if the file is not found the files is not found is the file not found

if the file is not found the files is not found is the file not found

[meino.cramer]

Hi,

 ... the cat has bitten into its own tail... somehow...


I wrote a script which handles dvbt streamed files. The process
creates some temporary files, which I want to remove afterwards,
because they normally big ones. 

First I wrote

    rm -f ${f}-[0-9]*.mp2 

which breaks which an error, if the certain has not created files
of that pattern ... despite the "-f" of the "rm" command.

Then I treid to check for the existence of such files in beforehand
this way:

     [ -f ${f}-[0-9]*.mp2 ]] && rm -f ${f}-[0-9]*.mp2

. Which fails for the same reason.

Did I get lost here? ;)

How many cats do I need to get one, which does not bit into
the tail of the next cat I enter into the script ?

Who knows of the according dog to chase the cats away and make
my script work? ;) :))

Thanks a lot for any help in advance!

Best regards,
mcc

Re: if the file is not found the files is not found is the file not found

[Moritz Bunkus]

Hey,

On Sun, Mar 4, 2012 at 15:31,  <meino.cramer@gmx.de> wrote:

> First I wrote
>
>    rm -f ${f}-[0-9]*.mp2
>
> which breaks which an error, if the certain has not created files
> of that pattern ... despite the "-f" of the "rm" command.

The error probably comes from zsh that it cannot match any file. Try
again with "setopt nullglob"; in that case the pattern not matching
anything will simply be replaced by nothing and the command executed.

Kind regards,
mo

Re: if the file is not found the files is not found is the file not found

[Mikael Magnusson]

On 4 March 2012 15:31,  <meino.cramer@gmx.de> wrote:
> Hi,
>
>  ... the cat has bitten into its own tail... somehow...
>
>
> I wrote a script which handles dvbt streamed files. The process
> creates some temporary files, which I want to remove afterwards,
> because they normally big ones.
>
> First I wrote
>
>    rm -f ${f}-[0-9]*.mp2
>
> which breaks which an error, if the certain has not created files
> of that pattern ... despite the "-f" of the "rm" command.
>
> Then I treid to check for the existence of such files in beforehand
> this way:
>
>     [ -f ${f}-[0-9]*.mp2 ]] && rm -f ${f}-[0-9]*.mp2
>
> . Which fails for the same reason.
>
> Did I get lost here? ;)
>
> How many cats do I need to get one, which does not bit into
> the tail of the next cat I enter into the script ?
>
> Who knows of the according dog to chase the cats away and make
> my script work? ;) :))
>
> Thanks a lot for any help in advance!

If you setopt extendedglob you can append (#qN) to the pattern to
suppress the error for that specific pattern, or setopt nullglob to
always do it. ie, it would be ${f}-[0-9]*.mp2(#qN)

-- 
Mikael Magnusson

Re: if the file is not found the files is not found is the file not found

[Bart Schaefer]

On Mar 4,  3:31pm, meino.cramer@gmx.de wrote:
}
}      [ -f ${f}-[0-9]*.mp2 ]] && rm -f ${f}-[0-9]*.mp2

This won't work even if the files do exist, because you can't apply
a single "-f" test to the multiple files that result from the glob.

Also you've used "[" on the left but "]]" on the right, which is
mostly nonsense.  If you instead used [[ on the left then the glob
would not be expanded and the test would again fail.

On Mar 4,  3:37pm, Mikael Magnusson wrote:
}
} If you setopt extendedglob you can append (#qN) to the pattern

You don't even need extendedglob -- you just need bare_glob_qual, which
is on by default unless you're in sh/ksh emulation modes:

    rm -f ${f}-[0-9]*.mp2(N)

I'm sure the archives of zsh-users hold many different answers to the
question, "Given a file pattern, how do I test whether at least one
matching file exists?"  Unfortunately, there's really no way to do so
in a single operation unless you set the no_nomatch option.  [[ ]]
does not perform globbing, and the test operators such as [ -f ] are
defined to return TRUE rather than FALSE on a *missing* file name
operand, so null_glob is not sufficient.

