Different mapping behavior for the different visual modes?

Question

I want to have different behavior for the different visual modes. As an example if I wanted to vnoremap S{ so that it has the following behavior:

• surrounds the selection in v mode, (puts { at the start of selection and and } at the end of selection)
• surrounds the lines in V mode, (puts { above and } below the selected lines)
• surrounds the block in ctrl-v mode, (puts{s on the left edge and and }s on the right edge)

what would that look like? I have this but its buggy (behavior isn't consistent, seems to toggle behavior on every attempt.

vnoremap <expr> S{ visualmode() ==# "v" ?
\ "<Esc>`>a}<Esc>`<i{<Esc>" : visualmode() ==# "V" ?
\ "<Esc>`>o}<Esc>`<O{<Esc>va{=" :
\ "A}<Esc>gvI{<Esc>"

Answer 1

You use visualmode() which returns the last used visual mode, not the current one.

                   vvvvvvvvvvvv
vnoremap <expr> S{ visualmode() ==# "v" ?
\ "<Esc>`>a}<Esc>`<i{<Esc>" : visualmode() ==# "V" ?
                              ^^^^^^^^^^^^

You need the current one, which is given by mode():

                   vvvvvv
vnoremap <expr> S{ mode() ==# "v" ?
\ "<Esc>`>a}<Esc>`<i{<Esc>" : mode() ==# "V" ?
                              ^^^^^^

---

Unrelated to the issue, but you don't need to call [visual]mode() several times, and xnoremap is better than vnoremap:

xno <expr> S{ {
\  'v': "\e`>a}\e`<i{\e",
\  'V': "\e`>o}\e`<O{\eva{=",
\  '<c-v>': "A}\egvI{\e",
\ }[mode()]