[gentoo-portage-dev] spring cleaning.

[Brian Harring]

Attached is a buttload of patches cleaning modules for the following 
things-

1) types module usage.  Use isinstance instead.
2) string module usage.  Use string methods instead.
3) stat module usage to get attributes.  Stat objects have named 
   attributes, use them (example would be st[stats.ST_MODE] vs 
   st.st_mode)
4) range usage.  Use xrange instead (former returns a list, latter is 
   a generator, eg, less mem usage and usually faster)
5) testing for a key, if it doesn't exist, setting a default.  Use 
   setdefault method instead.
6) testing for a key, if it exists, assign that to a var, else assign 
   a default to the var.  use get method instead.
7) ==/!= None tests.  use is/is not None instead (faster, ptr 
   comparison instead of going through the equality protocol).
8) len testing.  
   if len(blah):  # actually is if bool(len(blah)) effectively
   use 
   if blah:
   instead- the object knows it's being evaluated being evaluated in a 
   boolean context, thus avoids intermediate steps.
9) repeated regex compilation.  If it's a constant pattern, compile it
   once.
10) lack of awareness of iterating over dicts directly, and of 
    iteritems().  Conversion of for x in d.keys() -> to for x in d:
    Latter doesn't create an intermediate list, eg faster/cheaper mem 
    wise (note to anyone paying attention, only iterate over the dict
    directly if you're not mutating the dict within the loop).
    Final note, if you're doing 'if x in d.keys()', you need to be 
    back handed, forcing a linear search for no reason.
11) imports cleanup.  If it doesn't use it, no point in importing.
12) General exception cleanup.  Know what can be thrown, and catch 
    that instead of catching everything and requiring SystemExit crap.
13) Set exists, and python 2.3 is already forced.  Use it. :)
    Related note, portage_util forces sets.Set to be used which is stupid
    if set type exists (former is implemented in python, latter is 
    implemented in cpython, ie, there is a difference in speed).
14) Strings are immutable.  They can't be changed, as such attempting to 
    copy them is stupid (python just returns the string, but still, 
    don't do it).
15) Bundling your own version of spawn is Plain Idiotic (TM).
16) logic cleanup/simplification.

Modules involved here are mostly straightforward cleanup- changes 
should be obvious (sans getbinpkg).

Largest set of changes really is in getbinpkg- code in there is quite 
crufty, and a bit dense (IOW, check that chunk over carefully please).

With the potential exception of getbinpkg, these particular patches 
are all potential 2.1 material imo; no huge rush on that however 
considering that the code already works (these patches just makes the 
code suck less).

At the very least, include the portage_util set change; it'll result 
in a faster unique_array.

Either way, second set of eyes on the changes would be useful...

~harring