bash.viki

* redirection
#Quote <<---
>> {
>>   echo | exp file=/dev/fd/3 useri... 3>&1 >&4 |
>>     gzip | split -n 2047m - exp.dmp.

>> } 4>&1

> Why needs '>&4' and '4>&1' here? What do they do?

4>&1 dupplicates the outer stdout to fd 4, >&4, aka 1>&4
restores "exp's" stdout to that same resource. So, the overall
effect is to preserve exp's stdout. Otherwise, anything written
by "exp" on its stdout would go to the pipe to gzip as well.

If you want to make things tidy, you could do:
{
  echo 4>&- | exp file=/dev/fd/3 useri... 3>&1 >&4 4>&- |
    gzip 4>&- | split -n 2047m - exp.dmp. 4>&-

} 4>&1

To close that fd 4 fall all the commands as none of them use it
(nor should they use it).
---

note: pipeline connect pipe-item's stdin/stdout before each redirection
_the 'outer' fd 4, is the same (or duplicate of) the 'inner' fd 4._ I THINK.


* get the string form "..."
data:
#Quote <<---
WARNING: "src/third_party/WebKit/WebKitLibraries" is no longer part of
this client.  It is recommended that you manually remove it.

WARNING: "src/third_party/GTM" is no longer part of this client.  It
is recommended that you manually remove it.

...
---

methods:

#Quote <<---
awk -v RS=\" '!(NR%2)'

cut -d\" -sf2 filename

awk -F\" '/"/{print $2}'

awk -F\" '/".*"/{print $2}'

sed -ne 's/.*"\([^"]*\)".*/\1/p' input

awk 'match($0, /"[^"]*"/) { print substr($0, RSTART + 1, RLENGTH - 2); }'

---

http://groups.google.com/group/comp.unix.shell/browse_thread/thread/bc8a8404d945bfa7#


* a function for running commands
I have a function to run commands and check the results:
#Quote <<----
run_or_die()
{
    echo -n "Try to run '$@'... "
    eval "$@"

    [[ $? -eq 0 ]] && echo 'OK!' && return

    echo 'Failed!' 
    clean_up
    report_result "${TEST}" "FAIL" 0
    exec 1>&6 2>&7 
    cat "$OUTPUTFILE"
    exit -1
}
----

and then I find another similar function from here: http://groups.google.com/group/comp.unix.shell/browse_thread/thread/46b7ecaa924554c8
as this one:
#Quote <<----
run_cmd() {
        cmd=$1
        shift
        echo 'Command line: ' "$cmd" ${1+"$@"} # has paramters if $1 being set (null or no-null)
        "$cmd" ${1+"$@"}

} 
----

But the second method cannot run a pipeline!!

* read: seems always discards the spaces at the end of input, except ...

#Quote <<----
[hpt@localhost temp]$ read buf <<< "hello,world    ";echo "buf is: '$buf'"
buf is: 'hello,world'
[hpt@localhost temp]$ IFS= read buf <<< "hello,world    ";echo "buf is: '$buf'"
buf is: 'hello,world    '
[hpt@localhost temp]$ read  <<< "hello,world    ";echo "REPLY is: '$REPLY'"
REPLY is: 'hello,world    '
----

* Why is testing x$foo = "x" better than $foo = ""
#Quote <<----
Prepeding a dummy safely prevents the values from being mistaken as
operators.  Otherwise, a few very special values like "=" and "!"
become a problem in some shells:

[ -n "=" ]      Bourne shell and ksh88 give the = a higher priority
                and expect a comparison

[ "!" = "x" ]   Bourne shell, ksh88 and some earlier dash (e.g. 0.4.26)
                expect a negated comparison

BTW, test can become even more fragile, if the seldomly used operator -a
and the value "=" are used (instead of connecting two separate conditions)

[ -n "=" -a {another condition} ]   breaks in almost all shells,
                                    one exception is pdksh


http://www.in-ulm.de/~mascheck/various/test/
----
and 
#Quote <<----
You're on the portable and safe side, if you can rewrite -a/-o with &&/||.
But problems are less likely if you prepend dummies and avoid negation
(which I forgot at first).

Negation is ambiguous: "! condition -a condition" can be parsed as
    ( ! condition ) -a condition
or
    ! ( condition -a condition ) 
----
* Get the third line based on a patterned line

Hongyi Zhao wrote:

> Hi all,
> 
> I want to obtain the third line before each matched line.  For
> example:
> 
> ...
> _here_comes_line_a
> _here_comes_line_b
> _here_comes_line_c
> _here_comes_the_matched_line
> ...
> _here_comes_line_d
> _here_comes_line_e
> _here_comes_line_f
> _here_comes_the_matched_line
> ...
> 
> I want to based on the regex matched lines to the third line before
> each matched line, i.e., the  _here_comes_line_a and
> _here_comes_line_d in the above example.  Any hints on this issue?

** method 1
awk '/matched_line/{print s1}{s1=s2;s2=s3;s3=$0}'

** method 2
awk '/matched/{print a[NR-3]}{a[NR]=$0}' file # this one will unnecessarily memorize the whole file...

** method 3
awk '/matched/{print a[NR%3]}{a[NR%3]=$0}'

** method 4
$ tac inputfile | awk '/matched_line/{i=4}!--i' | tac

** method 5
txr 

**