Best way to generate permalinks

Question

I'm looking to create a database of goods and I don't think an incrementing id would be adequate:

http://example.com/0001
http://example.com/0002

So with 4 digits I can address 10000 items IIUC. I don't I want to show how many items I have in total or make it too guessable was the series is.

Of the examples of similar design I'm after I found:

http://uk.imdb.com/title/tt0083658/
http://stackoverflow.com/questions/2013091/
http://twitter.com/#!/kaihendry/status/160934896121032705
https://plus.google.com/107429617152575897589
http://bit.ly/xW5cGR

I think I like bit.ly the most, though the uppercase letters might be error prone. IIUC with bitl.y's design you can address Upper and lowercase plus numbers to the tune of 5 digits. 62*62*62*62*62=916 132 832 which is great size to work with.

I am not sure what the theory of this study is called. If you could point me to some good links on the topic I would be grateful. I want the codes to be short and usable as permalinks.

I'm wondering why Google plus settled for a very long 21 digit user id. Are there 3rd party tools or database features from MariaDB or MongoDB that sort of automatically generate sane short codes for you?

Answer 1

You want a base N encoding library/function, e.g. a base62 encoder will give you the 26 lower-cases, 26 upper-cases, and 10 digits. Then you just apply it to a counter variable, the number of existing elements, or the new item's ID.

In pseudocode:

shortID = base(number_of_existing_monkeys + 1)
save(new Monkey(shortID, url, description)

(This is using number of existing monkeys, you could alternatively use an auto-incrementing counter variable.)

Here are PHP base62 encode/decode functions for example: http://snipplr.com/view/22246/

As you can see, libraries like this usually have a character array somewhere, so you can just edit the array with your own character space.

As far as Google using 21-char IDs, that's really something different. It's a GUID (http://en.wikipedia.org/wiki/Globally_unique_identifier). The idea is to generate something guaranteed to be unique even without looking at existing records. It means you can generate IDs offline and without having to worry about transactional integrity. It's convenient for when you want to work with the new ID, maybe even save it somewhere else, before the object has been persisted.

Answer 2

If you are concerned about upper/lower case causing confusion then 6 lower case letters and/or digits (base 36) gives you over 2 billion entries to work with. This nicely covers the address space of positive 32 bit integers.

You can use a numeric ID (positive 32 bit integer) which can be generated randomly if you are concerned about letting people know how many products you have. Then just convert the integer to a six character base 36 number.

Note that randomly generating your integer key may have repercussions if you actually create new entries frequently enough. You can poke around on SO or dba.SE to learn about the pros and cons of monotonically increasing vs. random clustered indexes.