The keymap harness caught a regression in 5b6f789: returning '' from
the <expr> mapping and deferring cursor placement to vim.schedule
breaks any keypress immediately following Tab/<CR>. nvim_feedkeys with
the 'x' flag drains all queued keys synchronously, so the user's next
key lands at the pre-jump cursor position — and the scheduled
callback's startinsert leaks insert mode into the next test.
Mirror what the VimL side already does: have <expr> return
"<Cmd>lua require('nuwiki.commands')._helper(args)<CR>" so the
align + cursor-jump runs synchronously inside the same keypress. No
schedule, no mode juggling.
cr.adds_new_table_row ✓
tab.next_cell_in_table ✓
tab.from_first_cell_moves_to_second ✓
shift_tab.prev_cell_in_table ✓
39/39 in scripts/test-keymaps.sh (was 35/4-fail).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
gffranco
695702555f