Create a Timeline Query with date periods in SQL ACCESS - sql

I would like to write a Query in SQL Access (no choice with the software...) in order to create a timeline from a Table of records.
Hereunder a simple example in order to explain what I want to do.
I have a table with 2 records. Each record is defined by its ID, a Value, a date of begin, a date of end and a Level of confidence.
Id | vValue | dtBegin | dtEnd | lLevel
-------------------------------------------
1 | a |20/06/2016|28/06/2016| Low
2 | b |23/06/2016|25/06/2016| High
The query should return a timeline with the highest level of information (and its value) available for each period of time. In the example the result of the query should be:
vValue| dtBegin | dtEnd |lLevel
------------------------------------
a |20/06/2016|23/06/2016| Low
b |23/06/2016|25/06/2016| High
a |25/06/2016|28/06/2016| Low
From the 20/06/2016 to the 23/06/2016 the highest level of confidence available in the table is 'Low' and the associated value is 'a'/
From the 23/06 to the 25/06, the period is covered by two records but the highest level of confidence available is 'High', therefore the value is 'b'
Thanks a lot for your help
Thibaud

There may be an easier way of doing it, you could start with a base query to get all dates necessary:
SELECT dtBegin
FROM Table1
GROUP BY dtBegin
UNION
SELECT dtEnd AS dtBegin
FROM Table1
With this query you can get the non-overlapping date ranges:
SELECT Timeline.dtBegin, Min(Timeline_1.dtBegin) AS dtEnd
FROM Query1 AS Timeline, Query1 AS Timeline_1
WHERE (((Timeline_1.dtBegin) > Timeline.dtBegin))
GROUP BY Timeline.dtBegin
From this you could cross join back to your original table, and get the max level and vValue for any table entry that overlaps the date range using a subquery:
SELECT Query2.dtBegin, Query2.dtEnd, Table1.lLevel, Table1.vValue
FROM Query2, Table1
WHERE (((Table1.lLevel) =
(SELECT MAX(lLevel)
FROM Table1 T
WHERE T.dtBegin <=Query2.dtBegin AND
T.dtEnd >=Query2.dtEnd))) AND
((Table1.dtBegin)<=Query2.dtBegin) AND
((Table1.dtEnd)>=Query2.dtEnd)
You'll have to do some extra logic to get a max level, of course, depending on how your choices are configured.

It's hard to help without seeing the data structure. If the begin / end dates don't overlap in your data this could possibly work:
/* Finds distinct dates which only have a low record */
SELECT
vValue,dtBegin,dtEnd,'Low' AS lLevel
FROM
(
SELECT DISTINCT vValue,dtBegin,dtEnd
FROM tbl
WHERE lLevel = "Low"
EXCEPT
SELECT DISTINCT vValue,dtBegin,dtEnd
FROM tbl
WHERE lLevel = "High"
) AS X
/* Appends distinct date sets */
UNION
SELECT DISTINCT vValue,dtBegin,dtEnd,lLevel
FROM tbl
WHERE lLevel = "High"

Related

Need help in sql Query.neo

In this table there are three colum and in need the value for of data which are lesser than code = 28,this is my query
SELECT value,code,date
FROM table
order by date,vchcode
but when i ad where clouse like
SELECT value,code,date
FROM table
where code < 28
order by date,vchcode
is only shows 2 row with code 26 and 27... i need 26,27 and 32.. and table colums are variable its not fix..
I think you wnat to take the date into account -- what you really want are all rows before the date of the row with code 28.
One method uses a subquery:
SELECT t.value, t.code, t.date
FROM table t
WHERE date < (SELECT date FROM table t2 WHERE t2.code = 28)
ORDER BY t.date, t.vchcode

