a better way to exclude items other than a series of "grep -v" piped commands

[TwoProc]

This may be an awk question, a perl question, or a grep or sed problem...

I've got a list of file name extensions I don't want to see, and I want everything else from a "find" list.

e.g.

find . -type f | grep -v "\.xml" | grep -v "\.gif" | grep -v "\.htm" | grep -v "\.html"

...

and the list goes on for about 50 file extensions.

Can anyone please demonstrate a better way to exclude a list from a find than spawning a bazillion greps?

Note: handling it all in a single grep from the command line in the following fashion with something like "grep -e -v "\.gif" -e -v "\.htm" doesn't work.

Note2: I'd be find with putting all of the extensions in an input file of some sort, so that part doesn't have to come from the command line itself.

Thanks in advance

[Patrick Wallek · Accepted solution]

A couple of things come to mind:

1) Do it all with find:
# find . -type f \( ! -name "*.xml" -o ! -name "*.gif" -o ! -name "*.htm" -o ! -name "*.html" \)

Just keep listing your extensions. The '-o' means a 'logical or' and the '! -name "*.htm"' means NOT anything ending with .htm. Do a 'man find' for more info.

2) You could do it with find and grep:
# find . -type f | grep -v -E "\.xml|\.gif|\.htm|\.html"

[James R. Ferguson]

Hi John:

Put the patterns you don't want returned in your selection in a file; something like:

# cat /tmp/excludes
.awk
.c
.htm
.html
.log
.old
.pl
.pm
.sh

Now do:

# find /tmp -type f |grep -E -v -f /tmp/excludes

Regards!

...JRF...

[James R. Ferguson]

Hi (again) John:

Oops! In your exclusive file, if you truly want to skip files with dot (".") suffixes, escape the dot in the exclusion specification. The file contains regular expressions and a dot signifies any character when not escaped. My example '/tmp/excludes' should have looked like:


# cat /tmp/excludes
\.awk
\.c
\.htm
\.html
\.log
\.old
\.pl
\.pm
\.sh

Regards!

...JRF...

[TwoProc]

Patrick and James,

Thanks for the suggestions, I really like the ability to use the file with "grep -v -E -f" command -that exactly what I'm looking for.

Patrick - you're suggestion number #2 is quite cool, and I had BEEN wondering how to do an "or" in a grep for a word list *and* use it in an exclude grep function. I had only used it (and then rarely) before for include functions, and somehow never put 2+2 together to make it work for an exclude function. Thank you.

Upon reflection, the list method is going to win the day, because I want to account for all file extensions in the intended directory, and it will go into the hundreds of named extensions.

Thanks very much for the suggestions guys,

John

[Hein van den Heuvel]

And to get that list of file name extensions going you might want to use:

perl -le 'while (<*>) { m/\.(.*)/; $seen{$1}++} print "\\.$_" foreach (sort keys %seen)'

Hein.

[mobidyc]

Hello,

# find . -type f | grep -v "\.xml" | grep -v "\.gif" | grep -v "\.htm" | grep -v "\.html"

it's very ugly ;)

prefer a shortcut:
# touch foo.xml foo.txt foo.htm foo.html foo.avi foo.dat

# find . -type f |egrep -v "\.xml|\.gif|\.htm?"
./foo.txt
./foo.avi
./foo.dat
#

and better is:
# G_ARGS=".\xml|.gif|\.htm?"
# find . -type f |egrep -v "$G_ARGS"
./foo.txt
./foo.avi
./foo.dat
#

Regards,
Cedrick Gaillard

[OldSchool]

Hmm..
# cat /tmp/excludes
\.awk
\.c
\.htm
\.html
\.log
\.old
\.pl
\.pm
\.sh

You may want to "anchor" the end of line as well, esp if you want to exclude .htm but *not* .html's

something like:

\.htm$
\.log$

[James R. Ferguson]

Hi (again) John:

> OldSchool: You may want to "anchor" the end of line as well, esp if you want to exclude .htm but *not* .html's

Yes, absolutely true! I'm afraid that I was in too much of a hurry to catch a favorite television show and assumed the appropriate nuances applied. ;-)

Regards!

...JRF...

[TwoProc]

OldSchool - thanks for catching the fine point of end of line - much appreciated.

Hein -> I finished a script to do this as the initial part of the problem, but I'm intrigued by your perl script - because I've been slowly learning perl. I'm betting I will be spending quite a bit of time trying to figure out how it works, thank you. I really hate to ask - BUT... would you mind posting a little synopsis on the parts of that script and how it works, so I could learn? If you don't have the time, or its too lengthy of a request, I certainly understand, so feel free to ignore the request. Thanks again all.

