Delete all lines which contain any word not used anywhere else in the buffer

Question

Note: For those of you who do not know VHDL do not worry, I will explain the problem with an example.

Motivation

I have a VHDL file which is a big mess of buses. I have declared many signals and I think I have some unused signals. I want to delete unnecessary declarations.

Example

Imagine I have the following signal declarations

signal exmem_wb_we_in : std_logic;
signal exmem_const_in : std_logic_vector(15 downto 0);
signal exmem_data_in : std_logic_vector(15 downto 0);
signal exmem_addr_in : std_logic_vector(15 downto 0);
signal exmem_foo_in : std_logic;
signal exmem_mem_en_in : std_logic;
signal exmem_bar_in : std_logic_vector(15 downto 0);
signal exmem_jump_cond_in : std_logic_vector(3 downto 0);

Say the signals in bold (exmem_foo_in and exmem_bar_in) are never used. Then those words occur only once in the file. I want the lines where they appear to be deleted.

The question

Is it possible to get Vim to delete all lines which hold a single occurrence of a word in a file? The solution should use Vim commands that are 'always available', rather than OS dependent commands. A single command would be preferred. But multiple commands are also welcome.

Some assumptions

• assume only one signal is declared per line;
• there are no occurrences of numbers who appear only once in the file and could get us a line deleted when we in fact want that line;
• there are no comments which introduce a single word occurrence.

Bonus points if you can make it still work without some of those assumptions.

Clarifications:

• I don't know which words I want to delete beforehand.
• A VHDL file might look like this.

Answer 1

Just to delete all lines containing some specific word you can by just typing:

:g/exmem_bar_in/d
:g/exmem_foo_in/d

The signature is

:[range]g[lobal]/{pattern}/[cmd]

(can be found by :h :g)

If you need to know the list of unique words, it actually heavily depends on what you indeed want to treat as a unique word. I doubt that you actually mean any unique word but rather something unique in the signal section. If it is so, then I'd just create a separate buffer, copied eveything there, then remove all besides "signal :" prefix:

:%s/\(.*\):.*/\1/ "deleting all after last semicolon

then leave only unique value:

:sort u.

Now I have a list of unique words I can work with.