RE: [Gnu-arch-users] Re: grokking arch

[Parker, Ron]

> From: Charles Duffy [mailto:address@hidden

> On Thu, 2004-03-25 at 17:49, Stefan Monnier wrote:

> > Whether taglines use UUIDs or marshmallows doesn't make any 
> difference to
> > the fact that they may be mistakenly modified or even removed[...]

> That said, if it comes up in practice for people who aren't me who use
> UUIDs, I suppose it could be a genuine issue. Something 
> tree-lint could
> warn on, maybe, if you have a good detection algorithm?

Using "grep '^[A-Fa-f]*$' /usr/share/dict/words |wc" on SFU turned up 100
English words that theoretically could appear in a UUID and could therefore
fall victim to "s/word/other/".

At least ace, add, cab, cad, db, facade, and face seem likely to turn up in
some code somewhere, if not in a UUID.  The following list would seem to
fit:  a solitaire game, a calculator, a CAD program, anything with a
database, OO design pattern based code, and a 2D or 3D graphics program.
Looking at "db" alone, it would hit 27 out of every 256 UUIDs.  That's over
10%.

This is borne out by the following command:

         uuidgen -n30000|egrep -i
"ace|add|cab|cad|db|facade|face">,,dict-test

30,000 files is not an unreasonable size for a project.  10 runs of this
command produced 3,300-3,500 matches per run.  The final one produced 3,350
hits.  Even eliminating "db" from the pattern resulted in 605 hits, that's
still about 5%.  See, the break down of hits below.

        $ for word in ace add cab cad db facade face
        > do   
        >   echo -ne $word\\t  
        >   fgrep $word ,,dict-test|wc -l
        > done
        ace         170
        add         157
        cab         157
        cad         128
        db         2809
        facade        0
        face          9

I'm just supplying the raw data.  Someone else can decide what to do with
it.