if statement mySQL - mysql

I am trying to figure out how to do an if statement that will essentially calculate the stock of an item based on the dates in the tp_rental table.
I was hoping to do something like :
if Date_Due has an entry but Date_Returned does not
then set stock -1
if Date_Due has an entry and Date_returned has an entry entered
then set stock +1
See below for my attempt, not sure if i need a join here?
update title_platform
set Stock = case
when tp_rental.Date_Returned is not null then stock -1
else stock +1
end
Am I going about this the right way?


You need an UPDATE statement with a join between the 2 tables and a CASE expression:
update title_platform p inner join tp_rental r
on r.platformid = p.platformid
set p.stock = p.stock + case
when r.date_due is not null and r.date_returned is null then -1
when r.date_due is not null and r.date_returned is not null then 1
else ? -- if there is another option
end
Remove the else part if it is not needed.
Edit: if there are multiple rows in tp_rental for each platformid then you must aggregate first and then join:
update title_platform p inner join (
select platformid,
sum(date_due is not null and date_returned is not null) -
sum(date_due is not null and date_returned is null) result
from tp_rental
group by platformid
) r on r.platformid = p.platformid
set p.stock = case
when p.stock + r.result < 0 then 0
else p.stock + r.result
end

Related

MySQL Multiple Case When Exists Statement

I have two tables. Let's call it: SEATS and SEAT_ALLOCATION_RULE table.
Below are the table schema:
CREATE TABLE IF NOT EXISTS `SEATS` (
`SeatID` int(11) NOT NULL AUTO_INCREMENT,
`SeatName` varchar(255) NOT NULL DEFAULT '',
PRIMARY KEY (`SeatID`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1 AUTO_INCREMENT=9 ;
INSERT INTO `SEATS` (`SeatID`, `SeatName`) VALUES
(1, 'Super VIP'),
(2, 'VIP'),
(3, 'Business'),
(4, 'Economy'),
(5, 'Standing');
CREATE TABLE IF NOT EXISTS `SEAT_ALLOCATION_RULE` (
`SeatID` int(11) NOT NULL DEFAULT '0',
`Origin` varchar(50) NOT NULL DEFAULT '0',
`Destination` varchar(50) NOT NULL DEFAULT '',
`Passenger_Type` varchar(25) NOT NULL DEFAULT '',
PRIMARY KEY (`SeatID`,`Origin`,`Destination`,`Passenger_Type`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
INSERT INTO `SEAT_ALLOCATION_RULE` (`SeatID`, `Origin`, `Destination, `Passenger_Type`) VALUES
(1, 'Malaysia','',''),
(2, 'Malaysia','Singapore',''),
(3, 'Malaysia','Singapore','Senior_Citizen'),
(4, 'Bangkok','Japan','Student'),
(5, 'Cambodia','China','Senior_Citizen');
SEAT_ALLOCATION_RULE table determines which seat should a passenger be assigned to based on the following order in priority:
1. Origin, destination, and passenger_type match
2. Origin and destination match
3. Origin match
It means that if all the fields (origin, destination, and passenger_type) match, it should take higher priority than if it is just two fields match and so on. If a column is empty, it is considered as unspecified and hence has lower priority. So, for example:
If the Origin is Malaysia, Destination is Singapore, and Passenger_Type is Senior_Citizen, it should return seatID 3
If the Origin is Malaysia, Destination is Singapore, and Passenger_Type is Student, it should return seatID 2 (since it only match Origin and Destination)
If the Origin is Malaysia, Destination is US, and Passenger_Type is Student, it should return seatID 1 (since it only match Origin).
Now, based on the rules above, if the origin is Malaysia, destination is Singapore, and Passenger_Type is student, the query to return seatID is as follow:
SELECT s.SeatID, s.SeatName
FROM SEATS s
WHERE
CASE WHEN EXISTS(
select 1
from SEAT_ALLOCATION_RULE r
where s.SeatID = r.SeatID
AND r.Origin = 'Malaysia'
AND r.Destination = 'Singapore'
AND r.Passenger_Type='Student') Then 1
WHEN EXISTS(
select 1
from SEAT_ALLOCATION_RULE r
where s.SeatID = r.SeatID
AND r.Origin = 'Malaysia'
AND r.Destination = 'Singapore'
AND r.Passenger_Type='') Then 1
WHEN EXISTS(
select 1
from SEAT_ALLOCATION_RULE r
where s.SeatID = r.SeatID
AND r.Origin = 'Malaysia'
AND r.Destination = ''
AND r.Passenger_Type='') Then 1 END
However, the query above does not work as it will return seatID 1 and 2, but the expected output is only seatID 2 (since origin and destination matches and it takes higher precedence). Can someone help to correct my SQL query?

