[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gnu-arch-users] Re: two UI usability tweak suggestions
From: |
Tom Lord |
Subject: |
Re: [Gnu-arch-users] Re: two UI usability tweak suggestions |
Date: |
Fri, 19 Sep 2003 13:09:28 -0700 (PDT) |
> From: Miles Bader <address@hidden>
> I'd like _all_ revision arguments to allow `patch-N', where it makes
> sense (e.g. replay), presumably via a change to arch_parse_package_name
> -- e.g., add a `default version' arg, which could be either 0 (requiring
> fully qualified names) or a version to use for `patch-N' names.
It's harder than just a patch to arch_parse_package_name.
arch_parse_package_name has no context (other than its parameters)
about context. It has no default version to refer to.
Moreover, arguments to arch_parse_package_name have generally passed a
check of arch_valid_package_name.
So, what you are describing is a pervasive change -- and one that will
require tweaks in many cmd-*.c files.
I did, in larch, play around with forming higher-level abstractions
for argument validation, default-filling-in-ness, and parsing. Please
note that I didn't carry that over to tla (a cost/benefit decision).
(E.g., in larch there is the concept of "indicated_revision"....)
You _can_ do it globally. But I weakly doubt you can do it well
globally in a weekend.
I did have a social engineering goal when I decided to make the
validation and parsing in the cmd-*.c files roughly duplicate code
rather than seek out the perfect high-level abstractions. The goal
was to allow the cmd-*.c files to evolve independently and
irregularly. Part of my thinking, interpreted in this context, is
that people would make patches to add support for something like
`patch-N-as-revision-name' in a case-by-case manner. They're easy to
pick off one-by-one -- so I was going for fine-grained itches and
fine-grained solutions. The other part of my thinking was that
irregularity is what distinguishes awkward artificial-feeling
languages from natural fits-my-brain languages (the natural languages
displaying irregularity) -- I decided that _some_ aspects of perfect
regularity in the command set was likely to be a misfeature.
> BTW, if you don't want to make any of these tweaks, I can probably do
> them this weekend (it's nice to know one way or the other though, to
> avoid conflicting patches).
Overall, your patches so far are a big hit from the feedback I get.
So, do what you like -- just trying to give some hopefully helpful
hints.
-t