问题
So lets say I have some records that look like:
2011-01-01 Cat
2011-01-02 Dog
2011-01-04 Horse
2011-01-06 Lion
How can I construct a query that will return 2011-01-03 and 2011-01-05, ie the unused dates. I postdate blogs into the future and I want a query that will show me the days I don't have anything posted yet. It would look from the current date to 2 weeks into the future.
Update:
I am not too excited about building a permanent table of dates. After thinking about it though it seems like the solution might be to make a small stored procedure that creates a temp table. Something like:
CREATE PROCEDURE MISSING_DATES()
BEGIN
CREATE TABLE TEMPORARY DATES (FUTURE DATETIME NULL)
INSERT INTO DATES (FUTURE) VALUES (CURDATE())
INSERT INTO DATES (FUTURE) VALUES (ADDDATE(CURDATE(), INTERVAL 1 DAY))
...
INSERT INTO DATES (FUTURE) VALUES (ADDDATE(CURDATE(), INTERVAL 14 DAY))
SELECT FUTURE FROM DATES WHERE FUTURE NOT IN (SELECT POSTDATE FROM POSTS)
DROP TABLE TEMPORARY DATES
END
I guess it just isn't possible to select the absence of data.
回答1:
You're right — SQL does not make it easy to identify missing data. The usual technique is to join your sequence (with gaps) against a complete sequence, and select those elements in the latter sequence without a corresponding partner in your data.
So, @BenHoffstein's suggestion to maintain a permanent date table is a good one.
Short of that, you can dynamically create that date range with an integers table. Assuming the integers
table has a column i
with numbers at least 0 – 13, and that your table has its date column named datestamp
:
SELECT candidate_date AS missing
FROM (SELECT CURRENT_DATE + INTERVAL i DAY AS candidate_date
FROM integers
WHERE i < 14) AS next_two_weeks
LEFT JOIN my_table ON candidate_date = datestamp
WHERE datestamp is NULL;
回答2:
One solution would be to create a separate table with one column to hold all dates from now until eternity (or whenever you expect to stop blogging). For example:
CREATE TABLE Dates (dt DATE);
INSERT INTO Dates VALUES ('2011-01-01');
INSERT INTO Dates VALUES ('2011-01-02');
...etc...
INSERT INTO Dates VALUES ('2099-12-31');
Once this reference table is set up, you can simply outer join to determine the unused dates like so:
SELECT d.dt
FROM Dates d LEFT JOIN Blogs b ON d.dt = b.dt
WHERE b.dt IS NULL
If you want to limit the search to two weeks in the future, you could add this to the WHERE clause:
AND d.dt BETWEEN NOW() AND ADDDATE(NOW(), INTERVAL 14 DAY)
回答3:
The way to extract rows from the mysql database is via SELECT
. Thus you cannot select rows that do not exist.
What I would do is fill my blog table with all possible dates (for a year, then repeat the process)
create table blog (
thedate date not null,
thetext text null,
primary key (thedate));
doing a loop to create all dates entries for 2011 (using a program, eg $mydate is the date you want to insert)
insert IGNORE into blog (thedate,thetext) values ($mydate, null);
(the IGNORE keyword to not create an error (thedate is a primary key) if thedate exists already).
Then you insert the values normally
insert into blog (thedate,thetext) values ($mydate, "newtext")
on duplicate key update thetext="newtext";
Finally to select empty entries, you just have to
select thedate from blog where thetext is null;
回答4:
You probably not going to like this:
select '2011-01-03', count(*) from TABLE where postdate='2011-01-03'
having count(*)=0 union
select '2011-01-04', count(*) from TABLE where postdate='2011-01-04'
having count(*)=0 union
select '2011-01-05', count(*) from TABLE where postdate='2011-01-05'
having count(*)=0 union
... repeat for 2 weeks
OR
create a table with all days in 2011, then do a left join, like
select a.days_2011
from all_days_2011
left join TABLE on a.days_2011=TABLE.postdate
where a.days_2011 between date(now()) and date(date_add(now(), interval 2 week))
and TABLE.postdate is null;
来源:https://stackoverflow.com/questions/4559817/sql-query-that-returns-all-dates-not-used-in-a-table