This should do the trick:
select seatid
from seat_allocation_rule sar
order by ((sar.origin = :origin) << 2) + ((sar.destination = :destination) << 1) + (sar.passenger_type = :passenger_type) desc,
((sar.origin <> '') << 2) + ((sar.destination <> '') << 1) + (sar.passenger_type <> '') asc
limit 1
To understand how:
create table testcase (
origin varchar(255),
destination varchar(255),
passenger_type varchar(255),
expected_seat int(11)
);
insert into testcase values ('Malaysia','Singapore','Senior_Citizen',3),
('Malaysia','Singapore','Student',2),
('Malaysia','US','Student',1);
select * from (
select tc.*,
sar.seatid,
case when sar.seatid = tc.expected_seat then 'Y' else '-' end as pass,
((sar.origin = tc.origin) << 2)
+ ((sar.destination = tc.destination) << 1)
+ ((sar.passenger_type = tc.passenger_type) << 0) as score,
((sar.origin <> '') << 2)
+ ((sar.destination <> '') << 1)
+ ((sar.passenger_type <> '') << 0) as priority
from seat_allocation_rule sar
cross join testcase tc
) x order by expected_seat desc, score desc, priority asc;

This fixes the existing SQL:
SELECT DISTINCT s.SeatID, s.SeatName
FROM SEATS s
LEFT JOIN SEAT_ALLOCATION_RULE r ON r.SeatID = s.SeatID
AND r.Origin = 'Malaysia'
AND (
(r.Destination = 'Singapore' AND r.Passenger_Type IN ('Student', ''))
OR
(r.Destination = '' AND r.Passenger_Type = '')
)
WHERE r.SeatID IS NOT NULL
But it's only a partial solution, and it's hand-coding logic you really want to apply based solely on the data.
A complete solution will use hypothetical inputs for your passenger's ticket info to produce all eligible seats. This is a great use of lateral joins/apply, which are sadly lacking in MySql (all of their major competitors have had these for at least two release cycles, along with other gems that are absent from the current MySql release like windowing functions, ctes, full joins... I could go on). Here's how I'd do it in Sql Server:
SELECT p.PassengerID, s.SeatID, s.SeatName
FROM Passenger p
CROSS APPLY (
SELECT TOP 1 r.SeatID
FROM SEAT_ALLOCATION_RULE r
WHERE COALESCE(NULLIF(r.Origin, ''),p.Origin) = p.Origin
AND COALESCE(NULLIF(r.Destination,''), p.Destination) = p.Destination
AND COALESCE(NULLIF(r.Passenger_Type,''),p.Passenger_Type) = p.Passenger_Type
ORDER BY
CASE WHEN r.Origin <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Destination <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Passenger_Type <> '' THEN 1 ELSE 0 END DESC
) r
INNER JOIN SEATS s ON s.SeatID = r.SeatID
WHERE p.PassengerID = /* passenger criteria here */
I know the Sql Server solution isn't much immediate help to you, but perhaps it will suggest a better MySql solution.
Without APPLY, the only way I know to do this is to first compute the MAX() match count for your passengers (how many parts of the rules match):
SELECT p.PassengerID,
MAX(CASE WHEN r.Origin <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Destination <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Passenger_Type <> '' THEN 1 ELSE 0 END) AS MatchCount
FROM Passenger p
INNER JOIN SEAT_ALLOCATION_RULE r ON COALESCE(NULLIF(r.Origin, ''),p.Origin) = p.Origin
AND COALESCE(NULLIF(r.Destination,''), p.Destination) = p.Destination
AND COALESCE(NULLIF(r.Passenger_Type,''),p.Passenger_Type) = p.Passenger_Type
GROUP BY p.PassengerID
And then use that to filter down to results that have the same number of matches:
SELECT p
FROM Passenger p
INNER JOIN ( /* matchecounts */
SELECT p.PassengerID,
MAX(CASE WHEN r.Origin <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Destination <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Passenger_Type <> '' THEN 1 ELSE 0 END) AS MatchCount
FROM Passenger p
INNER JOIN SEAT_ALLOCATION_RULE r ON COALESCE(NULLIF(r.Origin, ''),p.Origin) = p.Origin
AND COALESCE(NULLIF(r.Destination,''), p.Destination) = p.Destination
AND COALESCE(NULLIF(r.Passenger_Type,''),p.Passenger_Type) = p.Passenger_Type
GROUP BY p.PassengerID
) m ON m.PassengerID = p.PassengerID
INNER JOIN SEAT_ALLOCATION_RULE r ON COALESCE(NULLIF(r.Origin, ''),p.Origin) = p.Origin
AND COALESCE(NULLIF(r.Destination,''), p.Destination) = p.Destination
AND COALESCE(NULLIF(r.Passenger_Type,''),p.Passenger_Type) = p.Passenger_Type
INNER JOIN SEATS s ON s.SeatID = r.SeatID
WHERE m.MatchCount =
(CASE WHEN r.Origin <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Destination <> '' THEN 1 ELSE 0 END
+ CASE WHEN r.Passenger_Type <> '' THEN 1 ELSE 0 END)
AND p.PassengerID = /* Passenger criteria here */
Which repeats a lot of code as well as effort in the DB, and is not very efficient. You can repeat the passenger criteria in the nested query, but that would only help a little. This option might also return multiple records for a passenger if they match two rules equally, though you can solve this easily enough with a GROUP BY expression.
In either case, note you can improve performance and simplify code by using actual NULL values instead of empty strings for missing parts of the SEAT_ALLOCATION_RULE table.

