I Have 2 tables in my DB and I want to compare values of 2 select queries Ive made on each one
Table 1: click_log
Query table 1:
SELECT *
FROM click_log
Table 2: km_articles
Query table 2:
SELECT km_article_no
FROM km_articles
WHERE km_article_date <= "2017-10-31" AND km_article_status = "Published" AND km_article_view_count <= "5"
The columns I want to compare are table link_clicked for table 1 with km_article_no and I know I will find repeated matched, nevertheless from those repeated matches I want to find the latest one that I want to get from another column in table 1 called "when_clicked" that contains data information, not sure How can i put together those to queries and then narrow them down.
this is how the tables look like:
Table 1:
|link_clicked|when_clicked
KB00001 | 2017-08-02
KB00001 | 2017-12-02
KB00002 | 2017-08-02
KB00002 | 2017-09-02
KB00003 | 2017-09-02
KB00003 | 2017-09-02
Table 2:
km_article_no|km_article_ti|km_article_status|km_article_view_count|km_article_date
KB00001 |outlook IOS | Published | 5 | 2017-01-02
KB00002 |outlook CSS | Published | 4 | 2017-01-05
KB00003 |outlook ZTE | Retired | 3 | 2017-01-09
If I understand correctly, you want to show all km_articlesrows, each with the latest related click_log.when_clicked date. So aggregate your click_log per link_clicked and find the maximum when_clicked. Then join this to km_articles.
select kma.*, cl.last_clicked
from km_articles kma
join
(
select link_clicked, max(when_clicked) as last_clicked
from click_log
group by link_clicked
) cl on cl.link_clicked = kma.km_article_no
where kma.km_article_date <= date '2017-10-31'
and kma.km_article_status = 'Published'
and kma.km_article_view_count <= 5;
(If you also want to show km_articles rows that have no match in click_log, then change join to left join.)
Related
Please note that I'm an absolute n00b in MySQL but somehow I managed to build some (for me) complex queries that work as they should. My main problem now is that for a many of the queries we're working on:
The querie is becoming too big and very hard to see through.
The same subqueries get repeated many times and that is adding to the complexity (and probably to the time needed to process the query).
We want to further expand this query but we are reaching a point where we can no longer oversee what we are doing. I've added one of these subqueries at the end of this post, just as an example.
!! You can fast foward to the Problem section if you want to skip the details below. I think the question can be answered also without the additional info.
What we want to do
Create a MySQL query that calculates purchase orders and forecasts for a given supplier based on:
Sales history in a given period (past [x] months = interval)
Current stock
Items already in backorder (from supplier)
Reserved items (for customers)
Supplier ID
I've added an example of a subquery at the bottom of this message. We're showing just this part to keep things simple for now. The output of the subquery is:
Part number
Units sold
Units sold (outliers removed)
Units sold per month (outliers removed)
Number of invoices with the part number in the period (interval)
It works quite OK for us, although I'm sure it can be optimised. It removes outliers from the sales history (e.g. one customer that orders 50 pcs of one product in one order). Unfortunately it can only remove outliers with substantial data, so if the first order happens to be 50 pcs then it is not considered an outlier. For that reason we take the amount of invoices into account in the main query. The amount of invoices has to exceed a certain number otherwise the system wil revert to a fixed value of "maximum stock" for that product.
As mentioned this is only a small part of the complete query and we want to expand it even further (so that it takes into account the "sales history" of parts that where used in assembled products).
For example if we were to build and sell cars, and we want to place an
order with our tyre supplier, the query calculates the amount of tyres we need to order based on the sales history of the various car models (while also taking into account the stock of the cars, reserved cars and stock of the tyres).
Problem
The query is becomming massive and incomprehensible. We are repeating the same subqueries many times which to us seems highly inefficient and it is the main cause why the query is becomming so bulky.
What we have tried
(Please note that we are on MySQL 5.5.33. We will update our server soon but for now we are limited to this version.)
