Long before fzf burst on the world
stage, I needed the ability to search on multiple patterns, AND-ed together,
as well as negative patterns (which would also be AND-ed). I also needed
the ability to search not just "line by line" but "paragraph by paragraph";
i.e., treat the para as the unit of search rather just one line.
mg was, and remains, a simple script to do that, plus a couple of other
tricks that I find very useful. The "Usage" message for mg is:
Usage:
mg [options] patt1 [patt2] [-patt3] ...
`mg` allows (a) multiple patterns, including negatives, a la `fzf` (except we
allow "-" or "!" as the negation prefix), (b) filtering based on paragraphs
instead of lines, and (c) taking filenames from STDIN.
-x STDIN is a list of filenames, not data. This is the only way you can
supply a list of filenames, because arguments are ALWAYS search terms.
-p search and report by para instead of by line. (Paragraphs are
separated by one or more blank lines).
-l like grep's -l, but remember filenames only come from STDIN!
Of course, it can't be interactive, like fzf is (I never really got the hang
of curses!), but mg is a part of many of my personal scripts.
I particularly use the "paragraph mode" a lot. The notes you see in this repo are but a small fraction of my enormous collection of tips and tricks and longer personal notes and explanations of all kinds of things I care about, and 99 times out of a hundred, a paragraph search gets me everything I want right there in STDOUT, without me having to browse the actual file.
(Sidenote: many search engines, have the notion of "near". For instance,
foo ~~ barmay, in some hypothetical search syntax, mean "foo within 10 words of bar". This of paragraph search as a slightly different form of that, where nearness is not measured by any notion of words, but of being in the same paragraph. It isn't always better but for most of my needs it has turned out to be better.)