Re: [gentoo-dev] [PATCH] eutils.eclass: Add awk wrapper - eawk - edit file in place

[Robin H. Johnson]

On Thu, May 19, 2016 at 10:50:48PM +0100, Amadeusz Żołnowski wrote:
> "Robin H. Johnson" <robbat2@g.o> writes:
> > 1.
> > If your AWK command has an error, then this clobbers the file, a
> > better variant, with die, would be:
> > awk "$@" "$f" >"${tmpf}" || die "awk failed..."
> > cp -f "${tmpf}" "$f" || die "copy back failed"
>
> Thanks! On the one hand it's a simple function, and on the other hand
> there are several things which can go wrong. I will submit corrected
> version with unit test.
>
> Why '-f' is required?
Corner cases w/ copying and bad ebuild usage (specifically running
ebuild as two different non-root users during development). Mostly
habit.

Another thought I had, is that if you're using eawk on something in
$ROOT, you're also going to clobber ownership, permissions and extended
attributes.

> >> # @USAGE: <file> <args>
> > This calling format is also a problem if they intended to pass multiple
> > files to the command, or if they do so accidentally.
> >
> > Eg, an unprotected glob in the first argument, eg:
> > eawk foo* '/foobar/{$1="x"; print $0}'
> >
> > If foo1 and foo2 existed, this would be considered as:
> > awk foo2 /foobar/{$1="x"; print $0}' foo1
> >
> > which really will NOT have the desired effect.
>
> I am aware of this. I considered before:
>
>   a) eawk files... -- awk_args...
>   b) eawk awk_args -- pattern files...
>
> making '--' required, but that just complicates usage and function
> itself. And anyway isn't it true for any other command with careless
> glob use?
>
> I have documented that function takes a file, not files. The effect of
> misuse eawk would be awk error. Although maybe it would be useful to
> support multiple files? Then I'd go for variant (b) probably.
If we declare the calling format in BSD style:
<args> <files>

Then this is negated, but means you need to be able to find the first
file in the list, which requires parsing the options correctly.

Another variation that you haven't considered above is passing multiple
programs to awk via -e or -f options, in which case there is no
non-option program argument.

I agree generally that this is a useful idea, but there's a lot of
caveats and corner-cases :-(.

-- 
Robin Hugh Johnson
Gentoo Linux: Dev, Infra Lead, Foundation Trustee & Treasurer
E-Mail   : robbat2@g.o
GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85
GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136