How to scroll large datafile while keeping header

On terminals that support setting the scrolling region:

tailf() ( # args: <file> [<number-of-header-lines>]
  trap 'tput csr 0 "$((LINES-1))"' INT
  tput csr "$((1+${2-1}))" "$((LINES-1))"
  tput clear
  {
    head -n"${2-1}"
    printf "%${COLUMNS}s\n" "" | tr ' ' =
    tail -n "$((LINES-1-${2-1}))" -f
  } < "$1"
)

(assumes a shell like zsh or bash that sets the $COLUMNS and $LINES variables based on the size of the terminal).

If you're familiar with vim, this is probably the best option for you. You can enable horizontal-scroll-bind-only by changing 'scrollopt':

set scrollopt=hor

So with vim -u NONE, you get the desired behavior with:

:set scrollopt=hor
:set nowrap
:1split
:windo set scrollbind

You may want to adjust 'sidescroll' and 'sidescrolloff' to change how many columns are skipped and how far from the edge skipping starts respectively.

Try this (you'll need to install multitail):

multitail -du -t "$(head -n 1 filename)" filename

or, for headers longer than one line:

multitail -wh 2 -l "head -n 2 filename" filename

If you want to follow command output instead of a file:

multitail -wh 2 -l "command | head -n 2" -l command

or use -t as appropriate. Note that you may need to use unbuffer so your command output appears immediately.

You can use -D to disable the display of status lines for the -wh forms (it would defeat the purpose of the -t form).