And yet we've never added a glob qualifier to invert nomatch (nor to
invert bad_pattern) for a single glob ... however, we did invent
anonymous functions, so you can make multiple operations look like a
single operation:

if (){ setopt localoptions no_nomatch; [ -f ${f}-[0-9]*.mp2([1]) ]; }
then rm -f ${f}-[0-9]*.mp2
fi

The "if (){ ... }" does not define a function named "if" because "if"
is a reserved word.  The ([1]) qualifier on the end of the pattern
extracts only the first matching file so that you aren't passing too
many arguments to "[ -f ... ]".

If you don't like that syntax and don't mind an extra process, you can
also do this with a subshell:

if ( setopt no_nomatch; [ -f ${f}-[0-9]*.mp2([1]) ]; )
then rm -f ${f}-[0-9]*.mp2
fi

Woof woof.

-- 
Barton E. Schaefer

Re: if the file is not found the files is not found is the file not found

[Mikael Magnusson]

On 4 March 2012 19:37, Bart Schaefer <schaefer@brasslantern.com> wrote:
> On Mar 4,  3:31pm, meino.cramer@gmx.de wrote:
> }
> }      [ -f ${f}-[0-9]*.mp2 ]] && rm -f ${f}-[0-9]*.mp2
>
> This won't work even if the files do exist, because you can't apply
> a single "-f" test to the multiple files that result from the glob.
>
> Also you've used "[" on the left but "]]" on the right, which is
> mostly nonsense.  If you instead used [[ on the left then the glob
> would not be expanded and the test would again fail.
>
> On Mar 4,  3:37pm, Mikael Magnusson wrote:
> }
> } If you setopt extendedglob you can append (#qN) to the pattern
>
> You don't even need extendedglob -- you just need bare_glob_qual, which
> is on by default unless you're in sh/ksh emulation modes:
>
>    rm -f ${f}-[0-9]*.mp2(N)

Ah, I've had them both on since forever and knew extendedglob wasn't
on by default, so I assumed bareglobquals was also off.

> I'm sure the archives of zsh-users hold many different answers to the
> question, "Given a file pattern, how do I test whether at least one
> matching file exists?"  Unfortunately, there's really no way to do so
> in a single operation unless you set the no_nomatch option.  [[ ]]
> does not perform globbing, and the test operators such as [ -f ] are
> defined to return TRUE rather than FALSE on a *missing* file name
> operand, so null_glob is not sufficient.

Here's one more for the collection,
if () { (( $# )) } arglblargh*(N[1]); then echo yes; else echo no; fi

But this needs the 'new' anon functions with arguments support.

-- 
Mikael Magnusson

Re: if the file is not found the files is not found is the file not found

[Bart Schaefer]

On Mar 4,  7:48pm, Mikael Magnusson wrote:
>
> On 4 March 2012 19:37, Bart Schaefer <schaefer@brasslantern.com> wrote:
> > I'm sure the archives of zsh-users hold many different answers to the
> > question, "Given a file pattern, how do I test whether at least one
> > matching file exists?"
> 
> Here's one more for the collection,
> if () { (( $# )) } arglblargh*(N[1]); then echo yes; else echo no; fi

Yes, I was thinking about that but it doesn't capture the "is a plain
file" semantics of [[ -f ]] -- which you can fix by adding qualifiers
to the glob instead, of course -- and even my formulation falls down
if the glob matches a mix of plain and not-plain files and the first
one happens to be the wrong kind.

(Also, the [1] in your formula is extraneous, but that's a nit.)

Really what one means with [ -f foo* ] is usually one of

# One existing file matching foo* is a plain file
(){ local f; for f; do [ -f "$f" ] && return 0; done; return 1; } foo*(N)

or

# All files matching foo* exist and are plain files
(){ local f; for f; do [ -f "$f" ] || return 1; done; (( $# )); } foo*(N)

Still another way to do "one foo* is plain":

(){ (( $# )) } foo*(Ne:'[ -f "$REPLY" ]':)

I'm not sure there's a way to use (e::) for "every foo* is plain".