Create a VIEW from the subqueries.
The main issue here is that we can't execute the view with parameters like supplier_id and interval period. Our subquery calculates the sum of the sold items for a given supplier within the given period. So even if we would build the VIEW so that it calculates this for ALL products from ALL suppliers we would still have the issue that we can't define the interval period after the VIEW has been executed.
A stored procedure.
Correct me if I'm wrong but as far as I know, MySQL only allows us to perform a Call on a stored procedure so we still can't run it against the parameters (period, supplier id...)
Even this workaround won't help us because we still can't run the SP against the parameters.
Using WITH at the beginning of the query
A common table expression in MySQL is a temporary result whose scope is confined to a single statement. You can refer this expression multiple times with in the statement.
The WITH clause in MySQL is used to specify a Common Table Expression, a with clause can have one or more comms-separated subclauses.
Not sure if this would be the solution because we can't test it. WITH is not supported untill MySQL version 8.0.
What now?
My last resort would be to put the mentioned subqueries in a temp table before starting the main query. This might not completely eliminate our problems but at least the main query will be more comprehensible and with less repetition of fetching the same data. Would this be our best option or have I overlooked a more efficient way to tackle this?
Many thanks for your kind replies.
SELECT
GREATEST((verkocht_sd/6*((100 + 0)/100)),0) as 'units sold p/month ',
GREATEST(ROUND((((verkocht_sd/6)*3)-voorraad+reserved-backorder),0),0) as 'Order based on units sold',
SUM(b.aantal) as 'Units sold in period',
t4.verkocht_sd as 'Units sold in period, outliers removed',
COUNT(*) as 'Number of invoices in period',
b.art_code as 'Part number'
FROM bongegs b -- Table that has all the sales records for all products
RIGHT JOIN totvrd ON (totvrd.art_code = b.art_code) -- Right Join stock data to also include items that are not in table bongegs (no sales history).
LEFT JOIN artcred ON (artcred.art_code = b.art_code) -- add supplier ID to the part numbers.
LEFT JOIN
(
SELECT
SUM(b.aantal) as verkocht_sd,
b.art_code
FROM bongegs b
RIGHT JOIN totvrd ON (totvrd.art_code = b.art_code)
LEFT JOIN artcred ON (artcred.art_code = b.art_code)
WHERE
b.bon_datum > DATE_SUB(CURDATE(), INTERVAL 6 MONTH)
and b.bon_soort = "f" -- Selects only invoices
and artcred.vln = 1 -- 1 = Prefered supplier
and artcred.cred_nr = 9117 -- Supplier ID
and b.aantal < (select * from (SELECT AVG(b.aantal)+3*STDDEV(aantal)
FROM bongegs b
WHERE
b.bon_soort = 'f' and
b.bon_datum > DATE_SUB(CURDATE(), INTERVAL 6 MONTH)) x)
GROUP BY b.art_code
) AS t4
ON (b.art_code = t4.art_code)
WHERE
b.bon_datum > DATE_SUB(CURDATE(), INTERVAL 6 MONTH)
and b.bon_soort = "f"
and artcred.vln = 1
and artcred.cred_nr = 9117
GROUP BY b.art_code
Bongegs | all rows from sales forms (invoices F, offers O, delivery notes V)
| art_code | bon_datum | bon_soort | aantal |
|:---------|:---------: |:---------:|:------:|
| item_1 | 2021-08-21 | f | 6 |
| item_2 | 2021-08-29 | v | 3 |
| item_6 | 2021-09-03 | o | 2 |
| item_4 | 2021-10-21 | f | 6 |
| item_1 | 2021-11-21 | o | 6 |
| item_3 | 2022-01-17 | v | 6 |
| item_1 | 2022-01-21 | o | 6 |
| item_4 | 2022-01-26 | f | 6 |
Artcred | supplier ID's
| art_code | vln | cred_nr |
|:---------|:----:|:-------:|
| item_1 | 1 | 1001 |
| item_2 | 1 | 1002 |
| item_3 | 1 | 1001 |
| item_4 | 1 | 1007 |
| item_5 | 1 | 1004 |
| item_5 | 2 | 1008 |
| item_6 | 1 | 1016 |
| item_7 | 1 | 1567 |
totvrd | stock
| art_code | voorraad | reserved | backorder |
|:---------|:---------: |:--------:|:---------:|
| item_1 | 1 | 0 | 5 |
| item_2 | 0 | 0 | 0 |
| item_3 | 88 | 0 | 0 |
| item_4 | 9 | 0 | 0 |
| item_5 | 67 | 2 | 20 |
| item_6 | 112 | 9 | 0 |
| item_7 | 65 | 0 | 0 |
| item_8 | 7 | 1 | 0 |
Now, on to the query. You have LEFT JOINs to the artcred table, but then include artcred in the WHERE clause making it an INNER JOIN (required both left and right tables) in the result. Was this intended, or are you expecting more records in the bongegs table that do NOT exist in the artcred.
