substring_index skips delimiter from right - mysql

Question

I have a table 'car_purchases' with a 'description' column. The column is a string that includes first name initial followed by full stop, space and last name.
An example of the Description column is
'Car purchased by J. Blow'
I am using 'substring_index' function to extract the letter preceding the '.' in the column string. Like so:
SELECT
Description,
SUBSTRING_INDEX(Description, '.', 1) as TrimInitial,
SUBSTRING_INDEX(
SUBSTRING_INDEX(Description, '.', 1),' ', -1) as trimmed,
length(SUBSTRING_INDEX(
SUBSTRING_INDEX(Description, '.', 1),' ', -1)) as length
from car_purchases;
I will call this query 1.
picture of the result set (Result 1) is as follows
As you can see the problem is that the 'trimmed' column in the select statement starts counting the 2nd delimiter ' ' instead of the first from the right and produces the result 'by J' instead of just 'J'. Further the length column indicates that the string length is 5 instead of 4 so WTF?
However when I perform the following select statement;
select SUBSTRING_INDEX(
SUBSTRING_INDEX('Car purchased by J. Blow', '.', 1),' ', -1); -- query 2
Result = 'J' as 'Result 2'.
As you can see from result 1 the string in column 'Description' is exactly (as far as I can tell) the same as the string from 'Result 2'. But when the substring_index is performed on the column (instead of just the string itself) the result ignores the first delimiter and selects a string from the 2nd delimiter from the right of the string.
I've racked my brains over this and have tried 'by ' and ' by' as delimiters but both options do not produce the desired result of a single character. I do not want to add further complexity to query 1 by using a trim function. I've also tried the cast function on result column 'trimmed' but still no success. I do not want to concat it either.
There is an anomaly in the 'length' column of query 1 where if I change the length function to char_length function like so:
select length(SUBSTRING_INDEX(
SUBSTRING_INDEX(Description, '.', 1),' ', -1)) as length -- result = 5
select char_length(SUBSTRING_INDEX(
SUBSTRING_INDEX(Description, '.', 1),' ', -1)) as length -- result = 4
Can anyone please explain to me why the above select statement would produce 2 different results? I think this is the reason why I am not getting my desired result.
But just to be clear my desired outcome is to get 'J' not 'by J'.
I guess I could try reverse but I dont think this is an acceptable compromise. Also I am not familiar with collation and charset principles except that I just use the defaults.
Cheers Players!!!!

Answer 1

CHAR_LENGTH returns length in characters, so a string with 4 2-byte characters would return 4. LENGTH however returns length in bytes, so a string with 4 2-byte characters would return 8. The discrepancy in your results (including SUBSTRING_INDEX) says that the "space" between by and J is not actually a single-byte space (ASCII 0x20) but a 2-byte character that looks like a space. To workaround this, you could try replacing all unicode characters with spaces using CONVERT and REPLACE. In this example, I have an en-space unicode character in the string between by and J. The CONVERT changes that to a ?, and the REPLACE then converts that to a space:
SELECT SUBSTRING_INDEX( SUBSTRING_INDEX("Car purchased by J. Blow", '.', 1),' ', -1)
Output:
by J
With CONVERT and REPLACE:
SELECT SUBSTRING_INDEX( SUBSTRING_INDEX(REPLACE(CONVERT("Car purchased by J. Blow" USING ASCII), '?', ' '), '.', 1),' ', -1)
Output
J
For your query, you would replace the string with your column name i.e.
SELECT SUBSTRING_INDEX( SUBSTRING_INDEX(REPLACE(CONVERT(description USING ASCII), '?', ' '), '.', 1),' ', -1)
Demo on DBFiddle