SQL Server iterating through time series data

I am using SQL Server and wondering if it is possible to iterate through time series data until specific condition is met and based on that label my data in other table?
For example, let's say I have a table like this:
Id Date Some_kind_of_event
+--+----------+------------------
1 |2018-01-01|dsdf...
1 |2018-01-06|sdfs...
1 |2018-01-29|fsdfs...
2 |2018-05-10|sdfs...
2 |2018-05-11|fgdf...
2 |2018-05-12|asda...
3 |2018-02-15|sgsd...
3 |2018-02-16|rgw...
3 |2018-02-17|sgs...
3 |2018-02-28|sgs...
What I want to get, is to calculate for each key the difference between two adjacent events and find out if there exists difference > 10 days between these two adjacent events. In case yes, I want to stop iterating for that specific key and put label 'inactive', otherwise 'active' in my other table. After we finish with one key, we start with another.
So for example id = 1 would get label 'inactive' because there exists two dates which have difference bigger that 10 days. The final result would be like that:
Id Label
+--+----------+
1 |inactive
2 |active
3 |inactive
Any ideas how to do that? Is it possible to do it with SQL?
When working with a DBMS you need to get away from the idea of thinking iteratively. Instead you need to try and think in sets. "Instead of thinking about what you want to do to a row, think about what you want to do to a column."
If I understand correctly, is this what you're after?
CREATE TABLE SomeEvent (ID int, EventDate date, EventName varchar(10));
INSERT INTO SomeEvent
VALUES (1,'20180101','dsdf...'),
(1,'20180106','sdfs...'),
(1,'20180129','fsdfs..'),
(2,'20180510','sdfs...'),
(2,'20180511','fgdf...'),
(2,'20180512','asda...'),
(3,'20180215','sgsd...'),
(3,'20180216','rgw....'),
(3,'20180217','sgs....'),
(3,'20180228','sgs....');
GO
WITH Gaps AS(
SELECT *,
DATEDIFF(DAY,LAG(EventDate) OVER (PARTITION BY ID ORDER BY EventDate),EventDate) AS EventGap
FROM SomeEvent)
SELECT ID,
CASE WHEN MAX(EventGap) > 10 THEN 'inactive' ELSE 'active' END AS Label
FROM Gaps
GROUP BY ID
ORDER BY ID;
GO
DROP TABLE SomeEvent;
GO
This assumes you are using SQL Server 2012+, as it uses the LAG function, and SQL Server 2008 has less than 12 months of any kind of support.
Try this. Note, replace #MyTable with your actual table.
WITH Diffs AS (
SELECT
Id
,DATEDIFF(DAY,[Date],LEAD([Date],1,0) OVER (ORDER BY [Id], [Date])) Diff
FROM #MyTable)
SELECT
Id
,CASE WHEN MAX(Diff) > 10 THEN 'Inactive' ELSE 'Active' END
FROM Diffs
GROUP BY Id
Just to share another approach (without a CTE).
SELECT
ID
, CASE WHEN SUM(TotalDays) = (MAX(CNT) - 1) THEN 'Active' ELSE 'Inactive' END Label
FROM (
SELECT
ID
, EventDate
, CASE WHEN DATEDIFF(DAY, EventDate, LEAD(EventDate) OVER(PARTITION BY ID ORDER BY EventDate)) < 10 THEN 1 ELSE 0 END TotalDays
, COUNT(ID) OVER(PARTITION BY ID) CNT
FROM EventsTable
) D
GROUP BY ID
The method is counting how many records each ID has, and getting the TotalDays by date differences (in days) between the current the next date, if the difference is less than 10 days, then give me 1, else give me 0.
Then compare, if the total days equal the number of records that each ID has (minus one) would print Active, else Inactive.
This is just another approach that doesn't use CTE.