-- 
Barton E. Schaefer

Re: if the file is not found the files is not found is the file not found

[Mikael Magnusson]

On 4 March 2012 20:44, Bart Schaefer <schaefer@brasslantern.com> wrote:
> On Mar 4,  7:48pm, Mikael Magnusson wrote:
>>
>> On 4 March 2012 19:37, Bart Schaefer <schaefer@brasslantern.com> wrote:
>> > I'm sure the archives of zsh-users hold many different answers to the
>> > question, "Given a file pattern, how do I test whether at least one
>> > matching file exists?"
>>
>> Here's one more for the collection,
>> if () { (( $# )) } arglblargh*(N[1]); then echo yes; else echo no; fi
>
> Yes, I was thinking about that but it doesn't capture the "is a plain
> file" semantics of [[ -f ]] -- which you can fix by adding qualifiers
> to the glob instead, of course -- and even my formulation falls down
> if the glob matches a mix of plain and not-plain files and the first
> one happens to be the wrong kind.
>
> (Also, the [1] in your formula is extraneous, but that's a nit.)

I like to imagine that it causes less data to be copied around. :)

-- 
Mikael Magnusson

Re: if the file is not found the files is not found is the file not found

[Bart Schaefer]

On Mar 4,  9:13pm, Mikael Magnusson wrote:
}
} >> if () { (( $# )) } arglblargh*(N[1]); then echo yes; else echo no; fi
} >
} > (Also, the [1] in your formula is extraneous, but that's a nit.)
} 
} I like to imagine that it causes less data to be copied around. :)

According to valgrind, for a pattern matching six files the ([1]) saves
3 mallocs and 3 frees but uses 346 *more* bytes of memory.  "time" on
10 runs of each shows no difference at all.

The ([1]) does begin to win on memory as well when the number of matched
files grows, but even at 360+ files no difference in time is detectable,
probably because the vast bulk is taken up by doing the glob in the first
place.

(e:'[ -f "$REPLY" ]':) takes about 3 times as long and uses 5 times as
much memory, with or without ([1]) added.  (e:'[[ -f "$REPLY" ]]') is
better, only double time and triple memory, and is on a par with adding
(.) as a qualifier.

So there you go. :-)

Re: if the file is not found the files is not found is the file not found

[Mikael Magnusson]

On 4 March 2012 22:47, Bart Schaefer <schaefer@brasslantern.com> wrote:
> On Mar 4,  9:13pm, Mikael Magnusson wrote:
> }
> } >> if () { (( $# )) } arglblargh*(N[1]); then echo yes; else echo no; fi
> } >
> } > (Also, the [1] in your formula is extraneous, but that's a nit.)
> }
> } I like to imagine that it causes less data to be copied around. :)
>
> According to valgrind, for a pattern matching six files the ([1]) saves
> 3 mallocs and 3 frees but uses 346 *more* bytes of memory.  "time" on
> 10 runs of each shows no difference at all.
>
> The ([1]) does begin to win on memory as well when the number of matched
> files grows, but even at 360+ files no difference in time is detectable,
> probably because the vast bulk is taken up by doing the glob in the first
> place.
>
> (e:'[ -f "$REPLY" ]':) takes about 3 times as long and uses 5 times as
> much memory, with or without ([1]) added.  (e:'[[ -f "$REPLY" ]]') is
> better, only double time and triple memory, and is on a par with adding
> (.) as a qualifier.
>
> So there you go. :-)

Heh, thanks for checking :). Btw,
% () { echo $# } *
2411
zsh -c 'repeat 1000 () { (($#)) } *'  6.53s user 0.40s system 98% cpu
6.999 total
zsh -c 'repeat 1000 () { (($#)) } *([1])'  5.97s user 0.38s system 99%
cpu 6.399 total
(numbers are consistent within 0.02s)

So maybe it doesn't save the time it takes to type the [1]. Adding oN
on the other hand,
zsh -c 'repeat 1000 () { (($#)) } *(oN)'  4.81s user 0.40s system 99%
cpu 5.252 total
zsh -c 'repeat 1000 () { (($#)) } *(oN[1])'  4.31s user 0.36s system
99% cpu 4.703 total

For those that don't know, (oN) disables sorting the results of the glob.

-- 
Mikael Magnusson