On: Why numbering should start at zero (1982)

I've long been a zero-based-index-loving geek, but I'm now coming around to preferring one-based indices in programming. Hopefully in the near future I'll write a blog post detailing why.

Regarding the article itself... sorry, Dijkstra, but I think conventions C and D are more beautiful, as I prefer discrete ranges where the signs on both ends are the same. Convention C is my favorite, as it only makes mention of members of the range, not numbers that lie outside of it.

I have an example of a similar journey I followed.

The Lua language has a 1-based indexing scheme. Furthermore, the only thing that differentiates a hash-map from an ordered array is the existence of only keys in the range [1, ∞) (and they must be entirely contiguous starting from 1 until the last present key).

The interface for arrays and hash-map data types is uniform, called a 'table', which has your standard `mapper[key] = value` syntax (and also, equivalently, `mapper.key = value` for well-formatted string-literal keys).

This results in the very surprising and confounding fact that if a naive programmer doesn't realize that Lua has this system, they will very likely end up starting from 0 and build an unordered hash map rather than an ordered array - complete with all the resulting performance penalties and unordered traversals.

For a long time this struck me as being absolutely absurd and very damning. After working with Lua for a while, though, there began to be a certain elegance to it that I could admire. There is a SO post that goes in to some of the benefits[0][1].

I still much prefer 0-based indexing though.

0: http://stackoverflow.com/questions/2785704/why-do-lua-arrays...

1: Irony intended.

Reading Dijkstra's own argument convinced me that 1-indexing is better. Convention C (consistently non-strict) is the one that we normally use to describe ranges in English.

While the article is explicitly only dealing with natural numbers, convention C is insufficiently expressive for anything more granular than the natural numbers.