handy_gnu.txt

# ------ GNU/Bash ------
# In-file replace a with b in file in.txt and save a backup copy of original as in.txt.bak. If '' or nothing is specified for i, then no backup is saved
  sed -i'.bak' 's/a/b/' in.txt
# To cat two files and immediately redirect the output back so that it becomes the content of the first file, one cannot use 'cat a b > a', instead use:
  echo '1' > a; echo '2' > b; echo "$(cat a b)" > a; cat a  # i.e. echo "$(cat a b)" > a
  echo '1' > a; echo '2' >> a; echo '3' > b; echo "$(echo 'inserted 1'; cat a | grep -v '1'; cat b)" > a; cat a  # More complex example with extra insert/remove
# Start output of a file after a certain string is seen (i.e. delete all output upto a certain string)    Second example uses 'delete (d)' line numbers instead
  sed '0,/3/d' test  # If test had 6 lines with numbers 1->6, then output would be 4,5,6 (3 lines)        sed '1,3d' test  # Note functionality is quite different
# Print a single line from a file
  sed -n "1p" file  # Where 1 is the line number to print
# Get all directories that consist of numbers (any lenght) and the for each of those do something (dir name will be in $i inside loop)
  VAR=`ls -d [0-9]*`; for i in ${VAR[*]}; do ...insertyourcode...; done
# Get the current workdirectory of this script
  SCRIPT_PWD=$(cd `dirname $0` && pwd)
# Random number generator (6 digits). If randomness is paramount/you need a large number of random numbers remember to init entropy pool
  RANDOMD=$(echo $RANDOM$RANDOM$RANDOM | sed 's/..\(......\).*/\1/')
# Random selection of echo's, with another type of entropy pool init
  RANDOM=`date +%s%N | cut -b14-19`; case $[$RANDOM % 3 + 1] in 1) echo '1';; 2) echo '2';; 3) echo '3';; *) echo '+';; esac
# Random number between 0-9 based on nanosecond clock. To get random number between 0 and 99 do -b18-19 and so on
  date +%s%N | cut -b19-19
# Random number between 1 and X
  echo $(( $RANDOM % 2000 + 1 ))  or  echo $[$RANDOM % 2000 + 1]  # Examples where X is 2000, change as needed
# Change all upper case letters to lower case (With thanks, http://www.linuxquestions.org/questions/linux-newbie-8/evaluate-strings-as-case-insensitive-676101/)
  echo "LoweR CasE" | tr [:upper:] [:lower:]  # Output: lower case
# Ensure fault-free grep operation when grepping through a file which contains non-ASCII characters
  grep --binary-files=text ...
# Shuffle lines of a file. Much better then sort -R --random-source=/dev/urandom. sort -R will still sort according to hash, leaving similar lines clustered
  shuf --random-source=/dev/urandom inputfile > outputfile
# The same, but with a set number of lines extracted
  shuf -n10 --random-source=/dev/urandom inputfile > outputfile
# The same as original, with random entropy pool initialization (TODO: confirm that RANDOM=x entropy pool init also inits /dev/urandom)
  RANDOM=`date +%s%N | cut -b14-19`; shuf --random-source=/dev/urandom inputfile > outputfile
# Delete all alpha chars from a string
  echo "string" | tr -d '[:alpha:]'
# Re-use filename based output as input for a new command
  ls | xargs -I{} echo {}  # {} is the filename passed and has become a variable, echo is the example new command
# Execute two (or more) commands from xargs, based on pipe input, and additionally use some variables from the parent shell in the subshell
  export TEST="path of parent shell: "; ls * | xargs -I{} -i sh -c 'echo {}; echo {}; echo $0 $1;' ${TEST} ${PWD}
# Discover the process name of the parent shell (output can be things like 'screen', 'bash', 'sh', 'sshd' etc.)
  if [ "`cat /proc/$PPID/comm`" == "sh" ]; then echo "Subshell started from sh"; else echo "Subshell was started from `cat /proc/$PPID/cmdline`"; fi
# Gotcha: always use quotes around variables in an if statement, which would avoid errors if the string is empty/zero-lenght
  if [ "${SOME_STRING}" == "${SOME_OTHER_STRING}" ]; then ...
# Arrays: * expands to a concatenated, IFS 1st char seperated, one line string, but only if it is double quoted. [@] expands to each-member-of-the-array, always
  IFS="";v=('1 2' '3 4');for i in ${v[*]};do echo $i;done;for i in ${v[@]};do echo $i;done;for i in "${v[*]}";do echo $i;done;for i in "${v[@]}";do echo $i;done
# Show the manual (man) for a particular command without using pager mode (i.e. continuous stdout output): use `cat` command as pager:
  man -Pcat ls
# Get full output of a command into a file, while at the same time providing parsed/partial output of the same command to stdout
  for i in `seq 1 10`; do echo $i; done | tee fulllog.txt | grep "[369]"  # fulllog.txt will have 1-10, screen will only have 3,6,9
# Track progress, on a single line, of a live income stream, while modifying the income stream to be in a different (i.e. usually "readable") format. Example;
  for i in `seq 1 10`; do echo $i; sleep 0.3; done | awk '/[2-8]/{sub(/[456]/,"middle",$0);printf "%s",$0}'  # Output: 23middlemiddlemiddle78
# Tracking visual output of Docker build process, on one line, live:  $ sudo docker build . | \
  awk '{if(match($0,/Step/)!=0){sub(/Step[ \t]*/,"",$0);sub(/[ \t]*:.*/,"",$0);printf("%s... ",$0)};if(match($0,/Successfully built/)!=0){printf("%s",$0);}}'
# Using input redirection ("here document") which removes the tabs (ref http://tldp.org/LDP/abs/html/here-docs.html) (the EOF can have a leading tab also);
  cat <<-EOF   ....   EOF  # Where .... and EOF are on subsequent lines. The "-" in "<<-" removes a leading tab from what follows in ....
