How do FULLTEXT INDEXES on multiple columns work?

Question

When adding a FULLTEXT INDEX on 3 columns, does that add 1 single index on 3 columns, or does it add 3 separate indexes?

I ask this, because at the moment I'm using FULLTEXT like this:

ALTER TABLE myTable ADD FULLTEXT all3colsIndex (col1,col2,col3);
SELECT * FROM myTable WHERE MATCH (col1, col2, col3) AGAINST ('word');

I have just added a user option to my search interface where the user can remove one of the columns from the search. So am I able to do this without losing the index, where I'm only using 2 of the 3 columns:

ALTER TABLE myTable ADD FULLTEXT all3colsIndex (col1,col2,col3);
If (UserOptionRemoveCol == "selected") {
    SELECT * FROM myTable WHERE MATCH (col1, col2) AGAINST ('word');
} else {
    SELECT * FROM myTable WHERE MATCH (col1, col2, col3) AGAINST ('word');
}

Or would I have to create a new FULLTEXT index on the two columns, as well as the three?

ALTER TABLE myTable ADD FULLTEXT all3colsIndex (col1,col2,col3);
ALTER TABLE myTable ADD FULLTEXT 2colsIndex (col1,col2);

Answer 1

Glancing over the manual for CREATE FULLTEXT INDEX, it indicates that you are able to specify multiple columns by repeating the column_name as such:

CREATE FULLTEXT INDEX ON table_name (column_name1 [...], column_name2 [...]) ...

Given this information, I would assume it creates a single index across 3 columns. Further, I'm assuming that it works under the left-to-right rule with regards to composite indexes (I would verify this by checking the execution plan for the following statements). Therefore, a composite index on (col1, col2, col3) would have to be selected in that order for it to be used (SELECT col1, col2 ...). If you were to call col2 it would not use the index.

Answer 2

Only one index, of type FULLTEXT, will be created. That index will "span" around multiple columns if necessary. However, the indexes themselves are not indexing the columns as much as their words. From the documentation:

InnoDB FULLTEXT indexes have an inverted index design. Inverted indexes store a list of words, and for each word, a list of documents that the word appears in. [...]

If you create one FULLTEXT grouping three columns, or three FULLTEXT grouping only one, these columns will in effect consist of entries in the FULLTEXT meta-data tables.

When incoming documents are tokenized, the individual words (also referred to as “tokens”) are inserted into the index tables along with position information and the associated Document ID (DOC_ID).

Basically, it seems multi-column FULLTEXT vs multiple single-column FULLTEXT indexes work the same, as the storage of the words is abstracted from the original separation of the columns.