Kernel prepatch 3.11-rc4 [LWN.net]

Kernel prepatch 3.11-rc4

Posted Aug 7, 2013 9:25 UTC (Wed) by dgm (subscriber, #49227)
In reply to: Kernel prepatch 3.11-rc4 by khim
Parent article: Kernel prepatch 3.11-rc4

How can you tell if an expression is correct if you don't read them in full?

to post comments

Kernel prepatch 3.11-rc4

Posted Aug 7, 2013 9:32 UTC (Wed) by mpr22 (subscriber, #60784) [Link] (2 responses)

You can't. Which is why making expressions hard to read is a bad idea.

Kernel prepatch 3.11-rc4

Posted Aug 14, 2013 13:24 UTC (Wed) by dgm (subscriber, #49227) [Link] (1 responses)

Which of those expressions is harder to read:

A == B

B == A

Kernel prepatch 3.11-rc4

Posted Aug 14, 2013 15:18 UTC (Wed) by mathstuf (subscriber, #69389) [Link]

Let's use some actual examples:

> i == container.end()
> container.end() == i

> result == (2 * idx * idx)
> (2 * idx * idx) == result

> pivot < container.size() / 2
> container.size() / 2 > pivot

> speed > threshold
> threshold < speed

In all of these cases except the last, I would naturally write the first line[1]. I think they read more naturally. To reduce errors, I have const on every variable possible (containers I have to build are the exception (I can't wait for initializer lists)), so my code tends to follow SSA a lot[2].

[1]I try and order all orderings to use '<' since I think "(a > min) && (a < max)" is harder to parse and using one comparison everywhere helps keep the mental overhead low.
[2]It helps in debugging since all intermediates are named and you don't get things like i->first.second.second in the middle of lines and have to backtrack what the iterator's pair nesting is all the time.