Prev: bash: how to transfer some non-ascii code
Next: very strange ksh behavior
From: Rakesh Sharma on 4 Apr 2010 05:37
On Apr 4, 7:14 am, Ben Bacarisse <ben.use...(a)bsb.me.uk> wrote:
> Rakesh Sharma <sharma...(a)hotmail.com> writes:
> > I want to match a string of numbers followed optionally by either a +
> > or a -.
> > The corner cases are: the number cannot begin with a 0 unless it's a
> > solitary 0.
>
> > So expressing this in terms of regular expressions I can write:
>
> >     [0][+-]? | [1-9][0-9]*[+-]?
>
> [0] is little odd!  You can make the ending more clear by re-writing
> this as:
>
>    (0|[1-9][0-9]*)[+-]?
>
> > So far so good, but how would I write the equivalent functionality in
> > terms
> > of shell file patterns of the kind found in the 'case'..'esac'
> > statments of bourne shell?
>
> bash can but not a POSIX shell.  You can get close if can limit the
> number of digits by simply writing out all the cases, but you can't
> even use [0-9] (you need [0123456789] instead) so the end result will
> be horrid and error prone.
>
> --
> Ben.

I came up with this, but not sure if it'll work: (yet to test it)

case $1 in
'' | *[!0-9+-]* | [+-] | *[+-]?* | [0][0-9]* )
echo >&2 "Incorrect arg: '$1'"; exit 1;;
*) :;;
esac


-- Rakesh
From: Janis Papanagnou on 4 Apr 2010 07:59
Rakesh Sharma schrieb:
> Hi,
>
> I want to match a string of numbers followed optionally by either a +
> or a -.
> The corner cases are: the number cannot begin with a 0 unless it's a
> solitary 0.
>
> So expressing this in terms of regular expressions I can write:
>
> [0][+-]? | [1-9][0-9]*[+-]?
>
> So far so good, but how would I write the equivalent functionality in
> terms
> of shell file patterns of the kind found in the 'case'..'esac'
> statments of bourne shell?

In addition to the already given answers; let me just point to the
expr(1) command that allows to check variables against BRE regexps
and you can conditionally branch according to its exit status.

BTW; are you really *required* to use a bourne shell? There are so
simple solutions in all modern shells.

Janis

>
> Thanks
>
>
> -- Rakesh
From: Ben Bacarisse on 4 Apr 2010 09:49
Rakesh Sharma <sharma__r(a)hotmail.com> writes:
<snip>
>> Rakesh Sharma <sharma...(a)hotmail.com> writes:
>> > I want to match a string of numbers followed optionally by either a +
>> > or a -.
>> > The corner cases are: the number cannot begin with a 0 unless it's a
>> > solitary 0.
<snip>
> I came up with this, but not sure if it'll work: (yet to test it)
>
> case $1 in
> '' | *[!0-9+-]* | [+-] | *[+-]?* | [0][0-9]* )
> echo >&2 "Incorrect arg: '$1'"; exit 1;;
> *) :;;
> esac

That looks right though it's not what you asked for!

--
Ben.
From: Rakesh Sharma on 4 Apr 2010 10:32
On Apr 4, 6:49 pm, Ben Bacarisse <ben.use...(a)bsb.me.uk> wrote:
> Rakesh Sharma <sharma...(a)hotmail.com> writes:
>
> <snip>
>
> >> Rakesh Sharma <sharma...(a)hotmail.com> writes:
> >> > I want to match a string of numbers followed optionally by either a +
> >> > or a -.
> >> > The corner cases are: the number cannot begin with a 0 unless it's a
> >> > solitary 0.
> <snip>
> > I came up with this, but not sure if it'll work: (yet to test it)
>
> > case $1 in
> >   '' | *[!0-9+-]* | [+-] | *[+-]?* | [0][0-9]* )
> >       echo >&2 "Incorrect arg: '$1'"; exit 1;;
> >   *) :;;
> > esac
>
> That looks right though it's not what you asked for!
>
> --
> Ben.


So what you mean is that this case..esac isn't equivalent to the
regular expression ( 0 | [1-9] [0-9]* ) [+-]?

-- Rakesh
From: Ben Bacarisse on 4 Apr 2010 11:41
Rakesh Sharma <sharma__r(a)hotmail.com> writes:

> On Apr 4, 6:49 pm, Ben Bacarisse <ben.use...(a)bsb.me.uk> wrote:
>> Rakesh Sharma <sharma...(a)hotmail.com> writes:
>>
>> <snip>
>>
>> >> Rakesh Sharma <sharma...(a)hotmail.com> writes:
>> >> > I want to match a string of numbers followed optionally by either a +
>> >> > or a -.
>> >> > The corner cases are: the number cannot begin with a 0 unless it's a
>> >> > solitary 0.
>> <snip>
>> > I came up with this, but not sure if it'll work: (yet to test it)
>>
>> > case $1 in
>> >   '' | *[!0-9+-]* | [+-] | *[+-]?* | [0][0-9]* )
>> >       echo >&2 "Incorrect arg: '$1'"; exit 1;;
>> >   *) :;;
>> > esac
>>
>> That looks right though it's not what you asked for!
>>
>> --
>> Ben.

Best not to quote sig blocks (even small ones).

> So what you mean is that this case..esac isn't equivalent to the
> regular expression ( 0 | [1-9] [0-9]* ) [+-]?

No. I was just explaining why you got negative answers. You
originally asked for a glob pattern than matched some set which was
not possible without some non-standard extensions. It turns out you
/can/ solve the practical problem by looking for every case that does
/not/ match, but that was not what you originally asked for.

Having "positive" patterns is, in general, more useful. They can be
used in places other than case...esac and they make maintaining a
larger set of case patterns much simpler. They are also much easier
to understand. For example, it is not immediately clear what patterns
fall through to the * case in your solution.

--
Ben.
First  |  Prev  | 
Pages: 1 2
Prev: bash: how to transfer some non-ascii code
Next: very strange ksh behavior