How to query last 2 business days only

I'm running a query that pulls the correct information I'm looking for, but I need it to pull the last 2 business days rather than the last 2 days. This comes into play when it's Monday and my results show information for Monday and Sunday rather than Monday and Friday. How can I change my query to pull in business days only?
USE [LetterGeneration]
SELECT g.LetterGenerationPrintJobId
,CAST(t.[TemplateKey] AS VarChar) AS LetterCode
,convert(char(12),r.CreatedDate,101) AS CreatedDate
,s.LetterGenerationStatusId AS Status
,s.StatusKey AS StatusDesc
,count(g.LetterGenerationId) as LetterCount
,c.BankingDateYorN
FROM [LetterGenerationTemplateRequest] AS r
INNER JOIN [LetterGenerationTemplate] AS t
ON t.[LetterGenerationTemplateId] = r.LetterGenerationTemplateId
INNER JOIN LetterGeneration g
ON g.LetterGenerationTemplateRequestId = r.LetterGenerationTemplateRequestId
INNER JOIN LetterGenerationStatus s
ON g.LetterGenerationStatusId = s.LetterGenerationStatusId
INNER JOIN Enterprise..Calendar C
ON c.BeginDate = g.LetterDate
WHERE ((DATEDIFF(d, r.CreatedDate, GETDATE()) = 0) OR (DATEDIFF(d, r.CreatedDate, GETDATE()) = 1))
--BankingDateYorN = 1
--AND RelativeTimeValue_BusinessDates =-1
AND t.[TemplateKey] NOT LIKE '%PLTV1%'
AND s.LetterGenerationStatusId NOT LIKE '4'
AND s.LetterGenerationStatusId NOT LIKE '16'
AND s.LetterGenerationStatusId NOT LIKE '19'
AND s.LetterGenerationStatusId NOT LIKE '20'
AND s.LetterGenerationStatusId NOT LIKE '38'
GROUP BY r.[LetterGenerationTemplateRequestId]
,r.LetterGenerationTemplateId
,g.Lettergenerationprintjobid
,t.[TemplateKey]
,r.[Loan_no]
,r.CreatedDate
,r.[CreatedBy]
,s.LetterGenerationStatusId
,s.StatusKey
,c.BankingDateYorN
ORDER BY r.CreatedDate DESC
UPDATE: I've recently discovered how to join a calendar table to my current query. The calendar query has a column called BusinessDayYorN with 1's for a business day and 0's for weekends and holidays. I've also updated the old query to now include the join.