Well to be honest I was not fully aware that this would essentially form an INNER JOIN but in this case it doesn't really matter. A record that exists in bongegs always exists in artcred as well (every sold product must have a supplier). That doesn't work both ways since a product can be in artcred without ever being sold.
You also have RIGHT JOIN on totvrd which implies you want every record in the TotVRD table regardless of a record in the bongegs table. Is this correct?
Yes it is intended. Otherwise only products with actual sales in the period would end up in the result and we also wanted to include products with zero sales.
One simplification:
and b.aantal < ( SELECT * from ( SELECT AVG ...
-->
and b.aantal < ( SELECT AVG ...
A personal problem: my brain hurts when I see RIGHT JOIN; please rewrite as LEFT JOIN.
Check you RIGHTs and LEFTs -- that keeps the other table's rows even if there is no match; are you expecting such NULLs? That is, it looks like they can all be plain JOINs (aka INNER JOINs).
These might help performance:
b: INDEX(bon_soort, bon_datum, aantal, art_code)
totvrd: INDEX(art_code)
artcred: INDEX(vln, cred_nr, art_code)
Is b the what you keep needing? Build a temp table:
CREATE TEMPORARY TABLE tmp_b
SELECT ...
FROM b
WHERE ...;
But if you need to use tmp_b multiple times in the same query, (and since you are not yet on MySQL 8.0), you may need to make it a non-TEMPORARY table for long enough to run the query. (If you have multiple connections building the same permanent table, there will be trouble.)
Yes, 5.5.33 is rather antique; upgrade soon.
(pre
By getting what I believe are all the pieces you had, I think this query significantly simplifies the query. Lets first start with the fact that you were trying to eliminate the outliers by selecting the standard deviation stuff as what to be excluded. Then you had the original summation of all sales also from the bongegs table.
To simplify this, I have the sub-query ONCE internal that does the summation, counts, avg, stddev of all orders (f) within the last 6 months. I also computed the divide by 6 for per-month you wanted in the top.
Since the bongegs is now all pre-aggregated ONCE, and grouped per art_code, it does not need to be done one after the other. You can use the totals directly at the top (at least I THINK is similar output without all actual data and understanding of your context).
So the primary table is the product table (Voorraad) and LEFT-JOINED to the pre-query of bongegs. This allows you to get all products regardless of those that have been sold.
Since the one aggregation prequery has the avg and stddev in it, you can simply apply an additional AND clause when joining based on the total sold being less than the avg/stddev context.
The resulting query below.
SELECT
-- appears you are looking for the highest percentage?
