1. 10

Computer graphics and statistics often use parameters with values defined in the [0‥1] interval. How would you name a type for such as value (as in typedef NNN float)

It seems that the only mathematical term that fits these requirements is “mantissa”‘ but it’s an overloaded term in computer science.

I settled for “norm” as in practice, the normalization of a value will bring it back to the [0‥1] interval. Thoughts? Recommendations? Alternatives?

Here’s another thread on the topic, btw.


  2. 2

    I’ve always used “fraction,” as some of the commenters on that thread suggest. While not precisely correct, I’ve never had a colleague misinterpret (famous last words).

    1. 1

      What I don’t like about “fraction” is that it suggests the value is “incomplete” (it’s a subset of something bigger). In computer graphics, people often use normalized texture coordinates. Let’s say you have an image that is 1024x1024 pixels, and you define a texture(s,t) function where s,t ∈ [0‥1]. I find that calling s and t fractions is odd in that case, because s and t would be no more or less complete than their non-normalized equivalent. Otherwise “fraction” works pretty well for most of the other uses of values within the interval (for instance, the k in lerp(a,b,k)).

    2. 2

      When ideas get useful enough, they deserve dedicated names. This definitely seems like an area where somebody should just make up a word and run with it.

      1. 1

        I’d call it a normalized index or coordinate, or in some other cases, parameter. It might be a good idea to reserve “coordinate” for continuously varying indices and build the type name around that.