select *
from LetterGenerationTemplateRequest
where createddate >= (
getdate() -
case datename(dw,getdate())
when 'Tuesday' then 5
when 'Monday' then 4
else 3
end
)
--and datename(dw,createdDate) not in ('Saturday','Sunday',datename(dw,getdate()))
and datename(dw,createdDate) not in ('Saturday','Sunday')
;

Assuming that you always want to include the last two non-weekend days you can try this:
; with aux as (
select diff = case
when datename(weekday, getdate()) in ('Tuesday', 'Wednesday ', 'Thursday', 'Friday') then 1
else
case datename(weekday, getdate())
when 'Saturday' then 2
when 'Sunday' then 3
when 'Monday' then 4
end
end
)
SELECT --r.[LetterGenerationTemplateRequestId]
--,r.LetterGenerationTemplateId
g.LetterGenerationPrintJobId
,CAST(t.[TemplateKey] AS VarChar) AS LetterCode
,r.[Loan_no]
,convert(char(12),r.CreatedDate,101) AS CreatedDate
-- ,g.ModifiedDate
-- ,convert(varchar(18), g.ModifiedDate - r.CreatedDate, 108) AS TimeSpan
,s.LetterGenerationStatusId AS Status
,s.StatusKey AS StatusDesc
,count(g.LetterGenerationId) as LetterCount
FROM [LetterGenerationTemplateRequest] AS r
INNER JOIN [LetterGenerationTemplate] AS t
ON t.[LetterGenerationTemplateId] = r.LetterGenerationTemplateId
INNER JOIN LetterGeneration g
ON g.LetterGenerationTemplateRequestId = r.LetterGenerationTemplateRequestId
INNER JOIN LetterGenerationStatus s
ON g.LetterGenerationStatusId = s.LetterGenerationStatusId
WHERE
DATEDIFF(day, r.CreatedDate, GETDATE()) <= (select diff from aux)
AND t.[TemplateKey] NOT LIKE '%PLTV1%'
AND s.LetterGenerationStatusId NOT LIKE '4'
AND s.LetterGenerationStatusId NOT LIKE '16'
AND s.LetterGenerationStatusId NOT LIKE '19'
AND s.LetterGenerationStatusId NOT LIKE '20'
AND s.LetterGenerationStatusId NOT LIKE '38'
GROUP BY r.[LetterGenerationTemplateRequestId]
,r.LetterGenerationTemplateId
,g.Lettergenerationprintjobid
,t.[TemplateKey]
,r.[Loan_no]
,r.CreatedDate
-- ,g.ModifiedDate
,r.[CreatedBy]
,s.LetterGenerationStatusId
,s.StatusKey
ORDER BY r.CreatedDate DESC
The CTE aux returns a dataset with only one record and only one field, the value of which is the number of days you need to go back in your WHERE statement.