-- typically NOT a good idea to name columns starting with numbers,
-- but ok. Typically let interface/output name the columns to end-users
GREATEST((b.verkocht_sdperMonth * ((100 + 0)/100)),0) as 'units sold p/month',
-- appears to be the total sold divided by 6 to get monthly average over 6 months query of data
GREATEST( ROUND(
( (b.verkocht_sdperMonth * 3) - v.voorraad + v.reserved - v.backorder), 0), 0)
as 'Order based on units sold',
b.verkocht_sd as 'Units sold in period',
b.AvgStdDev as 'AvgStdDeviation',
b.NumInvoices as 'Number of invoices in period',
v.art_code as 'Part number'
FROM
-- stock, master inventory, regardless of supplier
-- get all products, even though not all may be sold
Voorraad v
-- LEFT join to pre-query of Bongegs pre-grouped by the art_code which appears
-- to be basis of all other joins, std deviation and average while at it
LEFT JOIN
(select
b.arc_code,
count(*) NumInvoices,
sum( b.aantal ) verkocht_sd,
sum( b.aantal ) / 6.0 verkocht_sdperMonth,
avg( b.aantal ) AvgSale,
AVG(b.aantal) + 3 * STDDEV( b.aantal) AvgStdDev
from
bongegs b
JOIN artcred ac
on b.art_code = ac.art_code
AND ac.vln = 1
and ac.cred_nr = 9117
where
-- only for ORDERS ('f') and within last 6 months
b.bon_soort = 'f'
AND b.bon_datum > DATE_SUB(CURDATE(), INTERVAL 6 MONTH)
group by
b.arc_code ) b
-- result is one entry per arc_code, thus preventing any Cartesian product
ON v.art_code = b.art_code
GROUP BY
v.art_code
So I've got two tables:
ic:
id | created
---|--------------------
1 | 2016-03-31 16:20:03
2 | 2016-03-31 16:25:18
3 | 2016-03-31 16:28:09
status:
id | ic_id | timestamp
---|--------------------
1 | 1 | 2016-03-31 16:20:03
2 | 5 | 2016-03-31 16:25:18
3 | 5 | 2016-03-31 16:28:09
I now want to find the average difference between the ic.created and the status.timestamp of the first corresponding record in the status table. I started out with this:
SELECT status.`timestamp` - informed_consent.created as difference
FROM status
WHERE status.`timestamp` > '2017-06-19'
AND status.ic_id IN (
SELECT informed_consent.id
FROM informed_consent
WHERE informed_consent.id = status.ic_id;
);
But I immediately get an error in my mysql syntax. I guess I can also use a left join, but I'm kinda lost here.
Could anybody help me out in the right direction?
I think you can use inner join instead of sub query due to informed_consent.created is not present in outer sql,it's in the sub query
SELECT status.`timestamp`,status.`timestamp` - informed_consent.created as difference
FROM status
JOIN informed_consent ON informed_consent.id = status.ic_id
WHERE status.`timestamp` > '2017-06-19'
Ok, I think the answer of this is somewhere but I can't find it...
(and even my title is bad)
To be short, I want to get the fewest number of group I can make from a part of an association table
1st, Keep in mind this is already a result of a 5 table (+1k line) join with filter and grouping, that I'll have to run many time on a prod server as powerful as a banana...
2nd, This is a fake case that picture you my problem
After some Querying, I've got this data result :
+--------------------+
|id_course|id_teacher|
+--------------------+
| 6 | 1 |
| 6 | 4 |
| 6 | 14 |
| 33 | 1 |
| 33 | 4 |
| 34 | 1 |
| 34 | 4 |
| 34 | 10 |
+--------------------+
As you can see, I've got 3 courses, witch are teach by up to 3 teacher. I need to attend at one of every course, but I want as few different teacher as possible (I'm shy...).
My first query
Should answer : what is the smallest number of teacher I need to cover every unique course ?
With this data, it's a 1, cause Teacher 1 or Teacher 4 make courses for these 3 one.
Second query
Now that I've already get these courses, I want to go to two other courses, the 32 and the 50, with this schedule :
+--------------------+
|id_course|id_teacher|
+--------------------+
| 32 | 1 |
| 32 | 12 |
| 50 | 12 |
+--------------------+
My question is : For id_course N, will I have to get one more teacher ?
