Change MySQL query from WHERE to LEFT JOIN - mysql

I have a query which reports on Asterisk call usage and queue statistics. The query currently uses MySQL WHERE clauses to join the tables and filter the data.
This is how the query looks at present:
SELECT
c.name as 'Agent',
e.extended_number AS 'Extension',
COUNT(ql.`time`) AS 'Total Inbound Calls',
(SELECT
COUNT(1)
FROM
call_history ch,
ast_queue_mstatus qm
WHERE
ch.start >= (DATE_SUB(DATE(NOW()),
INTERVAL 10050 MINUTE))
AND ch.start <= (DATE(NOW()))
AND ch.calltype = 'out'
AND ch.flow = 'out'
AND ch.extension_number = qm.membername
AND qm.membername = e.extended_number
GROUP BY qm.membername) AS 'Total Outbound Calls',
FORMAT(SUM(CAST(ql.arg2 AS UNSIGNED)) / 60, 2) AS 'Total Inbound Duration',
FORMAT(AVG(CAST(ql.arg2 AS UNSIGNED)) / 60, 2) AS 'Avg Inbound Duration',
FORMAT(AVG(CAST(ql.arg1 AS UNSIGNED)) / 60, 2) AS 'Avg Caller Hold Time'
FROM
ast_queue_log ql,
ast_queue_mstatus qs,
client c,
extension e
WHERE
ql.queuename = '1234'
AND ql.`time` >= (DATE_SUB(DATE(NOW()), INTERVAL 7 DAY))
AND ql.`time` <= (DATE(NOW()))
AND (ql.event = 'COMPLETEAGENT'
OR ql.event = 'COMPLETECALLER'
OR ql.event = 'COMPLETETRANSFER')
AND RIGHT(ql.agent, 8) = qs.membername
AND qs.membername = e.extended_number
AND e.client_id = c.id
GROUP BY ql.agent
ORDER BY c.name;
The problem that I have is that I need to return outbound call results for queue members that have not received any inbound calls. At the moment if there are no records in ast_queue_log then the WHERE clause excludes the records, which means they do not get picked up by the scalar sub query.
How can I change this query so I get all call records from call_history where there is a corresponding record in ast_queue_mstatus?
Sorry if this is too complex or there is not enough info, I can try to simplify if this makes no sense!

It's time to ditch those old comma separated list of tables and use the join standard that was introduced in 1992 (yes). :-)
Because then you can use a LEFT JOIN. With this kind of join, the relationship is optional. That is, all rows in the first table are returned (if they match the conditions in the WHERE clause), but the joined table doesn't have to have records as well.
If it doesn't, a row is still returned, but the fields are filled with NULL value.
SELECT
c.name as 'Agent',
e.extended_number AS 'Extension',
COUNT(ql.`time`) AS 'Total Inbound Calls',
(SELECT
COUNT(1)
FROM
call_history ch
INNER JOIN ast_queue_mstatus qm
ON qm.membername = ch.extension_number
WHERE
ch.start >= (DATE_SUB(DATE(NOW()), INTERVAL 10050 MINUTE))
AND ch.start <= (DATE(NOW()))
AND ch.calltype = 'out'
AND ch.flow = 'out'
AND ch.extension_number = e.extended_number
GROUP BY
ch.extension_number) AS 'Total Outbound Calls',
FORMAT(SUM(CAST(ql.arg2 AS UNSIGNED)) / 60, 2) AS 'Total Inbound Duration',
FORMAT(AVG(CAST(ql.arg2 AS UNSIGNED)) / 60, 2) AS 'Avg Inbound Duration',
FORMAT(AVG(CAST(ql.arg1 AS UNSIGNED)) / 60, 2) AS 'Avg Caller Hold Time'
FROM
client c,
INNER JOIN extension e
ON e.client_id = c.id
INNER JOIN ast_queue_mstatus qs
ON qs.membername = e.extended_number
LEFT JOIN ast_queue_log ql
ON ql.queuename = '1234'
AND ql.`time` >= (DATE_SUB(DATE(NOW()), INTERVAL 7 DAY))
AND ql.`time` <= (DATE(NOW()))
AND (
ql.event = 'COMPLETEAGENT'
OR ql.event = 'COMPLETECALLER'
OR ql.event = 'COMPLETETRANSFER')
AND RIGHT(ql.agent, 8) = qs.membername
GROUP BY ql.agent
ORDER BY c.name;
You can use RIGHT JOIN as well to join the other way around (first table is optional). But in general this is harder to interpret.

