Custom delimiter workaround

[daneofmanythings]

I would like to be able to specify a delimiter. I see a closed issue regarding this from 7 years ago. Here is my use-case:
I feed the output of rg directly into fzf. fzf uses a delimiter to break each input line into segments as a means of customization. Mostly my results contain 4 segments: path, lnum, col, line. I see that there is an option for a NULL delimited path. That does not work for my use case. As a workaround, I feed each line through sed and replace : with 0x1f and hope there isn't a false positive. Is there a better way I could be doing this? I would think that rg > fzf is a fairly popular use-case and there may be a better way to do this. As an aside, my workflow is integrated into neovim, and here is the parameterized command I use:

local grep_cmd = function(opts)
    local cmd = [[
        rg \
        --trim \
        --no-heading \
        --smart-case \
        %s \
        --fixed-strings \
        %s \
        %s \
        --regexp="%s" \
        %s \
        | \
        sed 's/:/%s/; s/:/%s/; s/:/%s/' \
    ]]


    return string.format(
        cmd,
        opts.without_linenr and "--no-line-number" or "--column",
        opts.iglob and "--iglob=" .. opts.iglob or "",
        opts.with_filename and "--with-filename" or "",
        opts.pattern or ".",
        vim.fn.join(opts.search_in) or ".",
        delim,
        delim,
        delim
    )
end

[BurntSushi]

ripgrep's default output, like grep, isn't easily machine readable that is robust against false positives. Unlike grep, ripgrep provides a stable JSON output format via the --json flag. This is a robust way to get a zero false positive rate.

That is where I would suggest looking if you truly need something robust. There is a big gap between the default output format and JSON though. So there may be some work involved.

The middle ground is maintaining a tiny targeted patch in a fork of ripgrep that you control with the functionality you need.

[daneofmanythings]

Thank for the response! fzf is just a unix line filter. I would need to do much more complex preprocessing than just a sed call if I switched to a json output. I was hoping to avoid maintaining my own fork hah. Can you explain why a --delimiter option is out of scope for main? I'm guessing my use-case is fleetingly uncommon to justify the maintenance?

[ltrzesniewski]

How about rg --json | jq ... | fzf?

[daneofmanythings]

Thanks for the suggestion, but sed already works for me as an intermediate. I was hoping to remove the intermediate.