly be > subtle issues in any type, but my experience is that when you're trying to > decode someone's code, it's easiest when the code says what is actually > happening, not what it wishes were happening!
I posit that a column declared as: col VARCHAR(32) NOT NULL says a whole lot more about what is actually happening than: col TEXT NOT NULL And sure, I agree a mistyped word can be hard to decode by a future programmer, but that is a statistical probability in any case. > Of course, if you are _currently_ writing cross-platform code, absolutely > write the cross-platform code! And I will agree that the above position > has some issues when faced with things such as INT being 32 bits//.... etc. I *ALWAYS* write cross-platform code as far as SQL is concerned. I even think in this day and age every programmer should, or is there a case for the opposite?
