[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [IMP-dev] Directory structure




On Sep 26, 2008, at 9:16 AM, Ben Webb wrote:

Daniel Russel wrote:
On Sep 24, 2008, at 9:44 AM, Ben Webb wrote:
You really don't like to give up control :-) Anyway, the policies
aren't yours, they are ours. And, as far as I can tell, it isn't a
policy (i.e. it isn't documented as such or wasn't last I looked), it
is just the way things happen to be.

I think you missed my point, which was that (in general) I am wary of
blanket statements along the lines of "once I get write access I am
going to change everything" which seemed to be what you were saying.
That is precisely what others are worried about.
"Changing everything" would take a lot of time :-) I said I would clean up the formatting of those files (and, I didn't say it, but would like to provide a script to autogenerate them since they provide no extra information about IMP beyond the directory structure).


I was speaking generally, but on this issue specifically it doesn't make
a lot of difference to me whether you do the current:
  files = ['a', 'b', 'c', 'd']
or the proposed:
  files = []
  files.append('a')
  files.append('b')
  files.append('c')
  files.append('d')
Actually, in contrast the latter is less efficient Python and looks
uglier to me (but the latter is just aesthetic of course). If you really
want each file on a separate line, I suggest:
  files = [
     'a',
     'b',
     'c',
     'd',
     ]
At least that way the Python code is functionally the same as the
original, is still easy to auto-generate, and isn't as unnecessarily
verbose.
Actually I was proposing the last way since it has less text. An earlier version of my script had done the middle way, which as you point out is a bit ugly and hard to read. I can't imagine the efficiency makes any difference unless python is way worse than I imagine.

The only bugs I can think of were header files not being installed. That
is no longer an issue because they are staged to the build directory
now, so the compile (not just the install) will fail if a file is missed.
Yes, there were several instances of headers not being installed. I like that the failure occurs earlier now, but I still don't see any reason to make people maintain those lists by hand. And having the files on separate lines does reduce the number of conflicts and makes diffs clearer (I have spent too much time sorting out conflicts in the SConscript files--this is part of the reason I switched to a script which just rewrites them all).