#!/bin/bash # # Shows time in seconds to first byte of a url or urls # # Based on a gist https://gist.github.com/sandeepraju/1f5fbdbdd89551ba7925abe2645f92b5 # by https://github.com/sandeepraju # # Modified by jay@gooby.org, @jaygooby # # Usage: ttfb [options] url [url...] # -d debug # -l (infers -d) log response headers. Defaults to ./curl.log # -n of times to test time to first byte # -v verbose output. Show request breakdown during -n calls # # Examples: # # ttfb https://example.com/example/url # .098974 # # ttfb -n 5 https://example.com/ # ..... # fastest .099195 slowest .103138 median .099684 # # ttfb -n 5 bbc.co.uk news.bbc.co.uk # ..... # ..... # bbc.co.uk fastest .045873 slowest .046870 median .045999 # news.bbc.co.uk fastest .042286 slowest .060245 median .046035 # # ttfb bbc.co.uk news.bbc.co.uk # bbc.co.uk .048378 # news.bbc.co.uk .049303 # # Implicitly follows redirects using curl's -L option. # # Log all response headers (default log file is ./curl.log) by calling with -d # # Override the default log file by specifying -l /some/file # # Get min, max and median values by specifying the number of times to call # the URL; use -n2 for 2 tests, -n5 for 5 and so on. # # If you specify more than one url and have specified -d or -l, the log file # will be prefixed with the URL being requested. # # If you specify -n and -d or -l, the response headers from the consecutive # requests will be concatenated in the log file. # # See https://blog.cloudflare.com/a-question-of-timing/ # and https://curl.haxx.se/docs/manpage.html for an explanation # of how the curl variables relate to the various stages of # the transfer. # # To get a better approximation of devtool's TTFB, consider # the time without the connection overhead: # %{time_starttransfer} - %{time_appconnect} # # Uses a dirty eval to do the ttfb arithmetic. Depends # on bc and column commands. set -eu # check dependencies for dependency in curl bc column; do which $dependency > /dev/null || (echo "You need to have '$dependency' installed and in your \$PATH" >&2 && exit 1) done # Cribbed from https://stackoverflow.com/a/41762669/391826 median() { arr=($(printf '%s\n' "${@}" | sort -n)) nel=${#arr[@]} if (( $nel % 2 == 1 )); then # Odd number of elements val="${arr[ $(($nel/2)) ]}" else # Even number of elements (( j=nel/2 )) (( k=j-1 )) val=$(echo "scale=6;(${arr[j]}" + "${arr[k]})"/2|bc -l) fi echo $val } # defaults DEBUG="" LOG="" NUM_REQUESTS=0 VERBOSE=0 while getopts ":n:dl:v" OPTION do case $OPTION in d) DEBUG=1 ;; l) LOG="$OPTARG" ;; n) NUM_REQUESTS=$OPTARG ;; v) VERBOSE=1 ;; \?) echo -e "Usage: ttfb [options] url [url...]\n\t-d debug\n\t-l (infers -d) log response headers. Defaults to ./curl.log\n\t-n of times to test time to first byte\n\t-v verbose output. Show response breakdown (DNS lookup, TLS handshake etc)" >&2 exit 1 ;; esac done shift $((OPTIND - 1)) # shifts away every option argument, # leaving urls as $@ if [ -z "$1" ]; then echo "You didn't specify any urls to fetch" exit 1 else URLS="$@" fi # if we're given a custom log file, or log directory, implicitly set DEBUG=1 [ -n "$LOG" ] && DEBUG=1 # default the log file to curl.log in pwd or LOG_DIRECTORY if -o was specified LOG="${LOG:-curl.log}" # now work out if $LOG is relative or an absolute path # and then get the dirname [ "$LOG" != "${LOG#/}" ] && LOG_DIRECTORY=$(dirname "$LOG") || LOG_DIRECTORY=$(dirname "${PWD}/$LOG") if [ ! -d "$LOG_DIRECTORY" ]; then echo "Log directory $LOG_DIRECTORY doesn't exist" >&2 exit 1; fi # then set the actual log filename LOG=$(basename "$LOG") DEBUG=${DEBUG:-0} options=() options+=(-o /dev/null) options+=(-s) options+=(-L) options+=(--http2) options+=(-H 'Cache-Control: no-cache') options+=(-w 'echo DNS lookup: %{time_namelookup} TLS handshake: %{time_appconnect} TTFB including connection: %{time_starttransfer} TTFB: $(echo %{time_starttransfer} - %{time_appconnect} | bc) Total time: %{time_total} \n') if [ $DEBUG -eq 1 ]; then options+=(-D "${LOG_DIRECTORY}/${LOG}") fi for URL in $URLS; do # if we're checking more than one url # output the url on the results line if [ ${#@} -gt 1 ]; then SHOW_URL="${URL}|" if [[ $VERBOSE -eq 1 && -n "$NUM_REQUESTS" && "$NUM_REQUESTS" -gt 1 ]]; then echo $URL >&2 fi else SHOW_URL="" fi # if multiple requests have been specified, then show min, max & median values if [[ -n "$NUM_REQUESTS" && "$NUM_REQUESTS" -gt 1 ]]; then ttfbs=() for i in $(seq $NUM_REQUESTS); do # if we're checking more than one url, and debug is set, then log # the headers to a per-url file, but also for each request if [[ ${#@} -gt 1 && $DEBUG -eq 1 ]]; then LOGFILE="${URL//[^[:alnum:]]/_}" options+=(-D "${LOG_DIRECTORY}/${LOGFILE}-${LOG}_${i}") elif [ $DEBUG -eq 1 ]; then # we only have the one URL, but we still are requesting multiple # ttfb calls, so log the headers options+=(-D "${LOG_DIRECTORY}/${LOG}_${i}") fi request=$(eval $(curl "${options[@]}" "$URL")) ttfbs+=($(echo $request | grep -oE "TTFB: .{0,7}" | cut -d' ' -f2 | sort -n)); if [ $VERBOSE -eq 1 ]; then echo "$request" >&2 else printf "." >&2 fi done # tidy up - combine multiple request logs for the same url into a single file if [[ ${#@} -gt 1 && $DEBUG -eq 1 ]]; then cat "${LOG_DIRECTORY}/${LOGFILE}-${LOG}_"* > "${LOG_DIRECTORY}/${LOGFILE}-${LOG}" rm "${LOG_DIRECTORY}/${LOGFILE}-${LOG}_"* elif [ $DEBUG -eq 1 ]; then cat "${LOG_DIRECTORY}/${LOG}_"* > "${LOG_DIRECTORY}/${LOG}" rm "${LOG_DIRECTORY}/${LOG}_"* fi printf "\n" >&2 # sort the times ttfbs=( $( printf "%s\n" "${ttfbs[@]}" | sort -n ) ) # show quickest, slowest and median fftb printf "${SHOW_URL}\e[32mfastest \e[39m${ttfbs[0]} \e[91mslowest \e[39m${ttfbs[${#ttfbs[*]}-1]} \e[95mmedian \e[39m$(median ${ttfbs[*]})\e[39m\n"; else if [ $VERBOSE -eq 1 ]; then echo -e $SHOW_URL $(eval $(curl "${options[@]}" "$URL")) else echo -e $SHOW_URL $(eval $(curl "${options[@]}" "$URL") | grep -oE "TTFB: .{0,7}" | cut -d' ' -f2) fi fi done | column -s'|' -t