NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/misc-part2/10647

Internet Message Format | 1996-08-05 | 6.0 KB

Path: blue.cse.ogi.edu!leneis
From: [e-mail] (Tony Leneis)
Newsgroups: comp.sys.amiga.misc
Subject: Re: AmigaOS 4.x features
Date: 31 Mar 1996 05:56:11 GMT
Organization: Oregon Graduate Institute (OGI), Portland, Oregon
Message-ID: <4jl6pr$[e-mail]>
References: <4j7ein$[e-mail]> <4jfm8i$[e-mail]> <4jgc6q$[e-mail]> <4jkg5j$[e-mail]>
NNTP-Posting-Host: blue.cse.ogi.edu

In article <4jkg5j$[e-mail]>,
Charlie Moody <[e-mail]> wrote:
>On 29 Mar 1996 09:57:46 GMT, Tony Leneis wrote:
>:     I'll grant that * is easier than #?, but the Amiga pattern matching
>: is more powerful than traditional UNIX globbing.  As was pointed out earlier,
>: you can always enable a UNIX-ish * by using a utility like wildstar.
>
>It may be *as* powerful, but not moreso (and what the heck is 'UNIX
>globbing'?), and certainly not as well thought out.  And that's the point
>about enabling the '*':  it's easy, and not nearly as useful as it would
>be in context.

    First, globbing is filename wildcarding.  Since some UNIX programs
do regexp patterns and others do glob patterns, you need to make the
distinction.  Regexp pattern matching is much more powerful than globbing,
but it's not as convenient to express.  When I say that the Amiga pattern
matching is more powerful than globbing, I mean there are patterns that
Amiga pattern matching can express which globbing cannot.  For example:

test#a       <- Matches 'test', 'testa', 'testaa', 'testaaa', etc.
test#(ab)c   <- Matches 'testc', 'testabc', 'testababc', 'testabababc', etc.
~(#?.info)   <- Matches anything that doesn't end in '.info'
                * To be fair, a few UNIX shells provide this functionality
exp(ab|cd).c <- Matches 'expab.c' and 'expcd.c'
                * Again, _some_ UNIX shells provide this functionality

On the other hand, there are no UNIX globbing patterns which cannot be
expressed as an Amiga pattern.  Hence, Amiga patterns are more powerful.

>Yes, I can get the same effect as in can under unix pattern-matching...all
>I have to do is use a completely different, much more cumbersome, and much
>less intuitive (read: badly designed) method.

It's different, because it's essentially allowing you to do full regular
expressions.  Would you have preferred the standard UNIX regular expression
syntax?  Then you'd have to type '.*' instead of '#?', and you'd have to
remember to escape the '.' whenever you didn't want it to act as the Amiga
'?' metacharacter.  (Kind of a bummer if you're looking for '#?.c', or should
I say '.*\.c' as a UNIX regular expression.)

Where UNIX has two different pattern-matching systems that you have to learn,
the Amiga has one.  Seems much simpler to me.  (Never mind the fact that each
UNIX shell and utility seems to have it's own special little modification to
glob patterns and regexp patterns.)

>That's not "more flexible";  it's not "more powerful";  and it sure as
>h*ll ain't any easier.  Isn't it basic common sense to avoid re-inventing
>the wheel *whenever* possible?

As mentioned above, it is more flexible and more powerful, because anything
that can be expressed as a UNIX glob pattern can be expressed as an Amiga
pattern, and then some.  Just out of curiosity, whose wheel should they have
used instead of inventing their own?  As far as I know, there was no RFC or
other standard for file wildcarding when AmigaDOS was developed.  Which UNIX
shell should they have followed?  Or, should they have done it like VMS?  MVS?
MS-DOS?  Each of these does pattern matching in different ways.

>: For example, "dir sys:(a|b|c|e)#?".
>
>[additional references to documentation for more counter-intuitive,]
>[anti-logical methods deleted]
>
>        dir sys:[a-e]*
>
>Logical, simple, fast:  elegant.

    Well, you can certainly do it with a current version of AmigaDOS.
So, what's the problem?  One minor nit - that should be 'dir sys:[a-ce]*'
if you want it to be the same as the 1.3 AmigaDOS pattern matching example
above.  One thing I'd like to see added is the ability to compliment sets.
i.e. [^abc] matches anything except 'a', 'b', or 'c'.

>; The main problem,
>: if you want to call it a problem, is that it's different.  It also can be a
>: little more awkward to use for certain patterns, but it always can get the
>: job done.
>
>The fact that the kludge can be worked around does nothing to redeem the
>kludge.  It's still bad design with no significant benefit from making
>the sacrifices, and it should be thrown overboard at the first opportunity.

It's not a kludge, and there's nothing wrong with the design.  It is much
more useful than MS-DOS or UNIX globbing.  If I wanted that, I'd just run
Linux or CrossPC on my Amiga.

Throwing the Amiga pattern matching overboard would not only reduce the power
of AmigaDOS, it would also reduce the power of editors and other applicaitons
and utilities that allow you to do pattern matching.  I'd rather see Amiga
pattern matching extended and perhaps made more efficient (efficient as in
time it takes for the pattern-matching code to run.)

>: I'd certainly take the functionality of the # operator any day
>: over the UNIX * operator (provided I've got a match-any-character operator
>: to go with it.)
>
>You'd gladly throw away the match-any-character option, provided you've
>got a match-any-character option to replace it???

No.  The UNIX * operator is match-any-sequence-of-characters.  The UNIX ? and
the AmigaDOS ? operator is match-any-character.  The AmigaDOS # operator is
match-any-number-of.  So, I'd prefer the ability to match any number of
arbitrary patterns, provided I had a way to match any character, rather than
only being able to match any sequence of characters.

-Tony
--
=====///===Hi!==I'm=a=.signature=virus!==Copy=me=to=your=.signature!===///==
__  /// [e-mail]          From the desk of:   Hillary,  __  ///
\\\///  [e-mail]       -->Tony Leneis<--   Gennifer, \\\///Only
 \XX/   [e-mail]  HMC '91  Go Stags!  AND NOW US!\XX/Amiga