Using pipes inside of find -exec

Hi,

one more question related to LC: I'm trying to demonstrate
to some of my colleagues, how hopeless is trying to count
the lines of code in C/C++ files by just parsing them (i.e.
without invoking the "gcc -E" to remove comments and #if 0).

So I'm trying to extract simple strings out of C/C++ files:

boclu23:r0923 {106} find . \( -name \*.cpp -o -name \*.c \) -print | xargs sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' | head
"st.h"
"global.h"
"product.h"
"type_def.h"
"mac_def.h"
"hw.h"
"os.h"
"pn.h"
"cm.h"
"cm_time.h"


And then grep those extracted strings for the characters "/*" :

boclu22:r0923 {153} find . \( -name \*.cpp -o -name \*.c \) -print | xargs sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' | grep '/\*' | head
"/* This file is generated %s by\n** '%s'.\n"
" %d, /* PMM_INDEX_T base; */"
"\n { /* byte region_model[HW_FLASH_DEVICES]; */\n"
" { /* PMM_CATEGRY_STR categories[PMM_NF_CATEGRIES]; */\n"
"#endif /* PMM_FLASH_DEF_H */\n"
"\n\n/* End of file */\n"
"#define %-25s 0x%.2x /* %d */\n"
"/* routing table. Maps Id's onto task id's */"
" 0xFF , /* Empty */\n"
"/* Declaration of functions to appear in the function routing table */"


So there are quite a few files with weird strings. And here is
my problem: how could I print the names of the files who contain
the weird strings (i.e. with embedded comments)? I try:

boclu23:r0923 {116} find . \( -name \*.cpp -o -name \*.c \) -print -exec sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' {} \| grep '/\*' \;

but get the sed error messages because it doesn't recognize the pipe:

sed: can't read |: No such file or directory
sed: can't read grep: No such file or directory
sed: can't read /\*: No such file or directory

And if I don't escape the pipe like this, then the find complains:

boclu23:r0923 {116} find . \( -name \*.cpp -o -name \*.c \) -print -exec sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' {} | grep '/\*' \;
find: missing argument to `-exec'
grep: ;: No such file or directory


So the questions are:

1) is it possible to use pipes in "find -exec" and how?
2) is there a way to print the filename before sedding it?

I'm using RH Linux and thus GNU tools but would be interested
in a portable solution too. (And yes I know I could do it all
in Perl, but I'd like to learn some nice shell oneliners)

Thank you!
Alex

In article <3F1D0552.F2AA33C1@nokia.com>,
A. Farber <Alexander.Farber@nokia.comwrote:
>So the questions are:
>
>1) is it possible to use pipes in "find -exec" and how?

Yes, if you run a shell via -exec. The simplest way is to put the pipeline
in a shell script, and then specify that scriptname as the command in the
-exec option, e.g.

find -exec myscript {} \;

But if you really *must* do it as a one-liner, you can do:

find -exec sh -c 'some_command $1 | other_command' {} {} \;

>2) is there a way to print the filename before sedding it?

find -print -exec

--
Barry Margolin, barry.margolin@level3.com
Level(3), Woburn, MA
DN'T SEND TECHNICAL QUESTINS DIRECTLY T ME, post them to newsgroups.
Please DN'T copy followups to me -- I'll assume it wasn't posted to the group.

I'm a bit confused

If :

find . \( -name \*.cpp -o -name \*.c \) -print | xargs sed -ne
's/.*"\([^"]*\)\".*/"\1"/p'

outputs:

"st.h"
"global.h"
"product.h"
"type_def.h"
"mac_def.h"
"hw.h"
"os.h"
"pn.h"
"cm.h"
"cm_time.h"

How come you get any results when you pipe the output into grep
looking for /* ?

Anyhow, I don't know how/if you can pipe commands within "find -exec".
You can use exec multiple times though (-exec sed '' -exec grep
"") but that's not what you are looking for is it?

If what you are trying to achieve is to extract the filenames above and
to grep those files for your commented strings - why not just pass the
results of your find as a list of files to grep?

e.g.

grep "/\*" $(find . -exec sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' {} \;)

This should output the filename and the grep'd string together.

Am I understanding the problem correctly?

--
Posted via http://dbforums.com

"A. Farber" <Alexander.Farber@nokia.comwrote in message news:<3F1D0552.F2AA33C1@nokia.com>
Hi,
>
one more question related to LC: I'm trying to demonstrate
to some of my colleagues, how hopeless is trying to count
the lines of code in C/C++ files by just parsing them (i.e.
without invoking the "gcc -E" to remove comments and #if 0).
>
So I'm trying to extract simple strings out of C/C++ files:
>
boclu23:r0923 {106} find . \( -name \*.cpp -o -name \*.c \) -print | xargs sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' | head
"st.h"
"global.h"
"product.h"
"type_def.h"
"mac_def.h"
"hw.h"
"os.h"
"pn.h"
"cm.h"
"cm_time.h"
>
>
And then grep those extracted strings for the characters "/*" :
>
boclu22:r0923 {153} find . \( -name \*.cpp -o -name \*.c \) -print | xargs sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' | grep '/\*' | head
"/* This file is generated %s by\n** '%s'.\n"
" %d, /* PMM_INDEX_T base; */"
"\n { /* byte region_model[HW_FLASH_DEVICES]; */\n"
" { /* PMM_CATEGRY_STR categories[PMM_NF_CATEGRIES]; */\n"
"#endif /* PMM_FLASH_DEF_H */\n"
"\n\n/* End of file */\n"
"#define %-25s 0x%.2x /* %d */\n"
"/* routing table. Maps Id's onto task id's */"
" 0xFF , /* Empty */\n"
"/* Declaration of functions to appear in the function routing table */"
>
>
So there are quite a few files with weird strings. And here is
my problem: how could I print the names of the files who contain
the weird strings (i.e. with embedded comments)? I try:
>
boclu23:r0923 {116} find . \( -name \*.cpp -o -name \*.c \) -print -exec sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' {} \| grep '/\*' \;
>
but get the sed error messages because it doesn't recognize the pipe:
>
sed: can't read |: No such file or directory
sed: can't read grep: No such file or directory
sed: can't read /\*: No such file or directory
>
And if I don't escape the pipe like this, then the find complains:
>
boclu23:r0923 {116} find . \( -name \*.cpp -o -name \*.c \) -print -exec sed -ne 's/.*"\([^"]*\)\".*/"\1"/p' {} | grep '/\*' \;
find: missing argument to `-exec'
grep: ;: No such file or directory
>
>
So the questions are:




1) is it possible to use pipes in "find -exec" and how?

find . -type f \( -name \*.c -o -name \*.cpp \) \
-exec sh -c 'grep "[^\"]*" "$1" | sed -e "/\/\*/!d"' {} {} \;




2) is there a way to print the filename before sedding it?

find . -type f \( -name \*.c -o -name \*.cpp \) \
-exec sh -c 'grep "[^\"]*" "$1" /dev/null |
sed -e \
"/^\([^:]*:\).*\(\"[^\"]*\"\).*\$/s//\1\2/;/\/\*/!d"
' {} {} \;

(assuming that the filenames have no ':' character in them anywhere)