Related
Consider 2 tables, user and contact:
CREATE TABLE user (
id INT,
name VARCHAR(36)
);
CREATE TABLE contact (
id INT,
phone INT,
userID INT
);
INSERT INTO user (id,name) VALUES
(1,'Frank'),
(2,'Henry'),
(3,'John')
INSERT INTO contact (id,phone,userID) VALUES
(1,911,1),
(2,922,2),
(3,933,2)
I am interested in all user entries that do not have any contact.
The outer join of these tables returns 4 results:
SELECT contact.*, user.*
FROM contact
LEFT JOIN user
ON contact.userID = user.id
UNION ALL
SELECT contact.*, user.*
FROM user
LEFT JOIN contact
ON contact.userID = user.id
WHERE contact.userID IS NULL
How do I select all user where contact.userID is null (1 result, in this example)?
This is your second subquery:
SELECT c.*, u.*
FROM user u LEFT JOIN
contact c
ON c.userID = u.id
WHERE c.userID IS NULL;
It does exactly what you want, although I would remove the c.* from the SELECT.
How to find city when ContactID is provided and condition is if ContactID is coming as 123 then it will look whether it is P or C, If P then it will go to Person table and returns City(USA) as output and If C then it will go to Company table and gives City(AUS) as output.
NB: all tables contain thousands of record and City value comes from run time.
Unless you're dynamically generating the query (i.e. using some language other than SQL to execute it) then you need to join on both tables anyway. If you're joining on both tables then there's no need for a CASE statement:
select *
from contacts co
left outer join person p
on co.contactid = p.contactid
and co.person_company = 'P'
left outer join company c
on co.contactid = c.contactid
and co.person_company = 'C'
You'll start noting an issue here, for every column from PERSON and COMPANY you're going to have to add some business logic to work out which table you want the information from. This can get very tiresome
select co.contactid
, case when p.id is not null then p.name else c.name end as name
from contacts co
left outer join person p
on co.contactid = p.contactid
and co.person_company = 'P'
left outer join company c
on co.contactid = c.contactid
and co.person_company = 'C'
Your PERSON and COMPANY tables seem to have exactly the same information in them. If this is true in your actual data model then there's no need to split them up. You make the determination as to whether each entity is a person or a company in your CONTACTS table.
Creating additional tables to store data in this manner is only really helpful if you need to store additional data. Even then, I'd still put the data that means the same thing for a person or a companny (i.e. name or address) in a single table.
If there's a 1-2-1 relationship between CONTACTID and PID and CONTACTID and CID, which is what your sample data implies, then you have a number of additional IDs, which have no value.
Lastly, if you're not restricting that only companies can go in the COMPANY table and individuals in the PERSON table. You need the PERSON_COMPANY column to exist in both PERSON and COMPANY, though as a fixed string. It would be more normal to set up this data model as something like the following:
create table contacts (
id integer not null
, contact_type char(1) not null
, name varchar2(4000) not null
, city varchar2(3)
, constraint pk_contacts primary key (id)
, constraints uk_contacts unique (id, contact_type)
);
create table people (
id integer not null
, contact_type char(1) not null
, some_extra_info varchar2(4000)
, constraint pk_people primary key (id)
, constraint fk_people_contacts
foreign key (id, contact_type)
references contacts (id, contact_type)
, constraint chk_people_type check (contact_type = 'P')
);
etc.
you can LEFT JOIN all 3 tables and the using a CASE statement select the one that you need based on the P or C value
SELECT
CASE c.[Person/Company]
WHEN 'P' THEN p.NAME
WHEN 'C' THEN a.Name
END AS Name
FROM Contact c
LEFT JOIN Person p on p.ContactId = c.ContactId
LEFT JOIN Company a on a.ContachId = c.ContactId
Ben's answer is almost right. You might want to check that the first join has no match before doing the second one:
select c.*, coalesce(p.name, c.name) as p.name
from contacts c left outer join
person p
on c.contactid = p.contactid and
c.person_company = 'P' left join
company co
on c.contactid = co.contactid and
c.person_company = 'C' and
p.contactid is null;
This may not be important in your case. But in the event that the second join matches multiple rows and the first matches a single row, you might not want the additional rows in the output.
