How to speed up grep on large history?

First, if you aren't doing anything more complex than grep, limiting matches with history -m PATTERN may be faster than grep. Even if you're applying complex filters afterwards, try adding a simple filter with -m that gets rid of most undesired lines.

If the limiting factor is the time spent printing out the full history with history, there's a hackish way to print the part of the history selected by grep (or other filter). Use fc -p to start a new history, and make it read the selected part of the history. Untested:

function format_history {
  fc -p -a =(cat)
  fc -l 1
}
history 1 | grep curl | format_history

This isn't very convenient to use (you need something on both sides of the pipeline, so you can't just have a magic_history_filter prefix). And what speed it gains in not formatting the whole history, it might lose in having fancier plumbing and requiring the use of a temporary file (zsh can only read the history from a seekable file, not from a pipe).

Here's a simple function that will let you quickly search your entire history for any string:

histgrep() {
  fc -lm "*${1}*" 1 -1
}

or even faster:

histgrep() {
  zmodload -i zsh/parameter
  print -raC2 ${(kv)history[(R)*$1*]}
}

Either of these outputs a history line number in front of each line, so you can execute the line you want with !<number>. For example to insert history line number 2596 into your command line, do

!2596

then press Tab or Enter.

I have around 3400 lines of history. Doing histgrep cd produces over 220 lines for me but takes only about 0.004 seconds to complete with the first version of the function and half that with the second version.

From time to time you can run history >~/.zsh_history.cache. Then you can grep over that file plus the output of history for those commands missing in that file.

I am not familiar with zsh and thus cannot provide code for combining the file content with the command output.

You could try ripgrep ( https://github.com/BurntSushi/ripgrep ) Is actually faster and has many useful filtering options: the github page has a long list of features.

It's in the repositories of many distributions.