How to sort a list of people by birth and death dates when data are incomplete - mysql

Question

I have a list of people who may or may not have a birth date and/or a death date. I want to be able to sort them meaningfully - a subjective term - by birth date.
BUT - if they don't have a birth date but they to have a death date, I want to have them collated into the list proximal to other people who died then.
I recognize that this is not a discrete operation - there is ambiguity about where someone should go when their birth date is missing. But I'm looking for something that is a good approximation, most of the time.
Here's an example list of what I'd like:
Alice 1800 1830
Bob 1805 1845
Carol 1847
Don 1820 1846
Esther 1825 1860
In this example, I'd be happy with Carol appearing either before or after Don - that's the ambiguity I'm prepared to accept. The important outcome is that Carol is sorted in the list relative to her death date as a death date, not sorting the death dates in with the birth dates.
What doesn't work is if I coalesce or otherwise map birth and death dates together. For example, ORDER BY birth_date, death_date would put Carol after Esther, which is way out of place by my thinking.

Answer 1

I think you're going to have to calculate an average age people end up living (for those having both birth and death dates). And either subtract them from death date or add them to birth date for people who don't have the other one.
Doing this in one query may not be efficient, and perhaps ugly because mysql doesn't have windowing functions. You may be better of precalculating the average living age beforehand. But let's try to do it in one query anyway:
SELECT name, birth_date, death_date
FROM people
ORDER BY COALESCE(
birth_date,
DATE_SUB(death_date, INTERVAL (
SELECT AVG(DATEDIFF(death_date, birth_date))
FROM people
WHERE birth_date IS NOT NULL AND death_date IS NOT NULL
) DAY)
)

Answer 2

N.B.: I've tried with a larger dataset, and it is not working completely as I'd expect.
Try with this query (it needs an id primary key column):
SELECT * FROM people p
ORDER BY (
CASE WHEN birth IS NOT NULL THEN (
SELECT ord FROM (
SELECT id, #rnum := #rnum + 1 AS ord
FROM people, (SELECT #rnum := 0) r1
ORDER BY (CASE WHEN birth IS NOT NULL THEN 0 ELSE 1 END), birth, death
) o1
WHERE id = p.id
) ELSE (
SELECT ord FROM (
SELECT id, #rnum := #rnum + 1 AS ord
FROM people, (SELECT #rnum := 0) r2
ORDER BY (CASE WHEN death IS NOT NULL THEN 0 ELSE 1 END), death, birth
) o2
WHERE id = p.id
)
END)
;
What I've done is, basically, to sort the dataset two times, once by birth date and then by death date. Then I've used these two sorted lists to assign the final order to the original dataset, picking the place from the birth-sorted list at first, and using the place from the death-sorted list when a row has no birth date.
Here's a few problems with that query:
I didn't run it against lots of datasets, so I can't really guarantee it will work with any dataset;
I didn't check its performance, so it could be quite slow on large datasets.
This is the table I've used to write it, tested with MySQL 5.6.21 (I can't understand why, but SQL Fiddle is rejecting my scripts with a Create script error, so I can't provide you with a live example).
Table creation:
CREATE TABLE `people` (
`id` INT(11) NOT NULL AUTO_INCREMENT,
`name` VARCHAR(50) NOT NULL,
`birth` INT(11) NULL DEFAULT NULL,
`death` INT(11) NULL DEFAULT NULL,
PRIMARY KEY (`id`)
);
Data (I actually slightly changed yours):
INSERT INTO `people` (`name`, `birth`, `death`) VALUES ('Alice', 1800, NULL);
INSERT INTO `people` (`name`, `birth`, `death`) VALUES ('Bob', 1805, 1845);
INSERT INTO `people` (`name`, `birth`, `death`) VALUES ('Carol', NULL, 1847);
INSERT INTO `people` (`name`, `birth`, `death`) VALUES ('Don', 1820, 1846);
INSERT INTO `people` (`name`, `birth`, `death`) VALUES ('Esther', 1815, 1860);

Answer 3

you can use a subquery to pick a suitable birthdate for sorting purposes
and then a union to join with the records with a birthdate
for example:
select d1.name, null as birthdate, d1.deathdate, max(d2.birthdate) sort from
d as d1, d as d2
where d1.birthdate is null and d2.deathdate <=d1.deathdate
group by d1.name, d1.deathdate
union all
select name, birthdate, deathdate, birthdate from d
where birthdate is not null
order by 4
http://sqlfiddle.com/#!9/2d91c/1

Answer 4

Not sure if this will work, but worth a try (I can't test this on MySQL) so trying to guess:
order by case birth_date when null then death_date else birth_date end case