# The name of the calling procedure in a script
  caller 0 | awk '{print $2}'
# Extract the directory name, with leading slash, from a path with filename in the form of (example); "/some/paths/here/some_filename" to "/some/paths/here/"
  DIR_NAME=$(echo $PATH_AND_FILENAME | sed "s|/[^/]\+$|/|")
# Show signal masks for a running process - bit position 1 (2nd bit) is SIGHUP, etc..
  grep ^Sig /proc/$PPID/status
# Interleave, after each 75 lines in[to] file A.txt, the content of file B.txt
  sed -i "0~75 r B.txt" A.txt
# Interleave, after each 2nd line, a newly inserted string ($=EOL).          Second example does the same, just for a input file 'test' instead of for loop
  for i in $(seq 1 10); do echo $i; done | sed "0~2 s|$|\ninsert_text|";     cat test | sed "0~2 s|$|\ninsert_text|"
# If a variable [partially] matches a pre-determinted value, print another value instead (match needs to be from start of line)
  VAR1="abc"; echo "${VAR1/#ab/c}"  # output: cc ("ab" matches and is replaced by "c". A search for "bc" would fail, as it is not from start of line)
# Terminate a long-running/(or 'screen scrolling') command (which does not respond to CTRL+C quickly:
  1) Press CTRL+z, 2) jobs -l; kill -9 {PID_seen_in_jobs_-l_output}
# Timeout a given script/program after a set time. In the example, 1h=1 hour, but one can also use 10s=10 seconds etc.
  timeout --signal=9 1h ./pquery-run.sh
# Check for the existence of multiple files. (Btw, if you get '-bash: !": event not found' when running directly at the bash prompt, see 'Common Issues' below)
  if [ ! -r aaa -o ! -r bbb -o ! -r ccc ]; then echo "One or more files not found: aaa, bbb and ccc!"; fi
# Multi-threaded (3 threads in example) xargs call with direct input using cat. Output will be 1 then 2 then 3 after respective pauses (start at same time):
  CONCURRENCY=3; cat << EOC |
  sleep 1; echo "1" % sleep 2; echo "2" % sleep 3; echo "3"
  EOC
  xargs -d"%" -P${CONCURRENCY} -i^ sudo sh -c '^'
# xargs can be used as follows: INPUT | xargs | PROCESSOR - i.e. "pipe the INPUT directly into the PROCESSOR". Note this is conceptually somewhat similar to
  INPUT | xargs bash -c, but a much cleaner/clearer way to write it, as well as allowing some additional functionality + bash may not always be the shell used;
  echo "INPUT" | xargs | sed "s/INPUT/PROCESSOR/"
  echo ls | xargs | sh -x  # Conceptually similar to echo ls | xargs bash -c
  echo -n "echo\necho" | xargs | sed 's|DUMMY|NO ACTION|'  # When used like this, a \n is added to the output. This is not the case if 'xargs |' is dropped
# For all lines in a file, action something. Note that the for loop has limitations (spaces) which require IFS to be set. This is not so with the while loop
  ORIG_IFS=${IFS}; IFS=$'\n'; for LINE in $(cat ./file.txt); do echo ${LINE}; done; IFS=${ORIG_IFS}   #OR#  while read line; do echo ${line}; done < ./file.txt  # However, the best solution is:
    while IFS=$'\n' read LINE; do
    done < "${INPUTFILE}"
# Obtain a fully qualified directory and script name
  SCRIPT_AND_PATH=$(readlink -f $0); SCRIPT=$(echo ${SCRIPT_AND_PATH} | sed 's|.*/||')
# Change the file contents (inline replace used here) of multiple files in subdirectories
  ls ./*/somefiles | xargs -I{} sed -i 's|find|replace|' {}
# Rename file names from abc to xyz (where there is no numbers in the filename) in numerical subdirectories (where there is no letters in the directory name)
  ls ./*/abc | sed 's|[^0-9]||g' | xargs -I{} mv ./{}/abc ./{}/xyz
# Age of a file or directory in seconds (Thanks http://www.linuxquestions.org/questions/linux-software-2/how-to-calculate-file-age-file-xy-is-2-days-3-hours-old-164908/)
  echo $[ $(date +%s) - $(stat -c %Z filename.txt) ]  # where filename.txt is the file name (can also be a directory name) you want to know the age of
# For each file in a given filespec, action something. Example here: all reducer logs
  for file in /dev/shm/14*/reducer.log; do echo "$file"; done
# Two different ways to find today's directories, text process them, and then execute the processed text
  find . -mindepth 1 -maxdepth 1 -type d -mtime -1 -exec sh -c 'cd {};~/percona-qa/pquery-results.sh' \;
  ls -l | grep "^d" | grep "$(date +'%b %d')" | sed 's|.*:[0-9][0-9] ||' | xargs -I{} echo 'cd {};~/percona-qa/pquery-results.sh' | xargs -I{} sh -c '{}'
# Display a one-liner script, and append to it a given option, then execute the total. Example used here is a MyQSL client script which already has mysql -uroot etc.
