CharField max_length 2^n vs 2^n-1

Why does one use max_length of CharField equal to 2^n whilst others use 2^n-1?

For example:

• in django.contrib.gis.db.backends.postgis.models (django 1.3):

  class SpatialRefSys(models.Model, SpatialRefSysMixin):
      srtext = models.CharField(max_length=2048)
  

• in django_openid_auth.models (djano-openid-a开发者_开发问答uth 0.3):

  class Nonce(models.Model):
      server_url = models.CharField(max_length=2047)
  

Although it is not scientific measure, 2048 seems to be more popular than 2047, but 255 is more popular than 256. Django documentation says, that in MySQL max_length is restricted to 255 characters if you are using unique=True. But why would I use 2^n-1 instead od 2^n in other cases?

You've mostly got it. It's not just the unique=True argument for 255, it's also that strings up to 255 long can sometimes be stored more efficiently.

So the answer is that there is maybe a point to doing it for 255 vs. 256, but for other lengths it's very likely to be pointless. Picking power-of-two lengths to begin with is often not done for a scientific reason (most of us haven't actually benchmarked that our 512-long field made our application run measurably faster than a 513-long field).

That said, there may be application-specific cases where the Django application is a frontend to a C application, where having N^2-1 strings is useful further down the line to efficiently store the additional terminating \0 byte.