Select in specific days range even if there is no data [duplicate]

I'm building a quick csv from a mysql table with a query like:
select DATE(date),count(date) from table group by DATE(date) order by date asc;
and just dumping them to a file in perl over a:
while(my($date,$sum) = $sth->fetchrow) {
print CSV "$date,$sum\n"
}
There are date gaps in the data, though:
| 2008-08-05 | 4 |
| 2008-08-07 | 23 |
I would like to pad the data to fill in the missing days with zero-count entries to end up with:
| 2008-08-05 | 4 |
| 2008-08-06 | 0 |
| 2008-08-07 | 23 |
I slapped together a really awkward (and almost certainly buggy) workaround with an array of days-per-month and some math, but there has to be something more straightforward either on the mysql or perl side.
Any genius ideas/slaps in the face for why me am being so dumb?
I ended up going with a stored procedure which generated a temp table for the date range in question for a couple of reasons:
I know the date range I'll be looking for every time
The server in question unfortunately was not one that I can install perl modules on atm, and the state of it was decrepit enough that it didn't have anything remotely Date::-y installed
The perl Date/DateTime-iterating answers were also very good, I wish I could select multiple answers!
When you need something like that on server side, you usually create a table which contains all possible dates between two points in time, and then left join this table with query results. Something like this:
create procedure sp1(d1 date, d2 date)
declare d datetime;
create temporary table foo (d date not null);
set d = d1
while d <= d2 do
insert into foo (d) values (d)
set d = date_add(d, interval 1 day)
end while
select foo.d, count(date)
from foo left join table on foo.d = table.date
group by foo.d order by foo.d asc;
drop temporary table foo;
end procedure
In this particular case it would be better to put a little check on the client side, if current date is not previos+1, put some addition strings.
When I had to deal with this problem, to fill in missing dates I actually created a reference table that just contained all dates I'm interested in and joined the data table on the date field. It's crude, but it works.
SELECT DATE(r.date),count(d.date)
FROM dates AS r
LEFT JOIN table AS d ON d.date = r.date
GROUP BY DATE(r.date)
ORDER BY r.date ASC;
As for output, I'd just use SELECT INTO OUTFILE instead of generating the CSV by hand. Leaves us free from worrying about escaping special characters as well.
not dumb, this isn't something that MySQL does, inserting the empty date values. I do this in perl with a two-step process. First, load all of the data from the query into a hash organised by date. Then, I create a Date::EzDate object and increment it by day, so...
my $current_date = Date::EzDate->new();
$current_date->{'default'} = '{YEAR}-{MONTH NUMBER BASE 1}-{DAY OF MONTH}';
while ($current_date <= $final_date)
{
print "$current_date\t|\t%hash_o_data{$current_date}"; # EzDate provides for automatic stringification in the format specfied in 'default'
$current_date++;
}
where final date is another EzDate object or a string containing the end of your date range.
EzDate isn't on CPAN right now, but you can probably find another perl mod that will do date compares and provide a date incrementor.
You could use a DateTime object:
use DateTime;
my $dt;
while ( my ($date, $sum) = $sth->fetchrow ) {
if (defined $dt) {
print CSV $dt->ymd . ",0\n" while $dt->add(days => 1)->ymd lt $date;
}
else {
my ($y, $m, $d) = split /-/, $date;
$dt = DateTime->new(year => $y, month => $m, day => $d);
}
print CSV, "$date,$sum\n";
}
What the above code does is it keeps the last printed date stored in a
DateTime object $dt, and when the current date is more than one day
in the future, it increments $dt by one day (and prints it a line to
CSV) until it is the same as the current date.
This way you don't need extra tables, and don't need to fetch all your
rows in advance.
I hope you will figure out the rest.
select * from (
select date_add('2003-01-01 00:00:00.000', INTERVAL n5.num*10000+n4.num*1000+n3.num*100+n2.num*10+n1.num DAY ) as date from
(select 0 as num
union all select 1
union all select 2
union all select 3
union all select 4
union all select 5
union all select 6
union all select 7
union all select 8
union all select 9) n1,
(select 0 as num
union all select 1
union all select 2
union all select 3
union all select 4
union all select 5
union all select 6
union all select 7
union all select 8
union all select 9) n2,
(select 0 as num
union all select 1
union all select 2
union all select 3