[Hein van den Heuvel]

John,

It's actuallly no problem as I had it already written up two weeks ago.
And it is not perfect....

http://dcl.openvms.org/stories.php?story=07/03/22/8049141

There are more differences with the VMS version that I did not take care of.
- VMS needs to special case ";" as file version seperator
- and VMS typically is case-blind for filenames (Unless ODS-5 is used)
- VMS is strict about extentions allowing just 1 dot in the file name (unless ODS-5).

=> The ; provided a 'right side' anchor for the regexp. Need to use $ for "eol" under Unix.
=> Need to allow for multiple dots.

This gives:

$ perl -le 'while (<*>) {$seen{$1}++ if /\.([^.]*$)/} print "\\.$_" foreach (sort keys %seen)'

Explanation:

-l = print new line with each print

-e = program text to follow

while (<*>) { = loop over 'globbed' list of files, putting filename in automatic variable $_

$seen{uc($1)}++ = Increment (and create) an associative array elemement with name being last match from $1

if /\.([^.]*$)/ = Only do the aforementioned on a match of a piece of string to be called $1 with 'any non dot' after a period and before end of line. The first period is escaped with a backslash to make it real, not a wild character

} = loop end

print = print the default variable $_

foreach (sort = loop over sorted array from...

keys %seen = all the keys for associative array 'seen', stashed in default variable $_ one at a time.

Regards, Hein van den Heuvel

[TwoProc]

Hein, thank you very much for the posting, now I've got something to go to the books with and review. That was graciously posted, thank you.

I was laughing to myself looking over the code aspects as you've written them; because had I written them in Perl, they would have come out being a) MUCH longer, and b) looking like someone who is used to writing in both C and ksh "wrote a combination C/ksh script in perl".
:-)

[Dennis Handly]

>handling it all in a single grep from the command line ...

Your only problem was you only need one -v and the -e must be right before each pattern.
And if you didn't want to quote those ".", use fgrep.

>Patrick: # find . -type f | grep -v -E "\.xml|\.gif|\.htm|\.html"

There is no reason to use the egrep hammer.
Just use -f for that file solution. Or use multiple -e:
grep -v -e "\.xml$" -e "\.gif$" ...

[TwoProc]

Denis,

thanks for your response - however a bit of clarification here is needed. I said that the "grep -v -e" doesn't work, and it doesn't , at least in HPUX 11i (might work in the newer ones).

So, while the "grep -v -e" in repetition works fine in Linux, it does not on HPUX:

example:

$ cat > test
ehlllo
goodbye
bye
hello
no
yes
sayit
say
yell
scream
ice cream
ice
cream

$ cat test | grep -v -e "test" -v -e "yell" -v -e "ice"
ehlllo
goodbye
bye
hello
no
yes
sayit
say
yell
scream
ice cream
ice
cream

Notice that no lines are missing:

HOWEVER, on Linux the above test works as expected:

$ cat test | grep -v -e "test" -v -e "yell" -v -e "ice"
ehlllo
goodbye
bye
hello
no
yes
sayit
say
scream
cream

Which is why I put the question in the HPUX forum...

[James R. Ferguson]

Hi John:

The '-v' switch needs to occur *once*:

# grep -v -e "test" -e "yell" -e "ice" file

By the way, you can skip the extra process (the 'cat') and let 'grep' open the file(s) specified as its argument(s) as above.

Regards!

...JRF...

[OldSchool]

"Note: handling it all in a single grep from the command line in the following fashion with something like "grep -e -v "\.gif" -e -v "\.htm" doesn't work."

Actually, as noted above, it should be:

grep -v -e "\.gif" -e "\.html" .....

w/o repeating "-v".

Unfortunately, Linux isn't unix, its a work-alike, developed from observed behaviour / documentation of unix + "enhancements"

[TwoProc]

Well then,
it looks like the Linux version has figured out how to ignore the repeatiing series of "-v" to make it work...

thanks all for the clarification,
and I see that re-reading Dennis' post, he didn't repeat the "-v" over and over again.

Sorry Dennis, wish they had a do-over button on the points.

[Dennis Handly]

>I see that re-reading Dennis' post, he didn't repeat the "-v" over and over again.

I would assume you could repeat it and work, but I'm lazy. You just can't put the -v after the -e.

Ah, you're right, there is a bug in grep. They just increment vflag then they do bit stuff on it.

>Sorry Dennis, wish they had a do-over button on the points.

Ok, you can add the rest here. :-)
So you don't feel short changed, I filed a bug report on it:
CR JAGag37626:
Multiple -v in grep cause all to be ignored