Related

SQL how to create aggregate of two aggregated columns for each row

I'm working on a big query in SQL (MySQL 5.7) to calculate aggregated columns based on raw values in my table. I've created several aggregated columns (see attached screenshot and SQL) and I now need to create a conversion_percent column for each tlp_aff_id in my query.
This conversion_percent should be a division of the aggregated JoinedSessions.total_sessions and COUNT(Report.tlp_aff_id) as leads_total.
My current SQL:
SELECT
# Application details
Report.tlp_aff_id,
# Revenue
JoinedRevenue.total_revenue,
# Commission
JoinedCommission.total_commission,
# Profit
JoinedProfit.total_profit,
# Sessions
JoinedSessions.total_sessions,
# Submits
COUNT(Report.tlp_aff_id) as total_submits,
# Leads
COUNT(Report.tlp_aff_id) as leads_total,
SUM(case when Report.application_result = 'Accepted' then 1 else 0 end) leads_accepted,
SUM(case when Report.application_result = 'Rejected' then 1 else 0 end) leads_rejected
# Conversion percent
# JoinedConversion.conversion_percent
FROM
tlp_payout_report_minute AS Report
INNER JOIN
(
SELECT
JoinedRevenue.tlp_aff_id,
JoinedRevenue.minute_rounded_timestamp,
SUM(commission) AS total_revenue
FROM
tlp_payout_report_minute AS JoinedRevenue
WHERE
JoinedRevenue.minute_rounded_timestamp >= 1664841600
AND
JoinedRevenue.minute_rounded_timestamp <= 1664927999
GROUP BY
JoinedRevenue.tlp_aff_id
) AS JoinedRevenue
ON JoinedRevenue.tlp_aff_id = Report.tlp_aff_id
INNER JOIN
(
SELECT
ReportCommission.tlp_aff_id,
ReportCommission.seller_code,
ReportCommission.minute_rounded_timestamp,
SUM(commission) AS total_commission
FROM
tlp_payout_report_minute AS ReportCommission
WHERE
ReportCommission.minute_rounded_timestamp >= 1664841600
AND
ReportCommission.minute_rounded_timestamp <= 1664927999
AND
ReportCommission.seller_code != 44
GROUP BY
ReportCommission.tlp_aff_id
) AS JoinedCommission
ON JoinedCommission.tlp_aff_id = Report.tlp_aff_id
INNER JOIN
(
SELECT
ReportProfit.tlp_aff_id,
ReportProfit.seller_code,
ReportProfit.application_result,
ReportProfit.minute_rounded_timestamp,
SUM(commission) AS total_profit
FROM
tlp_payout_report_minute AS ReportProfit
WHERE
ReportProfit.minute_rounded_timestamp >= 1664841600
AND
ReportProfit.minute_rounded_timestamp <= 1664927999
AND
ReportProfit.application_result = 'Accepted'
AND
ReportProfit.seller_code = 44
GROUP BY
ReportProfit.tlp_aff_id
) AS JoinedProfit
ON JoinedProfit.tlp_aff_id = Report.tlp_aff_id
INNER JOIN
(
SELECT
Conversion.aff_id,
Conversion.conversion_type,
COUNT(Conversion.ip_address) as total_sessions
FROM
tlp_conversions AS Conversion
WHERE
Conversion.conversion_time >= '2022-10-04 00:00:00'
AND
Conversion.conversion_time <= '2022-10-04 23:59:59'
AND
Conversion.aff_id IS NOT NULL
AND
Conversion.conversion_type = 2
GROUP BY
Conversion.aff_id
) AS JoinedSessions
ON JoinedSessions.aff_id = Report.tlp_aff_id
WHERE
Report.minute_rounded_timestamp >= 1664841600
AND
Report.minute_rounded_timestamp <= 1664927999
GROUP BY
Report.tlp_aff_id
ORDER BY
JoinedRevenue.total_revenue DESC
I'm thinking something along the lines of:
INNER JOIN
(
...
) AS JoinedConversion
ON JoinedConversion.aff_id = Report.tlp_aff_id
But I don't think this is necessary for conversion_percent.
What's the right approach here?

