#1071 – Specified key was too long; max key length is 1000 bytes

As @Devart says, the total length of your index is too long.

The short answer is that you shouldn’t be indexing such long VARCHAR columns anyway, because the index will be very bulky and inefficient.

The best practice is to use prefix indexes so you’re only indexing a left substring of the data. Most of your data will be a lot shorter than 255 characters anyway.

You can declare a prefix length per column as you define the index. For example:

...
KEY `index` (`parent_menu_id`,`menu_link`(50),`plugin`(50),`alias`(50))
...

But what’s the best prefix length for a given column? Here’s a method to find out:

SELECT
 ROUND(SUM(LENGTH(`menu_link`)<10)*100/COUNT(`menu_link`),2) AS pct_length_10,
 ROUND(SUM(LENGTH(`menu_link`)<20)*100/COUNT(`menu_link`),2) AS pct_length_20,
 ROUND(SUM(LENGTH(`menu_link`)<50)*100/COUNT(`menu_link`),2) AS pct_length_50,
 ROUND(SUM(LENGTH(`menu_link`)<100)*100/COUNT(`menu_link`),2) AS pct_length_100
FROM `pds_core_menu_items`;

It tells you the proportion of rows that have no more than a given string length in the menu_link column. You might see output like this:

+---------------+---------------+---------------+----------------+
| pct_length_10 | pct_length_20 | pct_length_50 | pct_length_100 |
+---------------+---------------+---------------+----------------+
|         21.78 |         80.20 |        100.00 |         100.00 |
+---------------+---------------+---------------+----------------+

This tells you that 80% of your strings are less than 20 characters, and all of your strings are less than 50 characters. So there’s no need to index more than a prefix length of 50, and certainly no need to index the full length of 255 characters.

PS: The INT(1) and INT(32) data types indicates another misunderstanding about MySQL. The numeric argument has no effect related to storage or the range of values allowed for the column. INT is always 4 bytes, and it always allows values from -2147483648 to 2147483647. The numeric argument is about padding values during display, which has no effect unless you use the ZEROFILL option.

Leave a Comment