Re: [Gnu-arch-users] Re: Efficiently find out if <file> is source or not

[Tom Lord]

    > From: Stefan Monnier <address@hidden>

    > > You would need to (a) parse {arch}=tagging-method; (b) make sure it
    > > matches the source regexp and not the unrecognised one; (c) depending
    > > on the tagging method, look for an arch-tag: in the file, or a
    > > .arch-ids/$FILENAME.id, or whatever.

    > That's pretty damn nasty since you have to be extra careful to match
    > regexes in the same order as tla, and not to forget to check every
    > .arch-inventory file on the way to the file you're checking, and you have
    > to make sure you're using the same regex rules (in my case I'm writing 
this
    > in Emacs Lisp where the regexps are different), ...


Heh.

I don't disagree but arch wouldn't be here if I had stood my ground
about that.

Once upon a time, `inventory' was a hella-complicated `find'
invocation but not much more than that.  Configurable a _little_ bit,
but not so much.

Two issues undid that.

1) `find' was pretty slow for this purpose.   `inventory' was one of the
   first places of `larch' (arch in sh) to be rewritten in C.  

2) naming convention flexibility was, like, the #1 or #2 demand for
   the first year or 18 months or so of arch.   The simple `find'
   model got left in the dust because the rock of arch's naming
   conventions was softer than than the hard-place of naming
   conventions in previously existing projects.


The closest thing to hope, for K.I.S.S. fans, is that `inventory'
eventually gets factored out and takes its place in some abstract
pantheon alongside `find' as a fundamental way to look at a big tree.

-t