Is it possible to make a simple query to count how many records I have in a determined period of time like a year, month, or day, having a TIMESTAMP field, like:
SELECT COUNT(id)
FROM stats
WHERE record_date.YEAR = 2009
GROUP BY record_date.YEAR
Or even:
SELECT COUNT(id)
FROM stats
GROUP BY record_date.YEAR, record_date.MONTH
To have a monthly statistic.
Thanks!
GROUP BY DATE_FORMAT(record_date, '%Y%m')Note (primarily, to potential downvoters). Presently, this may not be as efficient as other suggestions. Still, I leave it as an alternative, and a one, too, that can serve in seeing how faster other solutions are. (For you can't really tell fast from slow until you see the difference.) Also, as time goes on, changes could be made to MySQL's engine with regard to optimisation so as to make this solution, at some (perhaps, not so distant) point in future, to become quite comparable in efficiency with most others.
try this one
SELECT COUNT(id)
FROM stats
GROUP BY EXTRACT(YEAR_MONTH FROM record_date)
EXTRACT(unit FROM date) function is better as less grouping is used and the function return a number value.
Comparison condition when grouping will be faster than DATE_FORMAT function (which return a string value). Try using function|field that return non-string value for SQL comparison condition (WHERE, HAVING, ORDER BY, GROUP BY).
I tried using the 'WHERE' statement above, I thought its correct since nobody corrected it but I was wrong; after some searches I found out that this is the right formula for the WHERE statement so the code becomes like this:
SELECT COUNT(id)
FROM stats
WHERE YEAR(record_date) = 2009
GROUP BY MONTH(record_date)
If your search is over several years, and you still want to group monthly, I suggest:
version #1:
SELECT SQL_NO_CACHE YEAR(record_date), MONTH(record_date), COUNT(*)
FROM stats
GROUP BY DATE_FORMAT(record_date, '%Y%m')
version #2 (more efficient):
SELECT SQL_NO_CACHE YEAR(record_date), MONTH(record_date), COUNT(*)
FROM stats
GROUP BY YEAR(record_date)*100 + MONTH(record_date)
I compared these versions on a big table with 1,357,918 rows (innodb), and the 2nd version appears to have better results.
version1 (average of 10 executes): 1.404 seconds
version2 (average of 10 executes): 0.780 seconds
(SQL_NO_CACHE key added to prevent MySQL from CACHING to queries.)
If you want to filter records for a particular year (e.g. 2000) then optimize the WHERE clause like this:
SELECT MONTH(date_column), COUNT(*)
FROM date_table
WHERE date_column >= '2000-01-01' AND date_column < '2001-01-01'
GROUP BY MONTH(date_column)
-- average 0.016 sec.
Instead of:
WHERE YEAR(date_column) = 2000
-- average 0.132 sec.
The results were generated against a table containing 300k rows and index on date column.
As for the GROUP BY clause, I tested the three variants against the above mentioned table; here are the results:
SELECT YEAR(date_column), MONTH(date_column), COUNT(*)
FROM date_table
GROUP BY YEAR(date_column), MONTH(date_column)
-- codelogic
-- average 0.250 sec.
SELECT YEAR(date_column), MONTH(date_column), COUNT(*)
FROM date_table
GROUP BY DATE_FORMAT(date_column, '%Y%m')
-- Andriy M
-- average 0.468 sec.
SELECT YEAR(date_column), MONTH(date_column), COUNT(*)
FROM date_table
GROUP BY EXTRACT(YEAR_MONTH FROM date_column)
-- fu-chi
-- average 0.203 sec.
The last one is the winner.
You can do this simply Mysql DATE_FORMAT() function in GROUP BY. You may want to add an extra column for added clarity in some cases such as where records span several years then same month occurs in different years.Here so many option you can customize this. Please read this befor starting. Hope it should be very helpful for you. Here is sample query for your understanding
SELECT
COUNT(id),
DATE_FORMAT(record_date, '%Y-%m-%d') AS DAY,
DATE_FORMAT(record_date, '%Y-%m') AS MONTH,
DATE_FORMAT(record_date, '%Y') AS YEAR
FROM
stats
WHERE
YEAR = 2009
GROUP BY
DATE_FORMAT(record_date, '%Y-%m-%d ');
I prefer to optimize the one year group selection like so:
SELECT COUNT(*)
FROM stats
WHERE record_date >= :year
AND record_date < :year + INTERVAL 1 YEAR;
This way you can just bind the year in once, e.g. '2009', with a named parameter and don't need to worry about adding '-01-01' or passing in '2010' separately.
Also, as presumably we are just counting rows and id is never NULL, I prefer COUNT(*) to COUNT(id).
Here's one more approach. This uses [MySQL's LAST_DAY() function][1] to map each timestamp to its month. It also is capable of filtering by year with an efficient range-scan if there's an index on record_date.
SELECT LAST_DAY(record_date) month_ending, COUNT(*) record_count
FROM stats
WHERE record_date >= '2000-01-01'
AND record_date < '2000-01-01' + INTERVAL 1 YEAR
GROUP BY LAST_DAY(record_date)
If you want your results by day, use DATE(record_date) instead.
If you want your results by calendar quarter, use YEAR(record_date), QUARTER(record_date).
Here's a writeup. https://www.plumislandmedia.net/mysql/sql-reporting-time-intervals/ [1]: https://dev.mysql.com/doc/refman/8.0/en/date-and-time-functions.html#function_last-day
I wanted to get similar data per day, after experimenting a bit, this is the fastest I could find for my scenario
SELECT COUNT(id)
FROM stats
GROUP BY record_date DIV 1000000;
If you want to have it per month, add additional zeroes (00) I would not recommend this from "make the code readable" perspective, it might also break in different versions. But in our case this took less then half the time compared to some other more clearer queries that I tested.
This is a MySQL answer (as MySQL is tagged in the question) and well documented in the manual https://dev.mysql.com/doc/refman/8.0/en/date-and-time-type-conversion.html
GROUP BY record_date.MONTHin your first code snippet? – chiccodoro