union all select 4
union all select 5
union all select 6
union all select 7
union all select 8
union all select 9) n3,
(select 0 as num
union all select 1
union all select 2
union all select 3
union all select 4
union all select 5
union all select 6
union all select 7
union all select 8
union all select 9) n4,
(select 0 as num
union all select 1
union all select 2
union all select 3
union all select 4
union all select 5
union all select 6
union all select 7
union all select 8
union all select 9) n5
) a
where date >'2011-01-02 00:00:00.000' and date < NOW()
order by date
With
select n3.num*100+n2.num*10+n1.num as date
you will get a column with numbers from 0 to max(n3)*100+max(n2)*10+max(n1)
Since here we have max n3 as 3, SELECT will return 399, plus 0 -> 400 records (dates in calendar).
You can tune your dynamic calendar by limiting it, for example, from min(date) you have to now().
Since you don't know where the gaps are, and yet you want all the values (presumably) from the first date in your list to the last one, do something like:
use DateTime;
use DateTime::Format::Strptime;
my #row = $sth->fetchrow;
my $countdate = strptime("%Y-%m-%d", $firstrow[0]);
my $thisdate = strptime("%Y-%m-%d", $firstrow[0]);
while ($countdate) {
# keep looping countdate until it hits the next db row date
if(DateTime->compare($countdate, $thisdate) == -1) {
# counter not reached next date yet
print CSV $countdate->ymd . ",0\n";
$countdate = $countdate->add( days => 1 );
$next;
}
# countdate is equal to next row's date, so print that instead
print CSV $thisdate->ymd . ",$row[1]\n";
# increase both
#row = $sth->fetchrow;
$thisdate = strptime("%Y-%m-%d", $firstrow[0]);
$countdate = $countdate->add( days => 1 );
}
Hmm, that turned out to be more complicated than I thought it would be.. I hope it makes sense!
I think the simplest general solution to the problem would be to create an Ordinal table with the highest number of rows that you need (in your case 31*3 = 93).
CREATE TABLE IF NOT EXISTS `Ordinal` (
`n` int(10) unsigned NOT NULL AUTO_INCREMENT, PRIMARY KEY (`n`)
);
INSERT INTO `Ordinal` (`n`)
VALUES (NULL), (NULL), (NULL); #etc
Next, do a LEFT JOIN from Ordinal onto your data. Here's a simple case, getting every day in the last week:
SELECT CURDATE() - INTERVAL `n` DAY AS `day`
FROM `Ordinal` WHERE `n` <= 7
ORDER BY `n` ASC
The two things you would need to change about this are the starting point and the interval. I have used SET #var = 'value' syntax for clarity.
SET #end = CURDATE() - INTERVAL DAY(CURDATE()) DAY;
SET #begin = #end - INTERVAL 3 MONTH;
SET #period = DATEDIFF(#end, #begin);
SELECT #begin + INTERVAL (`n` + 1) DAY AS `date`
FROM `Ordinal` WHERE `n` < #period
ORDER BY `n` ASC;
So the final code would look something like this, if you were joining to get the number of messages per day over the last three months:
SELECT COUNT(`msg`.`id`) AS `message_count`, `ord`.`date` FROM (
SELECT ((CURDATE() - INTERVAL DAY(CURDATE()) DAY) - INTERVAL 3 MONTH) + INTERVAL (`n` + 1) DAY AS `date`
FROM `Ordinal`
WHERE `n` < (DATEDIFF((CURDATE() - INTERVAL DAY(CURDATE()) DAY), ((CURDATE() - INTERVAL DAY(CURDATE()) DAY) - INTERVAL 3 MONTH)))
ORDER BY `n` ASC
) AS `ord`
LEFT JOIN `Message` AS `msg`
ON `ord`.`date` = `msg`.`date`
GROUP BY `ord`.`date`
Tips and Comments:
Probably the hardest part of your query was determining the number of days to use when limiting Ordinal. By comparison, transforming that integer sequence into dates was easy.
You can use Ordinal for all of your uninterrupted-sequence needs. Just make sure it contains more rows than your longest sequence.
You can use multiple queries on Ordinal for multiple sequences, for example listing every weekday (1-5) for the past seven (1-7) weeks.
You could make it faster by storing dates in your Ordinal table, but it would be less flexible. This way you only need one Ordinal table, no matter how many times you use it. Still, if the speed is worth it, try the INSERT INTO ... SELECT syntax.
Use some Perl module to do date calculations, like recommended DateTime or Time::Piece (core from 5.10). Just increment date and print date and 0 until date will match current.
I don't know if this would work, but how about if you created a new table which contained all the possible dates (that might be the problem with this idea, if the range of dates is going to change unpredictably...) and then do a left join on the two tables? I guess it's a crazy solution if there are a vast number of possible dates, or no way to predict the first and last date, but if the range of dates is either fixed or easy to work out, then this might work.

