os_explain

#!/bin/bash
################################################################################
## Copyright 2018 Tanel Poder. All rights reserved. More info at http://tanelpoder.com
## Licensed under the Apache License, Version 2.0. See LICENSE.txt for terms & conditions.
##
## File name:   os_explain (version 1.1)
## Purpose:     Script to explain the currently active branch of SQL plan
##              execution from Oracle server process stack trace
##
## Author:      Tanel Poder
## Copyright:   (c) http://www.tanelpoder.com
##
## Usage:       1) Take a stack trace of an Oracle process (either by using a
##                 a debugger, pstack or extract the stack from a corefile,
##                 save it to a file (pstack.txt for example)
##              2) run ./os_explain <stack_trace_file>
##                 For example: ./os_explain pstack.txt
##
##
##
##              Alternatively you can pipe pstack output directly to os_explain
##              STDIN:
##
##                 pstack <SPID> | ./os_explain
##
## Other:       Get stack using pstack on Solaris and Linux, using procstack
##              on AIX and by using pstack on recent versions of HP-UX.
##              Older HP-UX versions may require a debugger for getting stack
##              from OS. As an alternative (especially on Windows) you could
##              use ORADEBUG DUMP ERRORSTACK for dumping stack of a process.
##              Note that ORADEBUG's stack dump mechanism is not 100% safe
##              on active processes
##
##              Most of Oracle kernel function prefix translations are based
##              on Metalink note 175982.1 (google for it, it's gone from the
##              Oracle site)
##
##              On Solaris you may need to replace awk with nawk in this script
##              I don't remember why anymore, but for some reason I used it
##
################################################################################

# LIFO line reverser
f_lifo() {

        #echo Entering f_lifo()

        MAX_ARRAY=1023 # Max stack array depth. You can use 4095 here but older linuxes and hp-ux shell may not like it 

        i=$MAX_ARRAY

        IFS=^
        while read input ; do
            array[i]=$input
            ((i=i-1))
        done

        i=1

        for output in ${array[@]} ;  do

            if test "$1" = "-n" ; then
                 printf "%4d %s\n" $i $output
            else
                 printf "%s\n" $output
            fi
            ((i=i+1))
        done

}


TRANSLATION_STRING='
/-----------------  lwp# 2/,$d;
s/(.*//;
s/ [[0-9][a-f]]{16} /./;
s/qerae/AND-EQUAL: /g;
s/qerba/BITMAP INDEXAND : /g;
s/qerbc/BITMAP INDEX COMPACTION: /g;
s/qerbi/BITMAP INDEX CREATION: /g;
s/qerbm/MINUS: /g;
s/qerbo/BITMAP INDEX OR: /g;
s/qerbt/BITMAP CONVERT: /g;
s/qerbu/BITMAP INDEX UNLIMITED-OR: /g;
s/qerbx/BITMAP INDEX ACCESS: /g;
s/qercb/CONNECT BY: /g;
s/qercbi/SUPPORT FOR CONNECT BY: /g;
s/qerco/COUNT: /g;
s/qerdl/DELETE: /g;
s/qerep/EXPLOSION: /g;
s/qerff/FIFO BUFFER: /g;
s/qerfi/FIRST ROW: /g;
s/qerfl/FILTER DEFINITION: /g;
s/qerfu/UPDATE: /g;
s/qerfx/FIXED TABLE: /g;
s/qergi/GRANULE ITERATOR: /g;
s/qergr/GROUP BY ROLLUP: /g;
s/qergs/GROUP BY SORT: /g;
s/qerhc/HASH CLUSTERS: /g;
s/qerhj/HASH JOIN: /g;
s/qeril/IN-LIST: /g;
s/qerim/INDEX MAINTENANCE: /g;
s/qerix/INDEX: /g;
s/qerjot/NESTED LOOP JOIN: /g;
s/qerjo/NESTED LOOP OUTER: /g;
s/qerle/LINEAR EXECUTION  IMPLEMENTATION: /g;
s/qerli/PARALLEL CREATE INDEX: /g;
s/qerlt/LOAD TABLE: /g;
s/qerns/GROUP BY NO SORT: /g;
s/qeroc/OBJECT COLLECTION ITERATOR: /g;
s/qeroi/EXTENSIBLE INDEXING QUERY COMPONENT: /g;
s/qerpa/PARTITION: /g;
s/qerpf/QUERY EXECUTION PREFETCH: /g;
s/qerpx/PARALLELIZER: /g;
s/qerrm/REMOTE: /g;
s/qerse/SET IMPLEMENTATION: /g;
s/qerso/SORT: /g;
s/qersq/SEQUENCE NUMBER: /g;
s/qerst/QUERY EXECUTION STATISTICS: /g;
s/qertb/TABLE ACCESS: /g;
s/qertq/TABLE QUEUE: /g;
s/qerua/UNION-ALL: /g;
s/qerup/UPDATE: /g;
s/qerus/UPSERT: /g;
s/qervw/VIEW: /g;
s/qerwn/WINDOW: /g;
s/qerxt/EXTERNAL TABLE FETCH : /g;
s/opifch2/SELECT FETCH: /g
s/qergh/HASH GROUP BY: /g
'

# main()

case `uname -s` in
        Linux)
                FUNCPOS=4
                ;;
        SunOS)
                FUNCPOS=2
                ;;
        *)
                FUNCPOS=2
                ;;
esac

if [ "$1X" == "-aX" ] ; then
        FILTER="^\$|oracle|----------"
        INDENT=" "
        shift
else
        FILTER="^\$|\?\?|oracle|----------"
        TRANSLATION_STRING="/opifch /,\$d;/opiefn0/,\$d;/opiodr/,\$d;/opiexe/,\$d; $TRANSLATION_STRING"
        INDENT=" "
fi

if [ "$1X" == "-kX" ] ; then
        FUNCPOS=2 # this is for linux kernel stack samples from /proc/PID/stack
        TRANSLATION_STRING="s/+.*//g; $TRANSLATION_STRING"
        shift
fi

if [ $# -eq 0 ] ; then

                      sed -e "$TRANSLATION_STRING" \
                    | egrep -v "$FILTER" \
                    | sed 's/^ *//g;s/__PGOSF[0-9]*_//' \
                    | awk -F" " "{ for (f=$FUNCPOS; f <= NF; f++) { printf \"%s \", \$f }; printf \"\n\" }" \
                    | f_lifo \
                    | awk "
                           BEGIN{ option=\"  \" } /rwsfcd/{ option=\"* \" } !/rwsfcd/{ pref=(pref \"$INDENT\") ;
                           print pref option \$0 ; option=\"  \" }
                      "
else

        for i in $* ; do
                sed -e "$TRANSLATION_STRING" < $i \
                    | egrep -v "$FILTER" \
                    | sed 's/^ *//g;s/__PGOSF[0-9]*_//' \
                    | awk -F" " "{ for (f=$FUNCPOS; f <= NF; f++) { printf \"%s \", \$f }; printf \"\n\" }" \
                    | f_lifo \
                    | awk "
                           BEGIN{ option=\"  \" } /rwsfcd/{ option=\"* \" } !/rwsfcd/{ pref=(pref \"$INDENT\") ;
                           print pref option \$0 ; option=\"  \" }
                      "
        done

fi

# that's all!