I have a problem with a result set of two tables.
I create two tables.
create table person
(ID int not null PRIMARY KEY,
NAME varchar (50)
)
create table address
( ID int NOT NULL PRIMARY KEY,
PERSON_ID int not null,
ADDRESS VARCHAR (20) null,
ZIP int)
insert into person (ID, NAME) values ( 1,'Hans')
insert into person values (2,'Peter')
insert into address values (1,1, 'Andernach',56626)
insert into address values (2,2,'Koblenz',56000)
insert into address values (3,3,'Neuwied',56100)
I would like to get a result set of these tables.
If I use this query with p.ID = a.PERSON_ID
I get right result.
BUT if I use query with p.ID != a.PERSON_ID
I get a lot of results
select a.ADDRESS,
a.ZIP,
a.PERSON_ID,
p.NAME
from address a,
person p
where a.PERSON_ID != p.ID
My question: Why do I get a lot of results with !=?
Thanks
:-)
Because lots of rows in the first table don't match lots of rows in the second. In fact, almost all rows in the second will match each in the first. Presumably you want:
select a.ADDRESS,
a.ZIP,
a.PERSON_ID,
p.NAME
from person p left join
address a
on a.PERSON_ID = p.ID
where p.id is null;
This gets rows in person that don't have an address. (Putting the tables in the other order or using right join would get addresses with no person.)
My 'people' table has one row per person, and that person has a division (not unique) and a company (not unique).
I need to join people to p_features, c_features, d_features on:
people.person=p_features.num_value
people.division=d_features.num_value
people.company=c_features.num_value
... in a way that if there is a record match in p_features/d_features/c_features only, it would be returned, but if it was in 2 or 3 of the tables, the most specific record would be returned.
From my test data below, for example, query for person=1 would return
'FALSE'
person 3 returns maybe, person 4 returns true, and person 9 returns default
The biggest issue is that there are 100 features and I have queries that need to return all of them in one row. My previous attempt was a function which queried on feature,num_value in each table and did a foreach, but 100 features * 4 tables meant 400 reads and it brought the database to a halt it was so slow when I loaded up a few million rows of data.
create table p_features (
num_value int8,
feature varchar(20),
feature_value varchar(128)
);
create table c_features (
num_value int8,
feature varchar(20),
feature_value varchar(128)
);
create table d_features (
num_value int8,
feature varchar(20),
feature_value varchar(128)
);
create table default_features (
feature varchar(20),
feature_value varchar(128)
);
create table people (
person int8 not null,
division int8 not null,
company int8 not null
);
insert into people values (4,5,6);
insert into people values (3,5,6);
insert into people values (1,2,6);
insert into p_features values (4,'WEARING PANTS','TRUE');
insert into c_features values (6,'WEARING PANTS','FALSE');
insert into d_features values (5,'WEARING PANTS','MAYBE');
insert into default_features values('WEARING PANTS','DEFAULT');
You need to transpose the features into rows with a ranking. Here I used a common-table expression. If your database product does not support them, you can use temporary tables to achieve the same effect.
;With RankedFeatures As
(
Select 1 As FeatureRank, P.person, PF.feature, PF.feature_value
From people As P
Join p_features As PF
On PF.num_value = P.person
Union All
Select 2, P.person, PF.feature, PF.feature_value
From people As P
Join d_features As PF
On PF.num_value = P.division
Union All
Select 3, P.person, PF.feature, PF.feature_value
From people As P
Join c_features As PF
On PF.num_value = P.company
Union All
Select 4, P.person, DF.feature, DF.feature_value
From people As P
Cross Join default_features As DF
)
, HighestRankedFeature As
(
Select Min(FeatureRank) As FeatureRank, person
From RankedFeatures
Group By person
)
Select RF.person, RF.FeatureRank, RF.feature, RF.feature_value
From people As P
Join HighestRankedFeature As HRF
On HRF.person = P.person
Join RankedFeatures As RF
On RF.FeatureRank = HRF.FeatureRank
And RF.person = P.person
Order By P.person
I don't know if I had understood very well your question, but to use JOIN, you need your table loaded already and then use the SELECT statement with INNER JOIN, LEFT JOIN or whatever you need to show.