I want to check course by course, so "check for course 32", no need to check many at the same time
The best way I think is to count an inner join with a list of teacher of same fewest rank from the first query, so with our data we got only two : Teacher(1, 4).
For the Course 32, Teacher2 don't do this one, but as the Teacher1 do Courses(6, 33, 34, 32) I don't have to get another teacher.
For the Course 50, the only teacher to do it is the Teacher12, so I'll not find a match in my choice of teacher, and I'll have to get one more (so two in total with these data)
Here is a base [SQLFiddle
Best regards, Blag
You want to get a distinct count of ID_Teachers with the least count then... get a distinct count and limit the results to 1 record.
So perhaps something like...
SELECT count(Distinct ID_Teacher), Group_concat(ID_Teacher) as TeachersIDs
FROM Table
WHERE ID_Course in ('Your List')
ORDER BY count(Distinct ID_Teacher) ASC Limit 1
However this will randomly select if a tie exists... so do you want to provide the option to select which group of teachers and classes should ties exist? Meaning there are multiple paths to fulfill all classes involving the same number of teachers... For example teachers A, B and A, C fulfill all required classes.... should both records return in the result or is 1 sufficient?
So I've finally found a way to do what I want !
For the first query, as my underlying real need was "is there a single teacher to do everything", I've lower a bit my expectation and go for this one (58 lines on my true case u_u") :
SELECT
(
SELECT count(s.id_teacher) nb
FROM t AS m
INNER JOIN t AS s
ON m.id_teacher = s.id_teacher
GROUP BY m.id_course, m.id_teacher
ORDER BY nb DESC
LIMIT 1
) AS nbMaxBySingleTeacher,
(
SELECT COUNT(DISTINCT id_course) nb
FROM t
) AS nbTotalCourseToDo
[SQLFiddle
And I get back two value that answer my question "is one teacher enough ?"
+--------------------------------------+
|nbMaxBySingleTeacher|nbTotalCourseToDo|
+--------------------------------------+
| 4 | 5 |
+--------------------------------------+
The 2nd query use the schedule of new course, and take the id of one I want to check. It should tell me if I need to get one more teacher, or if it's ok with my actual(s) one.
SELECT COUNT(*) nb
FROM (
SELECT
z.id_teacher
FROM z
WHERE
z.id_course = 50
) t1
WHERE
FIND_IN_SET(t1.id_teacher, (
SELECT GROUP_CONCAT(t2.id_teacher) lst
FROM (
SELECT DISTINCT COUNT(s.id_teacher) nb, m.id_teacher
FROM t AS m
INNER JOIN t AS s
ON m.id_teacher = s.id_teacher
GROUP BY m.id_course, m.id_teacher
ORDER BY nb DESC
) t2
GROUP BY t2.nb
ORDER BY nb DESC
LIMIT 1
));
[SQLFiddle
This tell me the number of teacher that are able to teach the courses I already have AND the new one I want. So if it's over zero, then I don't need a new teacher :
+--+
|nb|
+--+
|1 |
+--+
I want to return all rows that have a certain value in a column and have more than 5 instances in which a number is that certain value. For example, I would like to return all rows of the condition in which if the value in the column M has the number 1 in it and there are 5 or more instances of M having the number 1 in it, then it will return all rows with that condition.
select *
from tab
where M = 1
group by id --ID is the primary key of the table
having count(M) > 5;
EDIT: Here is my table:
id | M | price
--------+-------------+-------
1 | | 100
2 | 1 | 50
3 | 1 | 30
4 | 2 | 20
5 | 2 | 10
6 | 3 | 20
7 | 1 | 1
8 | 1 | 1
9 | 1 | 1
10 | 1 | 1
11 | 1 | 1
Originally I just want to insert into a trigger so that if the number of M = 1's is greater than 5, then I want to create an exception. The query I asked for would be inserted into the trigger. END EDIT.
