Helix introduced a new feature to their fuzzy finder/picker, called tabular pickers. Basically, the feature adds support for columnar/tabular data with individually filterable columns.

For example, say you want to fuzzy search through ps aux to find process bar by user foo:

USER               PID  %CPU %MEM      VSZ    RSS   TT  STAT STARTED      TIME COMMAND
_windowserver      380  11.1  0.6 413026976  93328   ??  Rs   Wed10AM 216:09.73 /System/Library/PrivateFrameworks/SkyLight.framework/Resources/WindowServer -daemon
root               373   1.1  0.2 426996736  29248   ??  Ss   Wed10AM  18:30.80 /usr/sbin/bluetoothd
root               374   0.9  0.0 426912016   3424   ??  Ss   Wed10AM   7:04.58 /usr/sbin/notifyd
root                 1   0.4  0.1 426852528  15856   ??  Ss   Wed10AM  20:14.09 /sbin/launchd
foo                731   0.4  0.1 426961120  11424   ??  Ss   Wed10AM   2:50.66 /usr/libexec/bar
baz                731   0.4  0.1 426961120  11424   ??  Ss   Wed10AM   2:50.66 /usr/libexec/bar

In Helix, users would filter by specific columns with a query syntax like %COMMAND bar %USER foo (short-hand: %C ba %U fo, or fuzzily %CMD br %USR fo), which would match on /usr/libexec/bar command by user foo:

foo                731   0.4  0.1 426961120  11424   ??  Ss   Wed10AM   2:50.66 /usr/libexec/bar

This is an extremely useful & productive feature because many CLI programs output tabular data (e.g. kubectl, docker, ps, top) and users would be able to filter by arbitrary columns without having to "massage" the data before piping it to fzf (e.g. ps aux | fzf --query '%U foo %C bar' vs ps aux | awk '{print $1"\t"$2"\t"$11}' | fzf --header-lines=1).

The closest I was able to get was using --nth, like ps aux | fzf --nth 1,11 --header-lines 1 --query 'foo bar', but this approach is limited to pre-determined columns. Users cannot filter by arbitrary columns.

Unless I am missing a way that users can dynamically change --nth, it would be nice for this feature to be supported by fzf, either via a search syntax (e.g. %<column> <term>) or a bind + action (e.g. --bind "tab:increment-nth()" / --bind "shift-tab:decrement-nth()".