  cat ./cl | sed -e "s|$| -e 'SELECT 1;SELECT 2;'|" | xargs -0 -I{} sh -c {}
# Insert a line at the top of a file
  sed -ie '1i\your_text_here' file.txt
# Rename files, in subdirectories, from .log to .sql
  ls */pquery_thread*.log | xargs -I{} echo "mv {} {}DUMMY" | sed 's|logDUMMY|sql|' | xargs -0 | sh
# Conditionally execute commands based on presence of a string within a file
  if cat testfile.txt | egrep -qi "123"; then echo "yes"; else echo "no"; fi;   #OR#
# Uppercase/lowercase a string. With thanks, http://stackoverflow.com/a/19411918/1580826
  V="Aa"; echo ${V^^}; echo ${V,,}  # Will output AA and aa respectively
# Check a string for presence of a search/substring. With thanks, http://timmurphy.org/2013/05/13/string-contains-substring-in-bash/ and https://stackoverflow.com/questions/26320553/case-insensitive-comparision-in-if-condition
  if [[ "$string" == *"$searchstring"* ]]; then echo "found"; else echo "not found"; fi  # This is much faster then grep -q (which is not recommended for strings) and always works
  if [[ "$string" =~ $regex_searchstring ]]; then echo "found"; else echo "not found"; fi
  if [[ "$string" == *"={$searchstring1|$searchstring2}"* ]]; then echo "yes"; else echo "no"; fi;
  if [[ "${string^^}" == *"${searchstring^^}"* ]]; then echo "yes"; else echo "no"; fi;  # Case insensitive compare
  TEXT='test '; echo "${TEXT^^}"  # TEST
  TEXT='test '; echo "${TEXT,,}"  # test
  TEXT='test '; echo "${TEXT^t}"  # Test
  TEXT='test '; echo "${TEXT^^t}" # TesT
  TEXT='test '; echo "${TEXT^^e}" # tEst
# Check if sudo is installed and available
  if [ "$(sudo -A echo 'test' 2>/dev/null)" == "test" ]; then echo "SUDO installed & available"; else echo "SUDO not available, or password based access only"; fi
# Sent all stdout/stderr output to a file while still on screen (With thanks http://stackoverflow.com/questions/692000/how-do-i-write-stderr-to-a-file-while-using-tee-with-a-pipe)
  exec 1>> >(tee /tmp/stdout.log >&1); exec 2>> >(tee /tmp/stderr.log >&2)
# Start console selflogging from inside a script (With thanks http://stackoverflow.com/questions/5985060/bash-script-using-script-command-from-a-bash-script-for-logging-a-session)
  [ -z "$TYPESCRIPTLOG" ] && TYPESCRIPTLOG=1 exec /usr/bin/script -a -q -f /tmp/console_typescript.log -c "TYPESCRIPTLOG=1 $0 $@"  # http://stackoverflow.com/a/26308092
# Hack for per-subshell variables to ensure 'thread'(subshell) safety (With thanks http://askubuntu.com/questions/305858/how-to-know-process-pid-of-bash-function-running-as-child)
  function_runs_in_subshell(){ PID_OF_SUBSHELL=$BASHPID; VARIABLE[${PID_OF_SUBSHELL}]="some_var_value"; echo ${VARIABLE[${PID_OF_SUBSHELL}]}; }  function_runs_in_subshell &
# Read a file into an array (With thanks http://stackoverflow.com/questions/30988586/creating-an-array-from-a-text-file-in-bash)
  mapfile -t somearray < file.txt  # -t drops newline chars. See http://wiki.bash-hackers.org/commands/builtin/mapfile for more options
# Make a file unique/sort it without using a temporary file (similar to the -i function in sed). Note that something like: cat file.txt | sort -u > file.txt does NOT work.
  sort -u -o file.txt file.txt
# Using sed, replace any '-' into '_' in 'a-a-a=b-b_b' but only before the '=', which is definitely not straighforward. (With thanks http://stackoverflow.com/a/6637457)
  echo "a-a-a=b-b_b" | sed 'h;s|-|_|g;s|=.*|=|;x;s|.*=||;x;G;s|\n||';  # Output: a_a_a=b-b_b # h: place copy in hold buffer, x: swap buffers; G: dump buffers
# The same as the line above, but this one a bit more robust: ensure that if there is a secondary '=', it will still only change the first part
  echo 'a-a-a=b-b_b=c-c_c' | sed 'h;s|-|_|g;s|\([^=]\+\).*|\1|;x;s|[^=]\+||;x;G;s|\n||'  # Output: a_a_a=b-b_b=c-c_c # \([^=]\+\) : select the part, any char not =, upto first =
# Start a immediately-detached screen session, execute ls (in color) and provide a new prompt there. Name it 's1' so that  $ screen -d -r s1  will get you into the screen
  screen -admS s1 bash -c "ls --color=auto;bash"  # With thanks, http://www.cyberciti.biz/faq/how-to-turn-on-or-off-colors-in-bash/ and http://askubuntu.com/q/62562
# An easy way to get all text from a webpage, ready for further textual processing (removes all html tags etc)
  lynx --dump http://someurl.com
  lynx --listonly --dump https://www.percona.com/downloads/TESTING/pmm/ | grep  "pmm-client.*.tar.gz" |awk '{print $2}' | head -n1
# Sort a numeric array  # With thanks, https://www.reddit.com/r/bash/comments/3cbt1p/sorting_array_of_numeric_values/csu2150/
  mapfile -t ARRAY < <(printf '%s\n' "${ARRAY[@]}" | sort -n)  # You can also use -u with the sort to have unqiue values only and/or -r to have a reverse (large>small) sort
# Find a certain string within all files in a subdirectory (handy for finding text like function names etc. in source code etc.)
