I was looking to implement a year column and while researching I stumbled on the YEAR data type which sounded just right by its name, I assumed that it would just be something like an integer that can maybe hold only 4 digits, maybe more if negative?
But then I noticed while actually trying it out that some years I was inputting randomly by hand never went through giving an out of range error, so I went to look at the full details and, sure enough, it’s limited to years between 1901 and 2155, just 2155!
In terms of life of an application 2155 is just around the corner, well not that any software has ever lived that long, but you get what I mean in the sense that we want our programs to be as little affected by time within what’s reasonable given space constraints.
So what will they do when they get close enough to that year, because you don’t even have to be in that year to need it accessible, there could be references that point to the future, maybe for planning of some thing or user selected dates and whatnot; will they change the underlying definition of it as time passes so it’s always shifted forward? If that’s the approach they’ll take, will they just tell everyone who’s using this type that their older dates will just not be supported anymore and they need to migrate to a different type? YEAR-OLD? Then YEAR-OLDER? Then YEAR-OLDER-BUT-LIKE-ACTUALLY? Or, that if they plan to stay in business, they should move to SMALLINT?
Or will they take the opposite approach and put out a new YEAR datatype every time the 256 range is expired like YEAR-NEW, YEAR-NEW-1, YEAR-FINAL, YEAR-JK-GUYS-THE-WORLD-HASNT-COLLAPSED, etc.?

So I wonder, what’s the point of this data type? It’s just so incredibly restricted that I don’t see even a hypothetical use.
There exist other questions like this (example) but I think they all don’t address this point: has anyone from MariaDB or MySQL or an SQL committee (I don’t know if that’s a thing) wrote up some document that describes the plan for how this datatype will evolve as time passes? An RFC or anything like that?

  • bjorney@lemmy.ca
    link
    fedilink
    arrow-up
    6
    ·
    5 months ago

    I expect it won’t

    The year datatype is a 1 byte integer, but the engine adds/subtracts 1900 to the value under the hood and has special handling for zero.

    If you need to store more than 255 years range, you can use a 2 byte integer, which doesn’t need that special handling under the hood, because with 2 bytes you can store 65000+ years

    • QuazarOmega@lemy.lolOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      5 months ago

      Yes 2 bytes is absolutely fine for me in fact (waiting for this comment to age like milk in my cryo pod), but then if YEAR will just stay the same forever, will it become a relic of the past? If so, why YEAR in the first place, who would actually make use of it?

  • nebeker@programming.dev
    link
    fedilink
    English
    arrow-up
    2
    ·
    5 months ago

    To answer the part of your question I think is most fun, there is a standard for SQL. There are many dialects of SQL, but you’ll often hear of “ANSI SQL.” The latest version is SQL:2023.

    Looking at the MySQL manual entry for the YEAR type, I think we can conclude two things:

    1. The developers consider the possibility of deprecating and removing support for time data type features;

    2. They use “reasonable defaults” for conversions of 2 digit years, based on the current year.

    The good news is it sounds like this issue is being taken into account. I’m sure the conversion window will be adjusted in future version and the data type may be changed or deprecated altogether. I wouldn’t be surprised if they added a YEAR2 though. T-SQL has a datetime2, after all.