SQL Server Row totals in pivot query

I am trying to make a row in the end of the result set that shows the totals.
My query is this:
SELECT
[ ] = ISNULL(CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente, 'TOTAL'),
[2016-11-01] = MAX([2016-11-01]),
[2016-11-02] = MAX([2016-11-02]),
[2016-11-03] = MAX([2016-11-03]),
[2016-11-04] = MAX([2016-11-04]),
TOTAL = COUNT([2016-11-01]) + COUNT([2016-11-02]) + COUNT([2016-11-03]) + COUNT([2016-11-04])
FROM
(
SELECT GEN_Paciente.GEN_idPaciente,COALESCE(GEN_ape_paternoPaciente, '')+' '+COALESCE(GEN_ape_maternoPaciente, '')+' '+COALESCE(GEN_nombrePaciente, '') AS nombrePaciente,HOS_fechaCategorizacion,HOS_nivel_riesgoCategorizacion+CAST(HOS_nivel_dependenciaCategorizacion AS VARCHAR) as riesgoDependencia
FROM HOS_Categorizacion
INNER JOIN HOS_Hospitalizacion
ON HOS_Hospitalizacion.HOS_idHospitalizacion = HOS_Categorizacion.HOS_idHospitalizacion
INNER JOIN GEN_Paciente
ON GEN_Paciente.GEN_idPaciente = HOS_Hospitalizacion.GEN_idPaciente
WHERE HOS_nivel_riesgoCategorizacion IS NOT NULL
) src
PIVOT
(
MAX(riesgoDependencia)
for HOS_fechaCategorizacion in ([2016-11-01],[2016-11-02],[2016-11-03],[2016-11-04])
) p
GROUP BY
ROLLUP(CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente)
This gives me this result:
But as you can see the totals for the rows are right but the totals for the columns are wrong because I am using MAX instead of COUNT, but I only need COUNT in the TOTAL row, the others have to be MAX, so I wrote this query:
SELECT
[ ] = ISNULL(CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente, 'TOTAL'),
[2016-11-01] = CASE WHEN CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente IS NOT NULL THEN MAX([2016-11-01]) ELSE COUNT([2016-11-01]) END,
[2016-11-02] = CASE WHEN CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente IS NOT NULL THEN MAX([2016-11-02]) ELSE COUNT([2016-11-02]) END,
[2016-11-03] = CASE WHEN CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente IS NOT NULL THEN MAX([2016-11-03]) ELSE COUNT([2016-11-03]) END,
[2016-11-04] = CASE WHEN CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente IS NOT NULL THEN MAX([2016-11-04]) ELSE COUNT([2016-11-04]) END,
TOTAL = COUNT([2016-11-01]) + COUNT([2016-11-02]) + COUNT([2016-11-03]) + COUNT([2016-11-04])
FROM
(
SELECT GEN_Paciente.GEN_idPaciente,COALESCE(GEN_ape_paternoPaciente, '')+' '+COALESCE(GEN_ape_maternoPaciente, '')+' '+COALESCE(GEN_nombrePaciente, '') AS nombrePaciente,HOS_fechaCategorizacion,HOS_nivel_riesgoCategorizacion+CAST(HOS_nivel_dependenciaCategorizacion AS VARCHAR) as riesgoDependencia
FROM HOS_Categorizacion
INNER JOIN HOS_Hospitalizacion
ON HOS_Hospitalizacion.HOS_idHospitalizacion = HOS_Categorizacion.HOS_idHospitalizacion
INNER JOIN GEN_Paciente
ON GEN_Paciente.GEN_idPaciente = HOS_Hospitalizacion.GEN_idPaciente
WHERE HOS_nivel_riesgoCategorizacion IS NOT NULL
) src
PIVOT
(
MAX(riesgoDependencia)
for HOS_fechaCategorizacion in ([2016-11-01],[2016-11-02],[2016-11-03],[2016-11-04])
) p
GROUP BY
ROLLUP(CAST(GEN_idPaciente AS VARCHAR)+'-'+nombrePaciente)
But that is not working
Thanks for your help!!

If I understand this correctly you want to count all columns which are not null. In this case you should just look at the condition IS NULL and not at the actual value at all. Try this:
DECLARE #tbl TABLE(ID INT IDENTITY, val1 VARCHAR(100),val2 VARCHAR(100),val3 VARCHAR(100));
INSERT INTO #tbl VALUES
('row1_val1','row1_val2',NULL)
,('row2_val1','row2_val2','row2_val3')
,(NULL,'row2_val2',NULL)
,(NULL,NULL,'row2_val3')
,(NULL,NULL,NULL);
SELECT *
,CASE WHEN val1 IS NULL THEN 0 ELSE 1 END
+CASE WHEN val2 IS NULL THEN 0 ELSE 1 END
+CASE WHEN val3 IS NULL THEN 0 ELSE 1 END AS CountOfValNotNull
FROM #tbl
UPDATE: Add a final Totals Row
You'd need ugly fiddling with a CTE, an additional sort column, UNION ALL to add another row and a sub_select.
Use the outer-most ORDER BY to get the artificial Totals-Row to the end
hint: Use the #tbl variable from above!
WITH SortedRows AS
(
SELECT ROW_NUMBER() OVER(ORDER BY ID) AS SortColumn
,*
,CASE WHEN val1 IS NULL THEN 0 ELSE 1 END
+CASE WHEN val2 IS NULL THEN 0 ELSE 1 END
+CASE WHEN val3 IS NULL THEN 0 ELSE 1 END AS CountOfValNotNull
FROM #tbl
)
SELECT tbl1.*
FROM
(
SELECT * FROM SortedRows
UNION ALL
SELECT 999999,0,'','','',(SELECT SUM(CountOfValNotNull) FROM SortedRows)
) AS tbl1
ORDER BY tbl1.SortColumn

