From Arnout Engelen

I ran CANAPA on some DigiD case study code. As it turns out, modularizing on a per-class basis has a positive effect on time-efficiency. On a per-method basis, the overhead when analysing small methods nullifies the speed increase, but the performance isn't too bad either.

After running CANAPA on the modules seperately, the process is finalized by running it again on the entire code at once

Contents 1 Full 2 Per class 2.1 every class 2.2 final run 2.3 cumulative 3 Per method 3.1 every method 3.2 final run 3.3 cumulative 4 Quality

[edit] Full

full:

real    45m1.777s
user    45m39.547s
sys     0m14.396s

[edit] Per class

[edit] every class

real    25m34.865s
user    23m21.100s
sys     0m14.707s

[edit] final run

real    12m52.229s
user    13m11.513s
sys     0m6.278s

[edit] cumulative

~38 minutes

[edit] Per method

[edit] every method

real    27m10.597s
user    24m48.366s
sys     0m40.613s

[edit] final run

real    21m10.884s
user    21m51.949s
sys     0m8.615s

[edit] cumulative

~48 minutes

[edit] Quality

• Full has 4 more annotations compared to permethod-final and perclass-final
• permethod-final and perclass-final are identical
• permethod-final has 5 more annotations compared to permethod-nofinal
• perclass-final has 1 more annotation compared to permethod-nofinal