  grep -R --include=* -n "search string"  # From the highest directory
# Generate a random string  # With thanks, https://gist.github.com/earthgecko/3089509
  cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w ${1:-32} | head -n 1
# Redirect multiple files into mapfile or other commands  # With thanks, https://unix.stackexchange.com/questions/238967/controlling-the-order-in-which-files-are-put-into-an-array
  mapfile -t arrayname < <(cat a b)  # where a and b are input file[s][names]
# Obtain the parent pid (in-script)  # With thanks, https://superuser.com/questions/150117/how-to-get-parent-pid-of-a-given-process-in-gnu-linux-from-command-line
  ppid () { ps -p ${1:-$$} -o ppid=; }
  P=$(ppid); echo $P
# Combine multiple lines of sql untill a ';' is seen near EOL (; followed by spaces or tabs and then EOL will work as well). Ideal for combining multi-line SQL statements into one
  # With thanks, https://www.unix.com/shell-programming-and-scripting/80633-sed-combining-multiple-lines-into-one.html & https://www.gnu.org/software/sed/manual/sed.html#Multiline-techniques
  cat multiple_unterminated_lines.sql | sed -e :x -e '/;[ \t]*$/b' -e N -e 's/\n//' -e bx
# Sed can be used to "add/insert" a line behind another one. In the example (output is (on 3 lines): 1 2 3), 2 is inserted after the search term of "1". The 'a' is command for 'add'
  echo -e "1\n3" | sed '/1/a\2'
# Generate a random string of any lenght, with thanks @MichaelC
  sudo apt-get install pwgen; pwgen -ycns 100 1  # 100 chars, one result
# Find out what debugger and potentially what OS was used to compile mysqld, show GCC version, with thanks @DavidB
  objdump -s --section .comment bin/mysqld  # Then use a site like https://packages.debian.org/search?keywords=gcc-{version} etc.
# Find out the exit code of various elements used in a Linux pipeline  # With thanks, https://unix.stackexchange.com/a/14276/241016
  true | false; echo "${PIPESTATUS[0]} ${PIPESTATUS[1]}"  # Result: 0 1: 0 (correct exit code of true) 1 (correct exit code of false)
# Insert a line at the top of a file  # With thanks, https://linuxconfig.org/how-to-insert-line-to-the-beginning-of-file-on-linux
  sed -i "1 s/^/This is my first line with a variable: ${PWD}\n/" filename.txt
# If you have made a process background (bg) (and not 'disown'ed it), or interrupted it with CTRL+z where, as an example, it will show you the background ID: [1], you can kill this particular background process, by using:
  kill %1  # Where 1 is the index of the background or interrupted process
# Halt a process, move it to background, disown it (so that if you leave your shell session it will continue to execute in the background)
# List subdirectories only, with thanks https://linoxide.com/linux-command/linux-commad-to-list-directories-directory-names-only/
  ls --color=never -dF ${DIR}/* 2>/dev/null | grep \/$ | sed 's|/$||'  # Where DIR is the root/input
  Press CTRL+z on the running process, then type:  bg; disown
# See which processes are consuming the most memory, and how much memory they are consuming
  watch -n1 "ps --sort -rss -eo pid,pmem,rss,vsz,comm | head -16"
# A bit safer form of 'rm -Rf somedir' if you have no subdirectories:
  rm -df somedir/* somedir  # This will remove the directory, and all files in it

# ------ Debugging ------
# View complete output of all executed items, which is great for debugging
  bash -x ./your_script.sh
# In some complex redirection cases, using bash -x is not sufficient for finding the source a problem, for example a faulty sed command. In such cases use
  script out.txt
  bash -x ./your_script.sh; vi out.txt  # Then search for the error seen previously on normal execution. Verify surrounding output
# View all OS-level function/system calls/IO etc.
  strace ./your_script.sh  # You may need to add additional strace options, see 'man strace'
# View all file I/O operations for a running process, cleanly
  sudo strace -p $(pidof -s mysqld) -e trace=%file -f  # Change $(pidof -s mysqld) to any pid (pidof -s just returns one pid of one mysqld)
# View what mysqld (or any other process) is doing. Ideal for hangs/odd long query durations etc. (First install utility perf: $ sudo yum install perf)
  perf record -g --pid=`pidof -s mysqld`; perf report | c++filt; rm perf.data   # CTRL+C to interrupt the perf record once you think you have enough data
# Find out which compiler was used to compile an un-stripped binary
  objdump -s --section .comment /path/binary
# Attach gdb to a certain PID then write a coredump. In the example below, $! is used which is the PID for the last started process (sleep in this example)
  sleep 1000 &  gdb /usr/bin/sleep --pid=$! --ex 'generate-core-file'
# Check your bash scripts
  https://www.shellcheck.net/  https://github.com/koalaman/shellcheck
# Check if this script is being run interactively (i.e. $. ./script.sh instead of $./script.sh)
  if [[ "${-}" == *"i"* ]]; then echo "script was executed as . ./script.sh"; fi

# ------ Compiling / Compiler Issues ------
# If you see for example "/usr/bin/ld: cannot find -lzlib" in your compilation/build of MySQL, and the build fails because of it, then check:
  ld -lzlib --verbose  # Execute in a user writeable directory. If all entries fail, the lib could not be found.
  You need to just link one of them to the real zlib on your system  ## With thanks, https://stackoverflow.com/a/21647591/1208218
  sudo ln -s /usr/lib64/libz.so.1.2.7 /usr/lib/libzlib.so  ## Though you may have to search for "libz.so" in /usr (cd /usr; find . | grep libz.so)
  This likely works with any linker issue (ld -l{failed lib} --verbose && sudo ln -s ...)