But my table is always empty. Can anyone help me out? Thanks!
Try this :
select *
from tab
where M in (select M from tab where M = 1 group by M having count(id) > 5);
SQL Fiddle Demo
please try
select *,count(M) from table where M=1 group by id having count(M)>5
Since you group on your PK (which seems a futile excercise), you are counting per ID, whicg will indeed always return 1.
As i explain after this code, this query is NOT good, it is NOT the answer, and i also explain WHY. Please do not expect this query to run correctly!
select *
from tab
where M = 1
group by M
having count(*) > 5;
Like this, you group on what you are counting, which makes a lot more sense. At the same time, this will have unexpected behaviour, as you are selecting all kinds of columns that are not in the group by or in any aggregate. I know mySQL is lenient on that, but I don;t even want to know what it will produce.
Try indeed a subquery along these lines:
select *
from tab
where M in
(SELECT M
from tab
group by M
having count(*) > 5)
I've built a SQLFiddle demo (i used 'Test' as table name out of habit) accomplishing this (I don't have a mySQL at hand now to test it).
-- Made up a structure for testing
CREATE TABLE Test (
id INT NOT NULL AUTO_INCREMENT,
PRIMARY KEY(id),
M int
);
SELECT id, M FROM tab
WHERE M IN (
SELECT M
FROM Test
WHERE M = 1
GROUP BY M
HAVING COUNT(M) > 5
)
The sub-query is a common "find the duplicates" kind of query, with the added condition of a specific value for the column M, also stating that there must be at least 5 dupes.
It will spit out a series of values of M which you can use to query the table against, ending with the rows you need.
You shouldn't use SELECT * , it's a bad practice in general: don't retrieve data you aren't actually using, and if you are using it then take the little time needed to type in a list of field, you'll likely see faster querying and on the other hand the code will be way more readable.
I am new in MySQL, I am trying to find:
The difference between a given day's maximum value occurred and the previous day's maximum value.
I was able to get the maximum values for dates via:
select max(`bundle_count`), `Production_date`
from `table`
group by `Production_date`
But I don't know how to use SQL to calculate the differences between maximums for two given dates.
am expecting output like this
Please help me.
Update 1: Here is a fiddle, http://sqlfiddle.com/#!2/818ad/2, that I used for testing.
Update 2: Here is a fiddle, http://sqlfiddle.com/#!2/3f78d/10 that I used for further refining/fixing, based on Sandy's comments.
Update 3: For some reason the case where there is no previous day was not being dealt with correctly. I thought it was. However, I've updated to make sure that works (a bit cumbersome--but it appears to be right. Last fiddle: http://sqlfiddle.com/#!2/3f78d/45
I think #Grijesh conceptually got you the main thing you needed via the self-join of the input data (so make sure you vote up his answer!). I've cleaned up his query a bit on syntax (building off of his query!):
SELECT
DATE(t1.`Production_date`) as theDate,
MAX( t1.`bundle_count` ) AS 'max(bundle_count)',
MAX( t1.`bundle_count` ) -
IF(
EXISTS
(
SELECT date(t2.production_date)
FROM input_example t2
WHERE t2.machine_no = 1 AND
date_sub(date(t1.production_date), interval 1 day) = date(t2.production_date)
),
(
SELECT MAX(t3.bundle_count)
FROM input_example t3
WHERE t3.machine_no = 1 AND
date_sub(date(t1.production_date), interval 1 day) = date(t3.production_date)
GROUP BY DATE(t3.production_date)
), 0
)
AS Total_Bundles_Used
FROM `input_example` t1
WHERE t1.machine_no = 1
GROUP BY DATE( t1.`production_date` )
Note 1: I think #Grijesh and I were cleaning up the query syntax issues at the same time. It's encouraging that we ended up with very similar versions after we were both doing cleanup. My version differs in using IFNULL() for when there is no preceding data. I also ended up with a DATE_SUB, and I made sure to reduce various dates to mere dates without time component, via DATE()
Note 2: I originally had not fully understood your source tables, so I thought I needed to implement a running count in the query. But upon better inspection, it's clear that your source data already has a running count, so I took that stuff back out.