Case When Exists In Subquery

I am trying to check whether an ID is present in a subquery. I have ran just the subquery and it produces a list of all the ID's which have a fee against it, so what I want to do is check whether the ID in the main query is present in the subquery. If it's present then return 1, else return 0.
This is an easy query but I have no idea where i'm going wrong, I tried using exists rather than in but this does not work either.
case when debtor._rowid in (
select distinct note.debtorid from note
left join debtor on note.debtorid = debtor._rowid
left join fee on fee.debtorid = debtor._rowid
where fee.type = "Enforcement" and note.type = "Stage")
then 1 else 0 end) as `Enforcement`
Below is the entire code, when I remove the above code from the main query below, it works perfectly, so there's something wrong in my case statement.
with cte_1
as
(
select
debtor._rowid as casref
,concat(date_format(date_sub(debtor._createddate, interval 3 month), '%y'), '/', date_format(date_add(debtor._createddate, interval 9 month), '%y')) as `F/Y`
,date_format(debtor._createddate, '%M %Y') as `Loaded Month`
,ifnull(concat(date_format(date_sub(debtor.offence_date, interval 3 month), '%y'), '/', date_format(date_add(debtor.offence_date, interval 9 month), '%y')),'-') as `LO F/Y`
,coalesce(date_format(debtor.offence_date,'%M %Y'),'-') as `Liability Order Month`
,scheme.name as `Scheme`
,branch.name as `Branch`
,count(debtor._rowid) as `Cases Received`
,count(debtor.linkid) as `LinkID`
,(case
when concat(date_format(date_sub(debtor._createddate, interval 3 month), '%y'), '/', date_format(date_add(debtor._createddate, interval 9 month), '%y'))
= ifnull(concat(date_format(date_sub(debtor.offence_date, interval 3 month), '%y'), '/', date_format(date_add(debtor.offence_date, interval 9 month), '%y')),'-')
then 1 else 0 end ) as `Same Year`
, case when debtor._rowid in (
select distinct note.debtorid from note
left join debtor on note.debtorid = debtor._rowid
left join fee on fee.debtorid = debtor._rowid
where fee.type = "Enforcement"
and note.type = "Stage")
then 1 else 0 end) as `Enforcement`
from debtor
left join clientscheme on debtor.clientschemeID = clientscheme._rowid
left join scheme on clientscheme.schemeID = scheme._rowid
left join branch on clientscheme.branchID = branch._rowid
left join fee on debtor._rowid = fee.debtorid
left join note on debtor._rowid = note.debtorid
where clientscheme.branchID in (1,10,24)
and debtor._createddate >= '2017-04-01'
group by debtor._rowid
)
,
cte_2
as
(
select
`F/Y`
,`Loaded Month`
,`LO F/Y`
,`Liability Order Month`
,`Scheme`
,`Branch`
,sum(`Cases Received`) as `Case Count`
,sum(`LinkID`) as `Linked Accounts`
,sum(`Same Year`) as `In Year LO`
,sum(Enforcement) as `Enforcement Applied`
from cte_1
group by
`Loaded Month`
,`Liability Order Month`
,`Scheme`
, `Branch`
)
select
`F/Y`
,`Loaded Month`
,`LO F/Y`
,`Liability Order Month`
,`Scheme`
,`Branch`
,`Case Count`
,`Linked Accounts`
,round((`Linked Accounts`/`Case Count`),2) * 100 as `% of Linked Accounts`
,round((`In Year LO`/`Case Count`),2) * 100 as `In Year LO's`
,`Enforcement Applied`
from cte_2
It appears that you want to logically check if, for a given record in the result set, a _noteid value from the debtor table matches to a debtors from the note table. You could rephrase your query as follows:
SELECT
(d._rowid = n.debtorid) AS `Enforcement Allocated`
FROM note n
LEFT JOIN debtor d
ON n.debtorid = d._rowid
LEFT JOIN fee f
ON f.debtorid = d._rowid
WHERE
f.type = 'Enforcement' AND n.type = 'Stage';
Note that since the output of your CASE expression is just 1 or 0, you may take advantage of that MySQL allows boolean expressions as values.