  Another example;
  ld -ltensorflow --verbose  # Note the "{somelib}.so needed by {somelib}.so" and next line "found {somelib}.so {path}"
  ld /usr/local/lib/libtensorflow.so  # This part; "ld: warning: cannot find entry symbol _start; not setting start address" can be ignored (it is likely just looking for a start address, which will be there in the program which links with this library)
# If you get an error like this;
  /usr/local/lib/somlelib.so: undefined reference to `somefunction@GLIBC_2.23'
  # Then check first what GLIBC version you are running;
  sudo yum list glibc  # Or use ldd --version
  # If the version is older (in this example it was 2.17), then either update the OS or consider adding another glibc along the main OS one;
  https://unix.stackexchange.com/a/299665/241016
# An easy way to install the almost-latest gcc + binutils on Centos 7
  sudo yum install epel-release
  sudo yum install centos-release-scl-rh
  sudo yum install devtoolset-7-gcc-c++  # Or sudo yum install devtoolset-6-gcc-c++  # Or sudo yum install devtoolset-4-gcc-c++
  # Then, to enable the newly installed (and separate) environment, use;
  scl enable devtoolset-7 bash  # This sets the newly installed environment for that particular bash session
  # https://access.redhat.com/documentation/en-US/Red_Hat_Developer_Toolset/3/html/User_Guide/sect-Red_Hat_Developer_Toolset-Install.html
  # https://access.redhat.com/documentation/en-us/red_hat_developer_toolset/6/html/user_guide/sect-red_hat_developer_toolset-install
  # Note that devtoolset-4 is really gcc 5.3.1 (the native Centos7 gcc being 4.8.5)
# Capture all console output, including stderr:  sudo apt-get/yum install script
  script; ./somescript.sh; exit; vi typescript
# Easy way to see all live network connections on a system with DNS resolved hostnames
  sudo apt-get install iftop; sudo iftop  # And use for example:  iftop -i wlan0  to see wireless stats etc.

# ------ Common Issues ------
# Ever seen "-bash: !": event not found" when you were trying to use '!' in for example a sed command? Try for examle: echo "!" | sed 's|!|found|' - the fix;
  histchars=  # in ./bash_profile (or at the command line) will fix it. Reason; https://www.gnu.org/software/bash/manual/html_node/History-Interaction.html
  A better fix: use: set +H  # With thanks https://serverfault.com/a/208414
  See also https://www.gnu.org/software/bash/manual/html_node/The-Set-Builtin.html (note: +/-H != +/-h)
# Unlike echo -n, sed cannot parse a live input stream and print the result on a single line step-by-step as the input comes in. Proof;
  for i in `seq 1 10`; do echo $i; sleep 1; done | sed -n "H;/[369]/{x;s/\n/ /g;p}"  # Ref https://www.gnu.org/software/sed/manual/sed.html#fn-3
# While the above cannot be done with sed, it can be done quite easily with awk. Semi-similar example, showing output, progressively, on one line;
  for i in `seq 1 10`; do echo $i; sleep 0.3; done | awk '/[369]/{printf "%s",$0}'  # To make example above 100% similar, use: sed -n "/[369]/{s/\n/ /g;p}"
# When calling a function in the background (like the example on the next line) in a script, remember that such a function runs in a subshell
  function &  # By calling a function this way, all variables used within it will be fresh/new and any variables updated will not reflect back in the parent
# When checking for the existence of a directory, do not just use:  if [ -d ${SOME_VAR} ]; then echo "yes"; fi  The reason is that if ${SOME_VAR} is empty,
  the check will still work. Confirm the gotcha by executing:  if [ -d ]; then echo "yes"; fi  Instead, use:  if [ "${SOME_VAR}" != "" -a -d ${SOME_VAR} ];
# Along the same lines of the last issue, it is an easy mistake to make not to include double quotes around variables when checking for (in)equality. Use;
  if [ "${SOME_VAR}" == "Some Text" ]; then ...; fi  instead of not using double quotes (which fails when the variable is empty).
# If you see an error like '/usr/bin/ld: cannot find -lsnappy' when compiling, then it means that the snappy (for example) library cannot be found.
  To discover the real source, run  ld -lsnappy --verbose  and look near the end for the locations is scans for 'attempt to open ...'
  If you have the library in a specific location, you can use;  sudo ln -s /usr/lib64/libsnappy.so.1 /usr/lib64/libsnappy.so  or use -L option to your compiler!
  With thanks, http://stackoverflow.com/questions/16710047/usr-bin-ld-cannot-find-lnameofthelibrary
# Find out symbols for a library
  nm --synthetic some_plugin.so | c++filt
# Recursively copying with hidden files included: http://superuser.com/a/804367/457699
  cp -a  # Always use cp -a instead of cp -R (or -r) which does not copy hidden files/directories
# The first echo below will output an asterix, the second will output a directory listing
  echo "$(echo "*")"; echo "----"; echo $(echo "*");
# Except for searching files (if grep -qi "search string" file.txt...), it's best not to use: echo "text" | grep -q "text" to search strings for substrings. It is very slow, and
  has a gotcha: Test this:  LINE='test'; DONE=0; echo ${LINE} | if grep -q 'test'; then DONE=1; echo "DONE"; fi; echo $DONE  # DONE is 0 because (as per Bash team); 'All elements
  of a pipeline are executed in subshells.  A subshell cannot affect its parent's environment. The 'lastpipe' shell option can change this when job control is not active.'
