Speed up zsh compinit by only checking cache once a day.

compinit.zsh

# On slow systems, checking the cached .zcompdump file to see if it must be 
# regenerated adds a noticable delay to zsh startup.  This little hack restricts 
# it to once a day.  It should be pasted into your own completion file.
#
# The globbing is a little complicated here:
# - '#q' is an explicit glob qualifier that makes globbing work within zsh's [[ ]] construct.
# - 'N' makes the glob pattern evaluate to nothing when it doesn't match (rather than throw a globbing error)
# - '.' matches "regular files"
# - 'mh+24' matches files (or directories or whatever) that are older than 24 hours.
autoload -Uz compinit 
if [[ -n ${ZDOTDIR}/.zcompdump(#qN.mh+24) ]]; then
	compinit;
else
	compinit -C;
fi;

autoload -Uz compinit
for dump in ~/.zcompdump(N.mh+24); do
  compinit
done
compinit -C

works for me.

Timing with timezsh:

timezsh() {
  shell=${1-$SHELL}
  for i in $(seq 1 4); do /usr/bin/time $shell -i -c exit; done
}

Before:

        1.33 real         0.80 user         0.52 sys
        1.38 real         0.80 user         0.57 sys
        1.55 real         0.82 user         0.64 sys
        1.50 real         0.81 user         0.63 sys

After:

        0.78 real         0.32 user         0.42 sys
        0.81 real         0.33 user         0.45 sys
        0.96 real         0.37 user         0.55 sys
        1.21 real         0.42 user         0.71 sys

It was faster here

Definitive solution?

I am going for

() {
    emulate -L zsh
    setopt extendedglob
    autoload -Uz compinit complist
    local zcd=$1                # compdump
    local zcdc=$1.zwc           # compiled compdump
    local zcda=$1.last          # last compilation
    local zcdl=$1.lock          # lock file
    local attempts=30
    while (( attempts-- > 0 )) && ! ln $zcd $zcdl 2> /dev/null; do sleep 0.1; done
    {
        if [[ ! -e $zcda || -n $zcda(#qN.mh+24) ]]; then
            compinit -i -d $zcd
            : > $zcda
        else
            compinit -C -d $zcd
        fi
        [[ ! -f $zcdc || $zcd -nt $zcdc ]] && rm -f $zcdc && zcompile $zcd &!
    } always {
        rm -f $zcdl
    }
} $ZSH/run/u/$HOST-$UID/zcompdump

I didn't see a mention what happens when multiple shells are spawned at the same time. Maybe that's the reason we sometime run into empty dump file.

Where exactly do we put this code:

autoload -Uz compinit
for dump in ~/.zcompdump(N.mh+24); do
  compinit
done
compinit -C

@medwatt put the snippet somewhere in your .zshrc file, or put it in an external script file and call it from your .zshrc

@vincentbernat I solved the multiple shell problem by using a lockfile and a trap to remove it. This ensures that only one process is responsible for updating the file. Even though there's still a race condition, it's now almost impossibly small ¹

() {
	setopt local_options
	setopt extendedglob

	local zcd=${1}
	local zcomp_hours=${2:-24} # how often to regenerate the file
	local lock_timeout=${2:-1} # change this if compinit normally takes longer to run
	local lockfile=${zcd}.lock

	if [ -f ${lockfile} ]; then 
		if [[ -f ${lockfile}(#qN.mm+${lock_timeout}) ]]; then
			(
				echo "${lockfile} has been held by $(< ${lockfile}) for longer than ${lock_timeout} minute(s)."
				echo "This may indicate a problem with compinit"
			) >&2 
		fi
		# Exit if there's a lockfile; another process is handling things
		return
	else
		# Create the lockfile with this shell's PID for debugging
		echo $$ > ${lockfile}
		# Ensure the lockfile is removed
		trap "rm -f ${lockfile}" EXIT
	fi

	autoload -Uz compinit

	if [[ -n ${zcd}(#qN.mh+${zcomp_hours}) ]]; then
		# The file is old and needs to be regenerated
		compinit
	else
		# The file is either new or does not exist. Either way, -C will handle it correctly
		compinit -C
	fi
} ${ZDOTDIR:-$HOME}/.zcompdump 

---

For those who are new to some of this stuff, check out

• https://zsh.sourceforge.io/Doc/Release/Completion-System.html
  • For how compinit works
• https://zsh.sourceforge.io/Doc/Release/Expansion.html#Filename-Generation
  • For the (#q.N.....) syntax, search for "file access qualifier"

Footnotes

1. A second process would have to check for the file in between 2 (effectively) subsequent commands. if [ -f ${lockfile} ] and echo $$ > ${lockfile} ↩

@thefotios my solution is already handling this case (but my message wasn't quite clear on that, so I understand you thought this was not the case). This is the purpose of the ln command (which is atomic, so no race condition).

I commented out these lines in my .zshrc which sped it up a lot:

if type brew &>/dev/null; then
    FPATH=$(brew --prefix)/share/zsh-completions:$FPATH

    autoload -Uz compinit
    compinit
if

As this comment above pointed out, oh-my-zsh already runs compinit.