Combining conditions for CASE, mysql query

I'm trying to finish up a query to total different numbers for our people on the phones and I have the standard totals calculated fine, it seems, but I'm now trying to do this:
For every total column (total calls, total inbound, total outbound, total missed) I now need to have a new column partnered with it for known numbers in our database (so I'll have total calls known, total inbound known, etc.).
The way I'm doing this is by checking two fields in my session table: callingpartyno and finallycalledpartyno. So, for every call in session I need to check to see if the number in either of those fields are in my knownNumbers table in the phone_number field. If so, I need to count them for the known column.
About 8 lines down in my query below, I have my first instance of this which seems to be showing accurate numbers, but it's only incorporating callingpartyno. I first need to know how to combine conditions in this 'case whento addfinallycalledpartyno` as well.
The other issue is that for inbound and outbound, I need to base the phone number field I'm using on the LEGTYPE field. So, for example, IF I'm calculating outbound calls to known numbers I need something like
sum(if(LEGTYPE1 = 1,1,0)) AND finallycalledpartno = k.phone_number AS Total_Outbound_known.
I hope this makes sense, and it should be pretty straightforward for an advanced SQL programmer. I just can't seem to get the combination of conditions in a case when statement to work especially.
Query below:
SELECT u.firstn
,u.lastn
,c.extension
,SUM(IF(LEGTYPE1 = 1, 1, 0)) + -- outbound calls
SUM(IF(LEGTYPE1 = 2, 1, 0) AND ANSWERED = 1) + -- inbound calls
SUM(IF(Answered = 1, 0, 1)) AS Total_Calls
,sum(case when CALLINGPARTYNO = k.phone_number then 1 else 0 end ) AS total_known
,sum(if(Answered = 1,0,1)) AS Total_Missed
,sum(if(LEGTYPE1 = 2,1,0) and ANSWERED = 1) AS Total_Recieved
,sum(if(LEGTYPE1 = 1,1,0)) AS Total_Outbound
,round(sum(Duration) / 60,2) AS Total_Talk_Time_minutes
,sum(if(Answered = 1,0,1)) / (SUM(IF(LEGTYPE1 = 1, 1, 0)) + -- outbound calls
SUM(IF(LEGTYPE1 = 2, 1, 0) AND ANSWERED = 1) + -- inbound calls
SUM(IF(Answered = 1, 0, 1))) * 100 AS Percentage_Missed
FROM ambition.session a
INNER JOIN ambition.callsummary b
ON a.NOTABLECALLID = b.NOTABLECALLID
INNER join ambition.mxuser c
ON a.RESPONSIBLEUSEREXTENSIONID = c.EXTENSIONID
INNER join jackson_id.users u
on c.extension = u.extension
left join ambition.knownnumbers k
on a.callingpartyno = k.phone_number
WHERE b.ts between curdate() - interval 5 day and now()
AND c.extension IN (7276,7314,7295,7306,7357,7200,7218,7247,7331,7255,7330,7000,7215,7240,7358,7312)
group by c.extension;
Apart what I wrote in above comments, I would rewrite your query like this (I think at least it should be easier to read it, avoiding to repeat some SUMs)
SELECT firstn
, lastn
, extension
, Total_Outbound+Total_Missed+Total_Received AS Total_Calls
, Total_Known
, Total_Missed
, Total_Received
, Total_Outbound
, Total_Talk_Time_minutes
, Total_Missed / (Total_Outbound+Total_Missed+Total_Received) * 100 AS Percentage_Missed
FROM (
SELECT u.firstn
, u.lastn
, c.extension
, sum(case when CALLINGPARTYNO = k.phone_number then 1 else 0 end ) AS Total_Known
, sum(if(Answered = 1,0,1)) AS Total_Missed
, sum(CASE WHEN LEGTYPE1 = 2 AND ANSWERED = 1 THEN 1 ELSE 0 END) AS Total_Received
, sum(if(LEGTYPE1 = 1,1,0)) AS Total_Outbound
, round(sum(Duration) / 60,2) AS Total_Talk_Time_minutes
FROM ambition.session a
INNER JOIN ambition.callsummary b ON a.NOTABLECALLID = b.NOTABLECALLID
INNER join ambition.mxuser c ON a.RESPONSIBLEUSEREXTENSIONID = c.EXTENSIONID
INNER join jackson_id.users u ON c.extension = u.extension
LEFT JOIN ambition.knownnumbers k ON a.callingpartyno = k.phone_number
WHERE b.ts between curdate() - interval 5 day and now()
AND c.extension IN (7276,7314,7295,7306,7357,7200,7218,7247,7331,7255,7330,7000,7215,7240,7358,7312)
GROUP BY c.extension, u.firstn, u.lastn
) X;