MySQL Case in Update

I've browsed the forums. I've tested several different methods. I just can't get this to work.
GOAL:
Update the next column if an answer exists in the previous column
Q1 IS NULL at this point and I want it set to 1
Q2 IS NULL and I want it to stay NULL if Q1 IS NULL, If Q1 has an answer, update this column to 1
PROBLEM:
Both columns always update to 1
ATTEMPTS:
UPDATE Story1_Responses
SET Q1 = IFNULL(Q1,'1'),
Q2 = CASE WHEN Q1 IS NULL THEN NULL ELSE '1' END
WHERE UserID = 16
UPDATE Story1_Responses
SET Q1 = IFNULL(Q1,'1'),
Q2 = CASE WHEN IFNULL(Q1,'') = '' THEN NULL ELSE '1' END
WHERE UserID = 16
UPDATE Story1_Responses
SET Q1 = IFNULL(Q1,'1'),
Q2 = CASE WHEN COALESCE(Q1, '') = '' THEN NULL ELSE '1' END
WHERE UserID = 16
UPDATE Story1_Responses
SET Q1 = IFNULL(Q1,'1'),
Q2 = CASE WHEN Q1 IS NOT NULL THEN '1' ELSE IFNULL(Q2,NULL) END
WHERE UserID = 16
THOUGHTS:
It seems to me that MySQL updates the first column prior to the second column being evaluated even though they are in the same update statement execution.

Order is important. First set Q2 (because it checks the previous value of Q1), then set Q1.
UPDATE Story1_Responses
SET Q2 = CASE WHEN Q1 IS NULL THEN NULL ELSE '1' END,
Q1 = IFNULL(Q1,'1')
WHERE UserID = 16;
SQLFiddle

MySQL - join same table twice

I have a table called 'booking'
> id status
1 P
2 P
and another called 'call'
id calldate type booking
1 01/01/2012 DEL 1
2 01/02/2012 COL 1
3 01/03/2012 DEL 2
4 31/12/2019 COL 999
I want to list each record in 'bookings' ONCE, showing associated records from 'call' as another column like this:
bookingId deliverydate collectiondate
1 01/01/2012 01/02/2012
2 01/03/2012 null
I've tried:
select `b`.`bookingid` AS `bookingid`,
`del`.`calldate` AS `Delivery`,
`col`.`calldate` AS `Collection`
from `booking` `b`
left join `call` `del` on `b`.`bookingid` = `del`.`booking`
left join `call` `col` on `b`.`bookingid` = `col`.`booking`
where ((`del`.`type` = 'DEL') OR (`col`.`type` = 'COL') and (`b`.`status` = 'P'));
but I get bookingid 1 listed 3 times. Can someone please fix my joins?

I think you want to move your types into the join conditions:
select `b`.`bookingid` AS `bookingid`,
`del`.`calldate` AS `Delivery`,
`col`.`calldate` AS `Collection`
from `booking` `b`
left join `call` `del` on `b`.`bookingid` = `del`.`booking` AND `del`.`type` = 'DEL'
left join `call` `col` on `b`.`bookingid` = `col`.`booking` AND `col`.`type` = 'COL'
where `b`.`status` = 'P';

You don't need to join the same table twice, you could use this:
SELECT
`call`.booking,
max(case when type='DEL' then calldate end) as deliverydate,
max(case when type='COL' then calldate end) as collectiondate
FROM
booking inner join `call`
on booking.id=`call`.booking
WHERE
booking.status='P'
GROUP BY `call`.booking