SQL query group by nearby timestamp

I have a table with a timestamp column. I would like to be able to group by an identifier column (e.g. cusip), sum over another column (e.g. quantity), but only for rows that are within 30 seconds of each other, i.e. not in fixed 30 second bucket intervals. Given the data:
cusip| quantity| timestamp
============|=========|=============
BE0000310194| 100| 16:20:49.000
BE0000314238| 50| 16:38:38.110
BE0000314238| 50| 16:46:21.323
BE0000314238| 50| 16:46:35.323
I would like to write a query that returns:
cusip| quantity
============|=========
BE0000310194| 100
BE0000314238| 50
BE0000314238| 100
Edit:
In addition, it would greatly simplify things if I could also get the MIN(timestamp) out of the query.
From Sean G solution, I have removed Group By on complete Table. In Fact re adjected few parts for Oracle SQL.
First after finding previous time, assign self parent id. If there a null in Previous Time, then we exclude giving it an ID.
Now based on take the nearest self parent id by avoiding nulls so that all nearest 30 seconds cusip fall under one Group.
As There is a CUSIP column, I assumed the dataset would be large market transactional data. Instead using group by on complete table, use partition by CUSIP and final Group Parent ID for better performance.
SELECT
id,
sub.parent_id,
sub.cusip,
timestamp,
quantity,
sum(sub.quantity) OVER(
PARTITION BY cusip, parent_id
) sum_quantity,
MIN(sub.timestamp) OVER(
PARTITION BY cusip, parent_id
) min_timestamp
FROM
(
SELECT
base_sub.*,
CASE
WHEN base_sub.self_parent_id IS NOT NULL THEN
base_sub.self_parent_id
ELSE
LAG(base_sub.self_parent_id) IGNORE NULLS OVER(
PARTITION BY cusip
ORDER BY
timestamp, id
)
END parent_id
FROM
(
SELECT
c.*,
CASE
WHEN nvl(abs(EXTRACT(SECOND FROM to_timestamp(previous_timestamp, 'yyyy/mm/dd hh24:mi:ss') - to_timestamp
(timestamp, 'yyyy/mm/dd hh24:mi:ss'))), 31) > 30 THEN
id
ELSE
NULL
END self_parent_id
FROM
(
SELECT
my_table.id,
my_table.cusip,
my_table.timestamp,
my_table.quantity,
LAG(my_table.timestamp) OVER(
PARTITION BY my_table.cusip
ORDER BY
my_table.timestamp, my_table.id
) previous_timestamp
FROM
my_table
) c
) base_sub
) sub
Below are the Table Rows
Input Data:
Below is the Output
RESULT
Following may be helpful to you.
Grouping of 30 second periods stating form a given time. Here it is '2012-01-01 00:00:00'. DATEDIFF counts the number of seconds between time stamp value and stating time. Then its is divided by 30 to get grouping column.
SELECT MIN(TimeColumn) AS TimeGroup, SUM(Quantity) AS TotalQuantity FROM YourTable
GROUP BY (DATEDIFF(ss, TimeColumn, '2012-01-01') / 30)
Here minimum time stamp of each group will output as TimeGroup. But you can use maximum or even grouping column value can be converted to time again for display.
Looking at the above comments, I'm assuming Chris's first scenario is the one you want (all 3 get grouped even though values 1 and 3 are not within 30 seconds of eachother, but are each within 30 seconds of value 2). Also going to assume that each row in your table has some unique ID called 'id'. You can do the following:
Create a new grouping, determining if the preceding row in your partition is more than 30 seconds behind the current row (e.g. determine if you need a new 30 second grouping, or to continue the previous). We'll call that parent_id.
Sum quantity over parent_id (plus any other aggregations)
The code could look like this
select
sub.parent_id,
sub.cusip,
min(sub.timestamp) min_timestamp,
sum(sub.quantity) quantity
from
(
select
base_sub.*,
case
when base_sub.self_parent_id is not null
then base_sub.self_parent_id
else lag(base_sub.self_parent_id) ignore nulls over (
partition by
my_table.cusip
order by
my_table.timestamp,
my_table.id
) parent_id
from
(
select
my_table.id,
my_table.cusip,
my_table.timestamp,
my_table.quantity,
lag(my_table.timestamp) over (
partition by
my_table.cusip
order by
my_table.timestamp,
my_table.id
) previous_timestamp,
case
when datediff(
second,
nvl(previous_timestamp, to_date('1900/01/01', 'yyyy/mm/dd')),
my_table.timestamp) > 30
then my_table.id
else null
end self_parent_id
from
my_table
) base_sub
) sub
group by
sub.time_group_parent_id,
sub.cusip

