MySQL Query: How many times has a candidate won an election? - php

Question

I am writing a web app in PHP using mySQL that models an election.
I have three tables: Candidates, Elections, and Votes. Votes contains CandidateID, ElectionID and Count, which is the number of times that the given candidate was voted for in the given Election. Votes also contains TimeStamp which is the last time the row was modified which is used for breaking ties (the earlier vote wins). A candidate may have run in multiple elections. How do I find how many elections a given candidate has ever won?
All help greatly appreciated, thanks.
Some sample data:
CREATE TABLE IF NOT EXISTS `Votes` (
`ElectionID` int(11) unsigned NOT NULL,
`CandidateID` int(11) unsigned NOT NULL,
`Count` smallint(5) unsigned NOT NULL DEFAULT '0',
`stamp` int(11) unsigned NOT NULL,
PRIMARY KEY (`ElectionID`,`CandidateID`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8;
INSERT INTO `Votes` (`ElectionID`, `CandidateID`, `Count`, `stamp`)
VALUES
(1, 1, 3, 1332897534),
(4, 1, 3, 1333149930),
(4, 4, 2, 1333149947),
(4, 5, 3, 1333149947),
(1, 4, 4, 1333153373);
Desired output: One row, with one column, being the number of wins for a certain candidate

Answer 1

You can write:
SELECT COUNT(1)
FROM Elections AS e
INNER
JOIN Votes AS v1 -- representing the candidate of interest
ON v1.ElectionID = e.ID
AND v1.CandidateID = ...
LEFT
OUTER
JOIN Votes AS v2 -- representing a candidate who beat the candidate of interest
ON v2.ElectionID = e.ID
AND ( v2.Count > v1.Count
OR ( v2.Count = v1.Count
AND v2.stamp < v1.stamp
)
)
WHERE v2.ElectionID IS NULL -- meaning that no candidate beat the candidate of interest
;
(It's also possible to represent either or both of those joins with EXISTS and a correlated subquery; or the first join could be changed to IN with an uncorrelated subquery; but the above is the most likely to perform best, IMHO, and my experience on StackOverflow has been that people seem to like joins better than subqueries for some reason. If you'd prefer a subquery answer, let me know.)

Answer 2

SELECT CandidateID, MAX(Count) FROM Votes GROUP BY ElectionID
should do the trick

Answer 3

Your query basically needs to return every election and WHO Won it. Then apply that result to the specific candidate your are interested in finding out how many that person won out of all elections. Ex: in the U.S. Republican Race, you have 4 candidates... 2 are really the only real considered by most regardless of party affiliation. Each party runs their campaign in each state and they all have their respective votes tallied. So, at the end of ex: 20 states, you will only have 21 winners, but who won how many. Candidate "A" may win 10, "B" wins 6, "C" wins 3 and "D" wins 2. So if you wanted to know how many Candidate "B" won, your answer desired is 6... from my impression of your question.
This will give you all qualifying "First Place" elections for a given candidate. If all you care about is the HOW MANY, you can just change the Prequery.fields to COUNT(*). If you want to get the candidate's name and the name/info of the election, you can add that as join conditions AFTER the PreQuery has been executed.
select
PreQuery.idVotes,
PreQuery.CandidateID,
PreQuery.ElectionID,
PreQuery.Votes,
PreQuery.LastEntry
from
( select
v.*,
#WinRow := if( #LastElection = v.ElectionID, #WinRow +1, 1 ) as FinalPlace,
#LastElection := v.ElectionID as ignoreMe
from
Votes v,
( select #WinRow := 0, #LastElection := 0 ) sqlvars
order by
v.ElectionID,
v.Votes DESC,
v.LastEntry ASC ) PreQuery
where
PreQuery.FinalPlace = 1
AND PreQuery.CandidateID = CandidateIDYouAreInterestedIn

Answer 4

Basically what you want to do is group the Votes rows by ElectionID, and order by Count descending, stamp ascending. That will give you a result set of ordered Votes rows, with the "winners" of each election as the first row within each group.
Next, you want to select these first rows within each group and discard the rest (see here for how to do a Top-N query: http://www.sqlines.com/mysql/how-to/get_top_n_each_group).
Finally, you want to select count(*) from this result set where CandidateID = whatever candidate you're looking for. Alternatively, you can group by CandidateID and leave out the where clause if you want the number of wins for all candidates instead of a specific one.
Hope this helps.

Answer 5

Apparently I'm late to the party, but this is the first thing I though of:
First have one subquery to select the winning count for each unique electionid in Votes, call this table wins.
Then, join wins with Votes where electionid and count are equal. Because there may be a tie, we also need to choose the Votes row with the lowest stamp, so we'll group by electionid and count but this time choose the minimum stamp. We'll call this resulting table wins_with_stamp/wws for short.
Now, wins_with_stamp has all of the rows from Votes that are "winning" rows, so selecting how many a particular candidate won is just a matter of a where candidateid = ? clause.
-- Returns how many Votes rows that is the winner of its election
-- and candidateid is the candidate in question
select count(*)
from Votes v2
right join (
-- Gets the earliest stamp for the votes with the winning count for each election
select v.electionid, v.count, min(v.stamp) as minstamp
from Votes v
right join (
-- Gets the winning count for each election
select electionid, max(count) as max
from Votes
group by electionid
) wins on wins.max = v.count and wins.electionid = v.electionid
group by electionid, count
) wws on wws.count = v2.count and wws.electionid = v2.electionid and wws.minstamp = v2.stamp
where candidateid = [YOUR_CANDIDATEID]