If you post some more information, maybe turn it easy to understand.
There are some aspects of your schema I'm not understanding, like how to relate to the default_features table if there's no match in any of the specific tables. The only possible join condition is on feature, but if there's no match in the other 3 tables, there's no value to join on. So, in my example, I've hard-coded the DEFAULT since I can't think of how else to get it.
Hopefully this can get you started and if you can clarify the model a bit more, the solution can be refined.
select p.person, coalesce(pf.feature_value, df.feature_value, cf.feature_value, 'DEFAULT')
from people p
left join p_features pf
on p.person = pf.num_value
left join d_features df
on p.division = df.num_value
left join c_features cf
on p.company = cf.num_value
I have a table where each row has a few fields that have ID's that relate to some other data from some other tables.
Let's say it's called people, and each person has the ID of a city, state and country.
So there will be three more tables, cities, states and countries where each has an ID and a name.
When I'm selecting a person, what's the easiest way to get the names of the city, state and country in a single query?
Note: I know this is possible with joins, however as there are more related tables, the nested joins makes the query hard to read, and I'm wondering if there is a cleaner way. It should also be possible for the person to have those fields empty.
Assuming the following tables:
create table People
(
ID int not null primary key auto_increment
,FullName varchar(255) not null
,StateID int
,CountryID int
,CityID int
)
;
create table States
(
ID int not null primary key auto_increment
,Name varchar(255) not null
)
;
create table Countries
(
ID int not null primary key auto_increment
,Name varchar(255) not null
)
;
create table Cities
(
ID int not null primary key auto_increment
,Name varchar(255) not null
)
;
With the Following Data:
insert into Cities(Name) values ('City 1'),('City 2'),('City 3');
insert into States(Name) values ('State 1'),('State 2'),('State 3');
insert into Countries(Name) values ('Country 1'),('Country 2'),('Country 3');
insert into People(FullName,CityID,StateID,CountryID) values ('Has Nothing' ,null,null,null);
insert into People(FullName,CityID,StateID,CountryID) values ('Has City' , 1,null,null);
insert into People(FullName,CityID,StateID,CountryID) values ('Has State' ,null, 2,null);
insert into People(FullName,CityID,StateID,CountryID) values ('Has Country' ,null,null, 3);
insert into People(FullName,CityID,StateID,CountryID) values ('Has Everything', 3, 2, 1);
Then this query should give you what you are after.
select
P.ID
,P.FullName
,Ci.Name as CityName
,St.Name as StateName
,Co.Name as CountryName
from People P
left Join Cities Ci on Ci.ID = P.CityID
left Join States St on St.ID = P.StateID
left Join Countries Co on Co.ID = P.CountryID
JOINS are the only way to really do this.
You might be able to change your schema, but the problem will be the same regardless.
(A City is always in a State, which is always in a Country - so the Person could just have a reference to the city_id rather than all three. You still need to join the 3 tables though).
There is no cleaner way than joins. If the fields are allowed to be empty, use outer joins
SELECT c.*, s.name AS state_name
FROM customer c
LEFT OUTER JOIN state s ON s.id = c.state
WHERE c.id = 10
According to the description of the schema that you have given you will have to use JOINS in a single query.
SELECT
p.first_name
, p.last_name
, c.name as city
, s.name as state
, co.name as country
FROM people p
LEFT OUTER JOIN city c
ON p.city_id = c.id
LEFT OUTER JOIN state s
ON p.state_id = s.id
LEFT OUTER JOIN country co
ON p.country_id = co.id;
The LEFT OUTER JOIN will allow you to fetch details of person even if some IDs are blank or empty.