Query using group by month

I have problem with the query using group by month. This query returns total_revenue per month. but if month of year doesn't contain any data then total_revnue is increased unnecessarily.
SELECT COUNT(CT.cumTxnReportId),
CT.cumTxnReportId,
CT.ticketNum,
DATE_FORMAT(CT.exitDateTimeUtc,'%m-%Y'),
sum(netAmount) AS total_revenue,
D.name,
HOUR(CT.entranceDateTimeUtc) AS entryHour,
HOUR(CT.exitDateTimeUtc) AS exitHour,
CT.entranceDateTimeUtc,
CT.exitDateTimeUtc,
CT.netAmount AS netAmount,
CT.grossAmount,
CT.discountAmount,
CT.rate,
CT.txnType,
CT.ticketType,
CT.txnNum,
CT.numDiscounts
FROM Parkloco.ParkingArea PA
JOIN IParcPro.Device D ON PA.id = D.parkingAreaId
JOIN Parkloco.RateCard RC ON PA.id = RC.parkingAreaId
JOIN IParcPro.CumTxn CT ON D.id = CT.deviceId
WHERE PA.uuid = '27d842c1-7057-11e6-a0eb-1245b0d35d23'
AND (CT.txnType = 'Allowed'
OR CT.txnType = 'Add'
OR CT.txnType = 'Normal'
OR CT.txnType = 'Offline'
OR CT.txnType = 'Repay')
AND ((CT.entranceDateTimeUtc >= '2016-08-01 00:00:00'
AND CT.exitDateTimeUtc <= '2017-04-31 23:59:59'))
AND (RC.state = 'active'
OR RC.state = 'archived')
AND RC.fromDateTimeUtc <= '2017-04-31 23:59:59'
AND (RC.thruDateTimeUtc IS NULL
OR RC.thruDateTimeUtc >= '2016-08-01 00:00:00')
AND (TIMESTAMPDIFF (SECOND, CT.entranceDateTimeUtc, CT.exitDateTimeUtc) >= '0' * 60)
AND (TIMESTAMPDIFF (SECOND, CT.entranceDateTimeUtc, CT.exitDateTimeUtc) < '1441' * 60)
AND CT.numDiscounts=0
AND CT.ticketNum !=0
GROUP BY DATE_FORMAT(CT.exitDateTimeUtc,'%m-%Y')
but when I am increasing the range month - at that point of time I am getting unneccessary increment in total_revenue
SELECT COUNT(CT.cumTxnReportId),
CT.cumTxnReportId,
CT.ticketNum,
DATE_FORMAT(CT.exitDateTimeUtc,'%m-%Y'),
sum(netAmount) AS total_revenue,
D.name,
HOUR(CT.entranceDateTimeUtc) AS entryHour,
HOUR(CT.exitDateTimeUtc) AS exitHour,
CT.entranceDateTimeUtc,
CT.exitDateTimeUtc,
CT.netAmount AS netAmount,
CT.grossAmount,
CT.discountAmount,
CT.rate,
CT.txnType,
CT.ticketType,
CT.txnNum,
CT.numDiscounts
FROM Parkloco.ParkingArea PA
JOIN IParcPro.Device D ON PA.id = D.parkingAreaId
JOIN Parkloco.RateCard RC ON PA.id = RC.parkingAreaId
JOIN IParcPro.CumTxn CT ON D.id = CT.deviceId
WHERE PA.uuid = '27d842c1-7057-11e6-a0eb-1245b0d35d23'
AND (CT.txnType = 'Allowed'
OR CT.txnType = 'Add'
OR CT.txnType = 'Normal'
OR CT.txnType = 'Offline'
OR CT.txnType = 'Repay')
AND ((CT.entranceDateTimeUtc >= '2016-08-01 00:00:00'
AND CT.exitDateTimeUtc <= '2017-07-31 23:59:59'))
AND (RC.state = 'active'
OR RC.state = 'archived')
AND RC.fromDateTimeUtc <= '2017-07-31 23:59:59'
AND (RC.thruDateTimeUtc IS NULL
OR RC.thruDateTimeUtc >= '2016-08-01 00:00:00')
AND (TIMESTAMPDIFF (SECOND, CT.entranceDateTimeUtc, CT.exitDateTimeUtc) >= '0' * 60)
AND (TIMESTAMPDIFF (SECOND, CT.entranceDateTimeUtc, CT.exitDateTimeUtc) < '1441' * 60)
AND CT.numDiscounts=0
AND CT.ticketNum !=0
GROUP BY DATE_FORMAT(CT.exitDateTimeUtc,'%m-%Y')
output such as :
can anyone help me on this? Thanks in advance if you could let me know.
Despite MySQL allow this weird group by rules, in my opinion, you should to avoid use it. I explain, usually, all select clause non aggregate fields should appear on group by clause:
select a,b,c, sum(z)
from t
group by a,b,c
vs
select a,b,c, sum(z)
from t
group by a #<--- MySQL allow this!
Then, if b and c are not in group by, how MySQL figure up the right fields to be selected? Like this on <5.6:
The server is free to choose any value from each group, so unless they are the same, the values chosen are indeterminate.
In my opinion, in your query has no sense: Look entryHour and total_revenue. One is for an entry the other one is for all month.
I guess you should to rethink the hole sql statement. Because the result of this one is incoherent.
Also, remember this is not 'code revision service'. Please, read how to create a Minimal, Complete, and Verifiable example in order your question also help other users.

