MySQL use one subquery more times - mysql

Question

I'm just beginner in SQL, so please don't worry :)
I have table where I store dictionary word translates. The table have columns id, la (language), lb (target language), wa (original word), wb (translated word) and some more insignificant columns :). I want to show a table overview, where will be a list of all languages with count of words in each language plus row with SUM as a language with total words count (all counts of only distinct words).
I wrote this query:
SELECT `lng`, COUNT(`word`) AS `count`
FROM (
SELECT DISTINCT *
FROM (
SELECT DISTINCT `la` AS `lng`, `wa` AS `word`
FROM `dict_trans`
UNION ALL
SELECT DISTINCT `lb` AS `lng`, `wb` AS `word`
FROM `dict_trans`
) AS `tbla`
) AS `tblb` GROUP BY `lng`
UNION ALL
SELECT 'sum' AS `lng`, COUNT(`word`) AS `count`
FROM (
SELECT DISTINCT *
FROM (
SELECT DISTINCT `la` AS `lng`, `wa` AS `word`
FROM `dict_trans`
UNION ALL
SELECT DISTINCT `lb` AS `lng`, `wb` AS `word`
FROM `dict_trans`
) AS `tblc`
) AS `tbld` ORDER BY `count` DESC
But I think it's very silly and performance unfriendly doing one subquery more times.
SELECT DISTINCT *
FROM (
SELECT DISTINCT `la` AS `lng`, `wa` AS `word`
FROM `dict_trans`
UNION ALL
SELECT DISTINCT `lb` AS `lng`, `wb` AS `word`
FROM `dict_trans`
) AS `tbla`
I tried in second part of code to pass reference to the table from first part:
SELECT `lng`, COUNT(`word`) AS `count`
FROM (
SELECT DISTINCT *
FROM (
SELECT DISTINCT `la` AS `lng`, `wa` AS `word`
FROM `dict_trans`
UNION ALL
SELECT DISTINCT `lb` AS `lng`, `wb` AS `word`
FROM `dict_trans`
) AS `tbla`
) AS `tblb` GROUP BY `lng`
UNION ALL
SELECT 'sum' AS `lng`, COUNT(`word`) AS `count`
FROM `tblb`
ORDER BY `count` DESC
But error was thrown (#1146 - Table 'db.tblb' doesn't exist).
It's possible to solve this problem without of creating temporary tables?

Answer 1

You don't need all the complexity of your query. UNION means UNION DISTINCT so there is no need for the 3-level nesting and the extra count can be done with the WITH ROLLUP modifier:
SELECT lng AS language,
COUNT(word) AS WordCount
FROM
( SELECT la AS lng, wa AS word
FROM dict_trans
UNION -- DISTINCT is the default
SELECT lb, wb
FROM dict_trans
) AS t
GROUP BY lng
WITH ROLLUP ;
And if you want sorting, you'll need another nesting:
SELECT language, WordCount
FROM
( SELECT COALESCE(lng, 'Total') AS language,
COUNT(word) AS WordCount
FROM
( SELECT la AS lng, wa AS word
FROM dict_trans
UNION -- DISTINCT is the default
SELECT lb, wb
FROM dict_trans
) AS t
GROUP BY lng
WITH ROLLUP
) AS tmp
ORDER BY CASE WHEN language = 'Total' THEN 1 ELSE 0 END,
WordCount DESC ;
The most expensive part of this query will be the UNION (because you need DISTINCT) and the GROUP BY operations. If you want efficiency, it would be much better to have a separate table with all distinct language and word combinations and then group by on that table (no union would be needed). And the dict_trans would be a "junction" table and you'd have 2 foreign keys, pointing to that new distinct languages-words table.