OT: Why would you ever use numbers instead of strings for ids?

[Geoff Canyon]

On Thu, Sep 19, 2013 at 5:46 AM, David Bovill <david.bovill at gmail.com>wrote:

>
>    1. No good reason - historical
>    2. ID lookups tend to be faster
>    3. Some scenarios allow you to derive meaning from incremental increases
>    in the number - ie it is the next in a series
>
> Naturally in Livecode the difference between numbers and strings is
> deliberately and unusually blurred. Still in general it always seemed to me
> that reasons 2 & 3 were insufficient. Can someone explain why on earth the
> Twitter API does not just use screen names ( a unique lower case string of
> 15 chars without spaces) - what is the advantage of also having a 16
> integer number?



I'm not a DBA, but I'll take a swing at this.

1. Databases have auto-increment for numeric keys. It would require work to
do the same thing with strings.
2. Numeric keys are efficient -- both compact and fast. Maybe not as
important these days, but still a consideration.
3. Context does matter, so the fact that I can look at a row with a key of
12345 and know  that rows created around the same time will have similar
numbers is useful, even if the table also has columns for the creation date.

Finally -- I'm a purist, but primary keys should never be based on actual
data. In other words, they should be artificial -- some say surrogate. Here's
an essay on keys <http://www.agiledata.org/essays/keys.html>. I only
skimmed it, but it looks reasonable.