  Instead, see 'Check a string for presence of a search/substring' above for substring searches. Another gotcha can be seen in this one; echo "abc123xyz457" | grep -o '[^0-9]'
# In Bash, a calling function needs to be declared ABOVE the line it is called from. An example script (place the following 3 lines in a script) will show this (output on right):
  echo "does not work (error will be shown before this line): `works`"         # ./test.sh: line 1: works: command not found
  works(){ echo "works"; }                                                     # does not work (error will be shown before this line):
  echo "does work (will show 'works' output after semicolon): `works`"         # does work (will show 'works' output after semicolon): works
# However, the following does work, as all the functions are "read" before one of them is called (even though works1 calls a function defined LOWER then the calling line itself):
  works1(){ echo "`works2`"; }
  works2(){ echo "works"; }
  echo "does work: `works1`"                                                   # does work: works
# When wanting to execute multiple commands as one, the syntax is () not $(). For example, to time two commands executed sequentially;
  time $(echo "a";echo "a")  # WRONG: this will try and execute "a" and report: bash: a: command not found... (the total time will still be approx. correct, though bit higher)
  time OUTPUT=$(echo "a" ; echo "a"); echo $OUTPUT  # WRONG: while it works, just like the above, the time is slightly higher (due to the added variable assignment)
  time echo "a";echo "a"     # WRONG: this does not achieve the objective, it only times the first command
  time (echo "a";echo "a")   # CORRECT: this times both commands (without anything affecting time and without any sub-execution or variable assignment)
# Obtain the primary IP address of this machine
  IP_ADDRESS=$(ip route get 8.8.8.8 | head -n1 | cut -d' ' -f8); echo ${IP_ADDRESS}  # Fails on for example Centos 7.4 (needs -f7)
# Better methods to obtain the IP address of this machine
  IP_ADDRESS=$(ip route get 1 | head -n1 awk '{print $NF;exit}'); echo ${IP_ADDRESS}  # Or perhaps even better;
  IP_ADDRESS=$(hostname -I | head -n1 | cut -d' ' -f1); echo ${IP_ADDRESS}  # Current version we use
# Generate fake data
  https://faker.readthedocs.io/en/latest/providers.html  # Faker for Python
# When used in a function, the 'declare' command (for example DECLARE -a ARRAY) makes each variable local, as with the 'local' command, unless the '-g' option is used.
  With thanks, https://unix.stackexchange.com/questions/136720/bash-array-declared-in-a-function-is-not-available-outside-the-function
  It is thus likely best to use 'local' instead of 'declare' to remove any misunderstanding. This is also the default code syntax for percona-qa scripts
# Gotcha: do not use:  some command | uniq  assuming you will get unique items only. uniq and uniq -u can produce very unexpected results. To see some examples, try these;
  echo -e "1\n2\n2\n3\n2">test;echo "test:";cat test;echo "sort -u test:";sort -u test;echo "uniq test:";uniq test;echo "uniq -u test:";uniq -u test
  echo -e "1\n2\n2\n3">test;echo "test:";cat test;echo "uniq -u test:";uniq -u test
  With uniq -u it keeps any line which is truly unique, and this is why the first uniq, the first uniq -u and the second uniq -u results are all so different
  Thus; uniq is a versatile tool, but only if you are after very specific functionality. If not, always use sort -u
# Sort unique vs uniq: sort -u sorts the input list and only leaves unique entries. uniq on the other hand does not sort and removes only adjoining same entries
  cat need_sorted_file_with_unique_results_only.txt | sort -u; cat need_adjoining_same_entries_removed; | uniq; # cat need_all_duplicates_removed.txt | uniq -u
# Gotcha: This will not work: $ if [[ "a-b" == *"a[-_]b"* ]]; then echo "yes"; fi
          This will     work: $ if [[ "a-b" == *"a"[-_]"b"* ]]; then echo "yes"; fi  # With thanks, http://wiki.bash-hackers.org/syntax/pattern
# Gotcha: sh -x != bash -x. If you use sh -x instead of bash -x to see the debug output of a complex script, you may see errors like this:
  ./script.sh: 5: ./script.sh: [[: not found  # The reason is that sh -x cannot handle "[[" if statements whereas bash -x can
# Consider difference between:
  for i in $(seq 1 10); do echo $RANDOM | sed 's|\(^.\)|\1.|'; done | xargs -I{} echo "sleep {}; echo {}" | xargs -I{} bash -c "{}"
  # And the next line. The first one will run sequentially, whereas the -P10 version will start 10 threads at once
  for i in $(seq 1 10); do echo $RANDOM | sed 's|\(^.\)|\1.|'; done | xargs -I{} echo "sleep {}; echo {}" | xargs -P10 -I{} bash -c "{}"
# See what processes are accessing a given file
  fuser -a /path_to/your_file

# ------ GNU 'screen' usage ------
# Want to scroll inside a screen session?
  Use CTRL+a+ESC (i.e press CTRL+a, keep CTRL in, then press ESC), then use cursor up (or down) to scroll, ESC or q to escape
# Want to find which screen session is doing what?
  pstree -g  # Great for finding the screen PID (i.e. for use with screen -d -r {pid}) that is runnnig a particular script/binary

# ------ OS/Admin ------
# Running out of space on /dev/shm yet the actually used space is much smaller then allotted space? If so, is likely due to [dead] processes holding on to old files. Run this regularly:
  sudo lsof -s +L1 | awk '{print $7" "$2" "$3" "$10}' | grep 'dev/shm' | grep -E "^[0-9][0-9][0-9][0-9][0-9][0-9][0-9][0-9][0-9][0-9]|^[3-9][0-9][0-9][0-9][0-9][0-9][0-9][0-9][0-9]" | awk '{print $2}' | xargs kill -9 2>/dev/null  # Kill any processes holding onto 'deleted'/old files > 300GB. This should work for other volumes also, if the /dev/shm grep is modified or removed
# Monitor the current rate of syscalls/sec  ## With thanks, http://www.brendangregg.com/blog/2018-02-09/kpti-kaiser-meltdown-performance.html
  sudo perf stat -e raw_syscalls:sys_enter -a -I 1000
# If umount produces this error: "umount: /<mountpt>: target is busy.", consider using this (potentially risky!) lazy unmount instead: (<device>=actual disk)
  ## sudo umount -flv /dev/<device>                         ## Make sure you know what you are doing before executing this!   Another possible method is:
  ## sudo yum install psmisc; sudo fuser -km /dev/<device>  ## Make sure you know what you are doing before executing this!