Sorting twice on same column

I'm having a bit of a weird question, given to me by a client.
He has a list of data, with a date between parentheses like so:
Foo (14/08/2012)
Bar (15/08/2012)
Bar (16/09/2012)
Xyz (20/10/2012)
However, he wants the list to be displayed as follows:
Foo (14/08/2012)
Bar (16/09/2012)
Bar (15/08/2012)
Foot (20/10/2012)
(notice that the second Bar has moved up one position)
So, the logic behind it is, that the list has to be sorted by date ascending, EXCEPT when two rows have the same name ('Bar'). If they have the same name, it must be sorted with the LATEST date at the top, while staying in the other sorting order.
Is this even remotely possible? I've experimented with a lot of ORDER BY clauses, but couldn't find the right one. Does anyone have an idea?
I should have specified that this data comes from a table in a sql server database (the Name and the date are in two different columns). So I'm looking for a SQL-query that can do the sorting I want.
(I've dumbed this example down quite a bit, so if you need more context, don't hesitate to ask)
This works, I think
declare #t table (data varchar(50), date datetime)
insert #t
values
('Foo','2012-08-14'),
('Bar','2012-08-15'),
('Bar','2012-09-16'),
('Xyz','2012-10-20')
select t.*
from #t t
inner join (select data, COUNT(*) cg, MAX(date) as mg from #t group by data) tc
on t.data = tc.data
order by case when cg>1 then mg else date end, date desc
produces
data date
---------- -----------------------
Foo 2012-08-14 00:00:00.000
Bar 2012-09-16 00:00:00.000
Bar 2012-08-15 00:00:00.000
Xyz 2012-10-20 00:00:00.000
A way with better performance than any of the other posted answers is to just do it entirely with an ORDER BY and not a JOIN or using CTE:
DECLARE #t TABLE (myData varchar(50), myDate datetime)
INSERT INTO #t VALUES
('Foo','2012-08-14'),
('Bar','2012-08-15'),
('Bar','2012-09-16'),
('Xyz','2012-10-20')
SELECT *
FROM #t t1
ORDER BY (SELECT MIN(t2.myDate) FROM #t t2 WHERE t2.myData = t1.myData), T1.myDate DESC
This does exactly what you request and will work with any indexes and much better with larger amounts of data than any of the other answers.
Additionally it's much more clear what you're actually trying to do here, rather than masking the real logic with the complexity of a join and checking the count of joined items.
This one uses analytic functions to perform the sort, it only requires one SELECT from your table.
The inner query finds gaps, where the name changes. These gaps are used to identify groups in the next query, and the outer query does the final sorting by these groups.
I have tried it here (SQL Fiddle) with extended test-data.
