On Jul 16, 2010, at 10:39 PM, Derek M Jones wrote:
In any experiment it is necessary to restrict the attributes
that could change the experimental results to just those of
interest. The larger the amount of source used the larger
the likelihood that more uncontrolled attributes will occur.
In the setup I have described, the *only* thing that changes
in the code is the spelling of the identifiers.
What exactly is readability? Source is read for a variety of purposes
and readability could change for each purpose.
That's an excellent point, which is why I'm working from a
definition that I _thought_ was tolerably specific.
"Difficulty of finding correct information while performing
a simple maintenance task"
I would say that your definition of readability would be a small
component of this task. You are asking subjects to comprehend
an algorithm and then come up with a variation of that algorithm.
This is a big task in itself.
They will be *told* in brief outline what the algorithm is.
The bugs will all be of the "wrong identifier used" variety,
so calling it "a variation of that algorithm" is a bit of
a stretch.
This is an experiment in ease of maintenance of a particular program
experiment, not a readability experiment.
3 small programs,
3 versions of each,
*only* difference between versions the spelling of
the identifiers.
Given that my question is "does the spelling of identifiers
affect how hard it is to work with code", if there is a
detectable difference in how hard it is to work with the
code, there's nothing _else_ being varied that it _could_
be other than the spelling of the identifiers. That being
so, I don't really care whether it's called a readability
experiment or an ease of maintenance experiment, just as
long as it is relevant to the question of immediate interest.
Giving the same code to two groups of people, with the source
formatted different for each group and comparing group
performance might be a way of seeing performance differences
caused by layout differences.
Yes, but I said I wasn't interested in that.
The layout is *NOT* going to vary at all.
I accept that there almost surely are performance differences
caused by layout differences. I know that I find code
indented by eights so very painful to read that I have to
change the indentation before I even try, and I don't imagine
myself to be unusual. That's precisely why the indentation
will not change, and as I commented, that's precisely why there
are not going to be curly braces but will be end markers.
--
The Open University is incorporated by Royal Charter (RC 000391), an exempt charity
in England & Wales and a charity registered in Scotland (SC 038302).