# To recover text from a deleted file, use: (change 'your_search_string' to a known string that was in the file, and /dev/sda to the partition it was on)
  sudo grep -a -C 500 'your_search_string' /dev/sda | tee /tmp/recover  # Make sure to tee to another partition then the one the file was on
# Make a 12Gb tmpfs drive for reducer.sh/pquery-run.sh (also works on Centos7): edit fstab (sudo vi /etc/fstab;) and add:
  tmpfs     /dev/shm        tmpfs   size=12g,noatime,nodiratime       0 0
# Check the process limits of any process, mysqld example:
  cat /proc/`pidof mysqld`/limits
# Hanging processes/load average (ref 'uptime') increasing steadily/rogue processes? See which ones are causing issues by:
  top -n10 -b | head -n25
# Trace all TCP on localhost. Use in combination with mysql -uroot -h127.0.0.1 -P3306 to see TCP traffic generated by server/client communication
  sudo tcpdump -i lo -X -s 0 src host 127.0.0.1
# After running cmake/make, you can use make install in the following way to install the package to a non-default location (handy for testing):
  make install DESTDIR=/tmp/somedir  # With thanks to AlexeyB
# List packages on Ubuntu/Debian
  dpkg-query -l | egrep -i "percona|mysql|maria" | awk '{print $2}'  # With thanks to Evgeniy Patlan
# List installed version of jemalloc on Centos
  sudo yum list jemalloc
# Undelete utility: testdisk. example with /dev/sda below. Once browsing, use ':' to select files, then use 'C' to copy and again 'C' to specifiy target
  sudo yum install testdisk; sudo umount /dev/sda; sudo testdisk /dev/sda  # Enter on /dev/sda > 'None' partition > Advanced > Select patition > cursor right ('List') > browse
# Reset all directories to 755 (needed to be able to browse them) and all files to 644. This cannot be done in a single chmod command.
  find . -type d -exec chmod 0755 {} \; find . -type f -exec chmod 0644 {} \;  # With thanks https://help.directadmin.com/item.php?id=589
# Cleanup a broken installation of some program (example; someprog - note that someprog needs to be a recognized program name by the package manager). Confirmed to fix things.
  sudo apt-get purge someprog; sudo apt-get autoclean; sudo apt-get clean; sudo apt-get update; sudo apt-get install someprog; [optional:] sudo apt-get install -f someprog
# Get the IP address of the host (hostname --ip-address may produce 127.0.0.1 which is not usefull if you need the network interface IP)
  ip route get 1 | awk '{print $NF;exit}'  # With thanks, https://stackoverflow.com/questions/13322485/how-to-i-get-the-primary-ip-address-of-the-local-machine-on-linux-and-os-x
# Raise the priority of a certain process  # With thanks, https://serverfault.com/a/228632
  sudo renice -10 -p ${PROCESSID}  # Where ${PROCESSID} is the ID of the process to raise priority for
# Quickly profile a system inside scripts
  df -kTh; top -n1; cat /proc/cpuinfo; uname -a; cat /etc/*release* 2>&1; dpkg -l; printenv
# Clear OS cache, potentially avoid OOM on high use servers by running this regularly
  sudo sh -c "$(which echo) 3 > /proc/sys/vm/drop_caches"
# Use cgroups to limit application memory usage
  sudo service cgconfig restart
  sudo cgcreate -g memory:DBLimitedGroup
  sudo sh -c "$(which echo) 32G > /cgroup/memory/DBLimitedGroup/memory.limit_in_bytes"
  sudo sync;sudo sh -c "$(which echo) 3 > /proc/sys/vm/drop_caches"
  sudo cgclassify -g memory:DBLimitedGroup `pidof mysqld`
# apt != apt-get  # For example, to list application you need: sudo apt list <package_name> and list is not a valid option for apt-get
  sudo apt-cache policy <package_name>  # Can be used to find out what repo a package comes from etc.