Another way is to redesign your lookup tables. A city is always in a state and a state in a country. Hence your city table will have columns : Id, Name and state_id. Your state table will be : Id, Name and country_id. And country table will remain the same : Id and Name.
The person table will now have only 1 id : city_id
Now your query will be :
SELECT
p.first_name
, p.last_name
, c.name as city
, s.name as state
, co.name as country
FROM people p
LEFT OUTER JOIN city c
ON p.city_id = c.id
LEFT OUTER JOIN state s
ON c.state_id = s.id
LEFT OUTER JOIN country co
ON s.country_id = co.id;
Notice the difference in the last two OUTER JOINS
If the tables involved are reference tables (i.e. they hold lookup data that isn't going to change during the life time of a session), depending on the nature of your application, you could pre-load the reference data during you application start up. Then your query doesn't need to do the joins, instead it returns the id values, and in your application you do a decode of the ids when you need to display the data.
The easiest solution is to use the names as the primary keys in city, state, and country. Then your person table can reference them by the name instead of the pseudokey "id". That way, you don't need to do joins, since your person table already has the needed values.
It does take more space to store a string instead of a 4-byte pseudokey. But you may find the tradeoff worthwhile, if you are threatened by joins as much as you seem to be (which, by the way, is like a PHP programmer being reluctant to use foreach -- joins are fundamental to SQL in the same way).
Also there are many city names that appear in more than one state. So your city table should reference the state table and use these two columns as the primary key.
CREATE TABLE cities (
city_name VARCHAR(30),
state CHAR(2),
PRIMARY KEY (city_name, state),
FOREIGN KEY (state) REFERENCES states(state)
);
CREATE TABLE persons (
person_id SERIAL PRIMARY KEY,
...other columns...
city_name VARCHAR(30),
state CHAR(2),
country_name VARCHAR(30),
FOREIGN KEY (city_name, state) REFERENCES cities(city_name, state),
FOREIGN KEY (country_name) REFERENCES countries(country_name)
);
This just an example of the technique. Of course it's more complex than this, because you may have city names in more than one country, you may have countries with no states, and so on. The point is SQL doesn't force you to use integer pseudokeys, so use CHAR and VARCHAR keys where appropriate.
A disadvantage of standard SQL is the the return data needs to be in tabular format.
However some database vendors have added features that makes it possible to select data in non-tabular format. I don't know whether MySQL knows such features.
Create a view that does the Person, City, State, and Country joins for you. Then just reference the View in all other joins.
Something like:
CREATE VIEW FullPerson AS
SELECT Person.*, City.Name, State.Name, Country.Name
FROM
Person LEFT OUTER JOIN City ON Person.CityId = City.Id
LEFT OUTER JOIN State ON Person.StateId = State.Id
LEFT OUTER JOIN Country ON Person.CountryId = Country.Id
Then in other queries, you can
SELECT FullPerson.*, Other.Value
FROM FullPerson LEFT OUTER JOIN Other ON FullPerson.OtherId = Other.Id
All great answers but the questioner specified they didn't want to use joins. As one respondent demonstrated, assuming your Cities, States, and Countries tables have an Id and a Description field you might be able to do something like this:
SELECT
p.Name, c.Description, s.Description, ct.Description
FROM
People p, Cities c, States s, Countries ct
WHERE
p.Id = value AND
c.Id = value AND
s.Id = value AND
ct.Id = value;
Joins are the answer. With practise they will become more readable to you.
There may be special cases where creating a function would help you, for example you could do the following (in Oracle, I don't know any mysql):
You could create a function to return a formatted address given the city state and country codes, then your query becomes
SELECT first_name, last_name, formated_address(city_id, state_id, country_id)
FROM people
WHERE some_where_clause;
where formated_address does individual lookups on the city state and country tables and puts separators between the decoded values, or returns "no address" if they are all empty, etc