how to join after left join complex mysql queries

I have this query
SELECT
currency_code,
SUM(CASE WHEN TYPE = 'buy'THEN to_amount END ) AS BUY,
SUM(CASE WHEN TYPE = 'sell' THEN to_amount END ) AS SELL,
SUM(CASE WHEN TYPE = 'sell' THEN rate END ) AS SELL_RATE,
SUM(CASE WHEN TYPE = 'buy' THEN rate END ) AS BUY_RATE,
AVG(CASE WHEN TYPE = 'buy' THEN rate END ) AS AVG_BUY_RATE,
AVG(CASE WHEN TYPE = 'sell' THEN rate END ) AS AVG_SELL_RATE
FROM tb_currency
LEFT JOIN tb_bill
ON tb_currency.CURRENCY_ID = tb_bill.CURRENCY_ID
AND tb_bill.TYPE IN ('buy', 'sell')
AND date( DATE_TIME ) >= '2011-01-01'
AND date( DATE_TIME ) <= '2011-01-11'
GROUP BY currency_code
that will output this:
Right now i want to join this query with another table called tb_user
the tb_user have PK called user_id and the tb_bill that is use in the query above also have foreign key called user_id
tb_user
user_id (pk)| user_name | branch_id
tb_bill
bill_id (pk) | user_id (fk)|
Desired result should be the above picture plus one column branch_id.
If it doesnt have branch_id, return null.
I tried several times but still cant join it correctly. Hope you guys can help.
Thanks.
The three conditions in the join (the AND clauses) might be giving you trouble. Those three conditions are selection criteria, not join criteria.
Also, your use of CASE looks odd to me. I'm sure it works, but IF might be better suited for a one-condition function. In the below, if the fields are floating point rather than integer then replace the 0 with 0.0.
SELECT currency_code,
SUM(IF(TYPE = 'buy', to_amount, 0)) AS BUY,
SUM(IF(TYPE = 'sell', to_amount, 0)) AS SELL,
SUM(IF(TYPE = 'sell', rate, 0)) AS SELL_RATE,
SUM(IF(TYPE = 'buy', rate, 0)) AS BUY_RATE,
AVG(IF(TYPE = 'buy', rate, 0)) AS AVG_BUY_RATE,
AVG(IF(TYPE = 'sell', rate, 0)) AS AVG_SELL_RATE,
tb_user.whatever_field,
tb_user.whatever_other_field
FROM tb_currency
LEFT JOIN tb_bill ON tb_currency.CURRENCY_ID = tb_bill.CURRENCY_ID
LEFT JOIN tb_user ON tb_bill.user_id = tb_user.user_id
WHERE tb_bill.TYPE IN ('buy', 'sell')
AND date( DATE_TIME ) >= '2011-01-01'
AND date( DATE_TIME ) <= '2011-01-11'
GROUP BY currency_code, tb_user.user_id
Finally, all-cap field names look odd to my eye as well. Whatever works for you though.
add user_id to SELECT part
after
LEFT JOIN tb_bill ON tb_currency.CURRENCY_ID = tb_bill.CURRENCY_ID
place
LEFT JOIN tb_user ON tb_user.id = tb_bill.user_id
also you missing WHERE ( put instead first AND )
and
GROUP BY currency_code, user_id