SELECT name, dat
FROM (
SELECT name, dat, SUM(gap) over(ORDER BY dat, name) AS grp
FROM (
SELECT name, dat,
CASE WHEN LAG(name) OVER (ORDER BY dat, name) = name THEN 0 ELSE 1 END AS gap
FROM t
) x
) y
ORDER BY grp, dat DESC
Extended test-data
('Bar','2012-08-12'),
('Bar','2012-08-11'),
('Foo','2012-08-14'),
('Bar','2012-08-15'),
('Bar','2012-08-16'),
('Bar','2012-09-17'),
('Xyz','2012-10-20')
Result
Bar 2012-08-12
Bar 2012-08-11
Foo 2012-08-14
Bar 2012-09-17
Bar 2012-08-16
Bar 2012-08-15
Xyz 2012-10-20
I think that this works, including the case I asked about in the comments:
declare #t table (data varchar(50), [date] datetime)
insert #t
values
('Foo','20120814'),
('Bar','20120815'),
('Bar','20120916'),
('Xyz','20121020')
; With OuterSort as (
select *,ROW_NUMBER() OVER (ORDER BY [date] asc) as rn from #t
)
--Now we need to find contiguous ranges of the same data value, and the min and max row number for such a range
, Islands as (
select data,rn as rnMin,rn as rnMax from OuterSort os where not exists (select * from OuterSort os2 where os2.data = os.data and os2.rn = os.rn - 1)
union all
select i.data,rnMin,os.rn
from
Islands i
inner join
OuterSort os
on
i.data = os.data and
i.rnMax = os.rn-1
), FullIslands as (
select
data,rnMin,MAX(rnMax) as rnMax
from Islands
group by data,rnMin
)
select
*
from
OuterSort os
inner join
FullIslands fi
on
os.rn between fi.rnMin and fi.rnMax
order by
fi.rnMin asc,os.rn desc
It works by first computing the initial ordering in the OuterSort CTE. Then, using two CTEs (Islands and FullIslands), we compute the parts of that ordering in which the same data value appears in adjacent rows. Having done that, we can compute the final ordering by any value that all adjacent values will have (such as the lowest row number of the "island" that they belong to), and then within an "island", we use the reverse of the originally computed sort order.
Note that this may, though, not be too efficient for large data sets. On the sample data it shows up as requiring 4 table scans of the base table, as well as a spool.
Try something like...
ORDER BY CASE date
WHEN '14/08/2012' THEN 1
WHEN '16/09/2012' THEN 2
WHEN '15/08/2012' THEN 3
WHEN '20/10/2012' THEN 4
END
In MySQL, you can do:
ORDER BY FIELD(date, '14/08/2012', '16/09/2012', '15/08/2012', '20/10/2012')
In Postgres, you can create a function FIELD and do:
CREATE OR REPLACE FUNCTION field(anyelement, anyarray) RETURNS numeric AS $$
SELECT
COALESCE((SELECT i
FROM generate_series(1, array_upper($2, 1)) gs(i)
WHERE $2[i] = $1),
0);
$$ LANGUAGE SQL STABLE
If you do not want to use the CASE, you can try to find an implementation of the FIELD function to SQL Server.