  apt-cache search <package name>       # Search
# Display space used by subdirs, handy to find out what is using space. Note that the sort is not perfect
  ls -l | grep "^d" | awk '{print $9}' | xargs -I{} du -sh ./{} | sort -n | more
# Configure timezone on a server (does not change any other works of the server, only the 'local time')
  timedatectl && timedatectl list-timezones | grep "Australia"  # or similar
  sudo timedatectl set-timezone Australia/Sydney                # or similar
  # If timedatectl is not available (older systems), use dpkg-reconfigure tzdata
# Configure timezone on a per-user basis 
  timedatectl list-timezones | grep yourplace  # To see a list to swap the last bit of:
  echo 'export TZ=\"/usr/share/zoneinfo/your_country/your_place\"' >> ~/.basrc

# ------ Tools ------
# Use shellcheck (See https://github.com/koalaman/shellcheck and http://www.shellcheck.net/) to check script syntax
  sudo yum install cabal-install; cabal update; cabal install shellcheck; ~/.cabal/bin/shellcheck
# Expand DEB files
  ar x yourfile.deb; tar -xf data.tar.xz  # Each deb file has control.tar.xz and data.tar.xz
# Compressing files while sending
  ssh hostname 'tar cf - -C /path/to/source directory'|tar xvf - -C /path/to/target

# ------ Shortcuts ------
# Pquery, single-threaded run, mysqld pre-started, socket, testcase and ./lib/ in this directory, percona-qa bzr tree pulled to ~/percona-qa/
  export LD_LIBRARY_PATH=${PWD}/lib;~/percona-qa/pquery/pquery --infile=./testc.sql --database=test --threads=1 --no-shuffle --user=root --socket=./socket.sock

# ------ Git ------
# git: git checkout (checkout a branch), git branch [-a] [-r] (show current branch, -a is all branches local and remote, -r is remote branches only)
# Undo last commit, leaving everything else as-is (for example to change commit message)  # With thanks, https://stackoverflow.com/questions/2845731/how-to-uncommit-my-last-commit-in-git
  git reset --soft HEAD^         # Undoes commit
  git reset HEAD^                # Undoes commit AND everything staged (i.e. new files added)
  clean_tree.sh (in percona-qa)  # Resets tree completely to last online commit
# Register author details
  git config --global user.name "your name"
  git config --global user.email email@domain.com
# Store git credentials, and turn off warnings for whitespace errors (with thanks, http://stackoverflow.com/a/12396793/1580826) - run from within clone directory
  git config credential.helper store  # After this, you only need to enter your password (or app token) once. It will then be saved for the future
  git config apply.whitespace nowarn
# Diff ALL changes, including newly added (i.e. staged) files (with thanks, http://stackoverflow.com/a/27299842)
  git diff HEAD
# Apply a diff and then patch (with thanks, http://www.thegeekstuff.com/2014/03/git-patch-create-and-apply/)
  git diff HEAD > patch.diff; # Change to another place #; git apply < patch.diff
# Retrieve a single file in full from Github using wget. With thanks, http://stackoverflow.com/a/4605068
  wget https://raw.githubusercontent.com/Percona-QA/percona-qa/master/ldd_files.sh  # The filename, repository etc. in the url can be changed to get the needed file
  wget https://raw.githubusercontent.com/Percona-QA/percona-qa/master/pmm-info.sh && chmod +x pmm-info.sh && ./pmm-info.sh
# Fixing a set of incorrect pushes:
  # cd .. ; cp -r ${YOURBRANCH} ${YOURBRANCH_COPY}  # Important: Make a copy of your branch!
  git reset --hard ${SHA}  # Reset the branch to a known good state/commit (warning; deletes everything newer than the state of the branch at the moment the ${SHA} commit was made
  # Now, make any changes necessary, like for example copying in files etc. from ${YOURBRANCH_COPY}
  git commit -a -m "My new commit"  # Commit any changes
  git push -f  # Forces an overwrite of history. Anyone else using your branch must rebase or repul
# Handy git fix reference; https://sethrobertson.github.io/GitFixUm/fixup.html

# ------ Must read's ------
# http://tldp.org/LDP/abs/html/globbingref.html on globbing (and not regex) and filename expansion
  echo *  # expands "*" to all filenames     ls -l [^ab]*  # Lists all files not starting with a or b, but [^ab]* is not a regex, it's globbing
# Various tools to extract information from files etc. From http://stackoverflow.com/questions/1124571/get-list-of-static-libraries-used-in-an-executable
  file, ldd, nm, objdump, readelf, strings
# On Redirection
  http://www.tldp.org/LDP/abs/html/io-redirection.html
# On Arrays
  http://wiki.bash-hackers.org/syntax/arrays
# On using a named pipe
  http://stackoverflow.com/a/14893732
# A search engine which does not ignore special characters, so you can search for code specific items
  http://symbolhound.com/

# ------ ~/.vimrc (copy and paste into file, sets mariadb-qa coding guidelines for vi) ------
" tabstop:          Width of tab character
" softtabstop:      Fine tunes the amount of white space to be added
" shiftwidth        Determines the amount of whitespace to add in normal mode
" expandtab:        When on uses space instead of tabs
set tabstop     =2
set softtabstop =2
set shiftwidth  =2
set expandtab
set nocompatible
colo torte
syntax on

# ------ SQL ------
# For testing, generate a random string variable to assign it to a variable
  SET @str=CAST(RAND() AS CHAR(10));
  SET default_master_connection=@str;  # For example

# ------ Google cloud ------
# Configure a serial port for console level access:
  https://cloud.google.com/compute/docs/instances/interacting-with-serial-console
  Basically, add a variable here (change {yourproject} and {yourorgid} to yours):
  https://console.cloud.google.com/compute/metadata?project={yourproject}&folder&organizationId={yourorgid}
  Variable to add:  serial-port-enable	value:  TRUE
  Then go back to (change {yourzone}, {yourserver}, {yourproject} and {yourorgid} to yours):
  https://console.cloud.google.com/compute/instancesDetail/zones/{yourzone}/instances/{yourserver}?project={yourproject}&folder&organizationId={yourorgid}
  And click the "Connect to serial console" button
# Remount root as r/w to enable changing files like /etc/fstab
  mount -o remount,rw /
# Check if entries in dmesg/journal give more information
  journalctl -xb
# Check if system has correct boot info
  update-initramfs -u  # If you get an error like 'cannot find boot device in /etc/fstab', debug fstab
# Issue with a specific driver on Ubuntu which can be blacklisted
  echo "blacklist i2c-piix4" >> /etc/modprobe.d/blacklist.conf
  # With thanks, https://askubuntu.com/questions/691729/piix4-smbus-0000007-3-host-smbus-controller-bus-not-enabled