I am not sure but you need something like this, Hope it will be helpful to you upto some extend:
Try this:
SELECT t1.`Production_date` ,
MAX(t1.`bundle_count`) - MAX(t2.`bundle_count`) ,
COUNT(t1.`bundle_count`)
FROM `table_name` AS t1
INNER JOIN `table_name` AS t2
ON ABS(DATEDIFF(t1.`Production_date` , t2.`Production_date`)) = 1
GROUP BY t1.`Production_date`
EDIT
I create a table name = 'table_name', as below,
mysql> SELECT * FROM `table_name`;
+---------------------+--------------+
| Production_date | bundle_count |
+---------------------+--------------+
| 2004-12-01 20:37:22 | 1 |
| 2004-12-01 20:37:22 | 2 |
| 2004-12-01 20:37:22 | 3 |
| 2004-12-02 20:37:22 | 2 |
| 2004-12-02 20:37:22 | 5 |
| 2004-12-02 20:37:22 | 7 |
| 2004-12-03 20:37:22 | 6 |
| 2004-12-03 20:37:22 | 7 |
| 2004-12-03 20:37:22 | 2 |
| 2004-12-04 20:37:22 | 1 |
| 2004-12-04 20:37:22 | 9 |
+---------------------+--------------+
11 rows in set (0.00 sec)
My query: to find difference in bundle_count between two consecutive dates:
SELECT t1.`Production_date` ,
MAX(t2.`bundle_count`) - MAX(t1.`bundle_count`) ,
COUNT(t1.`bundle_count`)
FROM `table_name` AS t1
INNER JOIN `table_name` AS t2
ON ABS(DATEDIFF(t1.`Production_date` , t2.`Production_date`)) = 1
GROUP BY t1.Production_date;
its output:
+---------------------+-------------------------------------------------+--------------------------+
| Production_date | MAX(t2.`bundle_count`) - MAX(t1.`bundle_count`) | COUNT(t1.`bundle_count`) |
+---------------------+-------------------------------------------------+--------------------------+
| 2004-12-01 20:37:22 | 4 | 9 |
| 2004-12-02 20:37:22 | 0 | 18 |
| 2004-12-03 20:37:22 | 2 | 15 |
| 2004-12-04 20:37:22 | -2 | 6 |
+---------------------+-------------------------------------------------+--------------------------+
4 rows in set (0.00 sec)
This is PostgreSQL syntax (sorry; it's what I'm familiar with) but should fundamentally work in either database. Note this doesn't exactly run in PostgreSQL either because group is not a valid table name (it's a reserved keyword). The approach is a self-join as others have mentioned but I've used a view to handle the max-by-day and the difference as separate steps.
create view max_by_day as
select
date_trunc('day', production_date) as production_date,
max(bundle_count) as bundle_count
from
group
group by
date_trunc('day', production_date);
select
today.production_date as production_date,
today.bundle_count,
today.bundle_count - coalesce(yesterday.bundle_count, 0)
from
max_by_day as today
left join max_by_day yesterday on (yesterday.production_date = today.production_date - '1 day'::interval)
order by
production_date;
PostgreSQL also has a construct called window functions which is useful for this and a bit easier to understand. Just had to stick in a bit of advocacy for a superior database. :-P
select
date_trunc('day', production_date),
max(bundle_count),
max(bundle_count) - lag(max(bundle_count), 1, 0)
over
(order by date_trunc('day', production_date))
from
group
group by
date_trunc('day', production_date);
These two approaches differ in how they handle missing days in the data - the first will treat it as a 0, the second will use the previous day which is present. There wasn't a case like this in your sample so I don't know if this is something you care about.