i'm looking for a way to, on my query, dynamically set the beginning of the window function on Sql Server using ROWS BETWEEN.
Something like:
SUM(field) OVER(ORDER BY field2 ROWS BETWEEN field3 PRECEDING AND CURRENT ROW)
field3 holds the amount of items (via group by from a CTE) that represent a group.
Is that possible or should i try a different approach?
>> EDIT
My query is too big and messy to share here, but let me try to explain what i need. It's from a report builder which allows users to create custom formulas, like "emplyoees/10". This also allows the user to simply input a formula like "12" and i need to calculate subtotals and the grand total for them. When using a field, like "employees", everything works fine. But for constant values i can't sum the values without rewriting a lot of stuff (which i'm trying to avoid).
So, consider a CTE called "aggregator" and the following query:
SELECT
*,
"employees"/10 as "ten_percent"
12 as "twelve"
FROM aggregator
This query returns this output:
row_type counter company_name department_name employees ten_percent twelve
data 1 A A1 10 1 12
data 1 A A2 15 1,5 12
data 1 A A3 10 1 12
subtotal 3 A 35 3,5 12
data 1 B B1 10 1 12
subtotal 1 B 10 1 12
total 4 45 4,5 12
As you can see, the values fot "twelve" are wrong for subtotal and total row types. I'm trying to solve this without changing the CTE.
ROLLUP won't work because i already have the sum for other columns.
I tried this (i ommited "row_type_sort" on the table above, it defines the sorting):
CASE
WHEN row_type = 'data' THEN
MAX(aggregator.[twelve])
ELSE
SUM(SUM(aggregator.[twelve]))
OVER (ORDER BY "row_type_sort" ROWS BETWEEN unbounded PRECEDING AND CURRENT ROW)
END AS "twelve"
This would work OK if i could change "unbounded" by the value of column "counter", which was my original question.
LAG/LEAD wasn't helpful neither.
I'm out of ideas. Is it possible to achieve what i need only by changing this part of the query, or the result of the CTE should be changed as well?
Thanks
Related
I have a table that looks like this with three columns From, To, and Symbol:
From To Symbol
0 2 dog
2 5 dog
5 9 cat
9 15 cat
15 20 dog
20 40 dog
40 45 dog
I was trying to write an SQL query that groups records in a way that produces the following result:
From To Symbol
0 5 dog
5 15 cat
15 45 dog
That is, if the From and To values are continuous for the same Symbol, one result record is created with the smallest From and the largest To values and the Symbol. In the above example table, since the second record has a value of 2 in the To column which is not the same as the From value in the next record with the same Symbol (15, 20, dog), two result records are created for the same Symbol (dog).
I have tried to join the table to itself, then group by. But I could not figure out how exactly that can be done. I have to do this in Microsoft Access. Any help would be greatly appreciated. Thanks!
Assuming the values have no overlaps and that gaps separate values, you can do this in MS Access with a trick. You need to identify the adjacent symbols that are the same. Well, you can identify them by counting the number of previous rows with different symbols (using a subquery). Once you have this information, the rest is aggregation:
select symbol, min(from) as from, max(to) as to
from (select t.*,
(select count(*)
from t as t2
where t2.from < t.from and t2.symbol <> t.symbol
) as grp
from t
) t
group by symbol, grp;
Gaps would make this problem much harder in MS Access.
Note: Don't use reserved words or keywords for column names. This code uses the names supplied in the question, but doesn't bother to escape them. I think that just makes it harder to understand the query.
I need some advice in google query language.
I want to count rows depending on date and a condition. But if the condition is not met, it should return 0.
What I'm trying to achieve:
Date Starts
05.09.2018 0
06.09.2018 3
07.09.2018 0
What I get:
Date Starts
06.09.2018 3
The query looks like =Query(Test!$A2:P; "select P, count(B) where (B contains 'starts') group by P label count(B) 'Starts'")
P contains ascending datevalues and B an event (like start in this case).
How can I force output a 0 for the dates with no entry containing "start"?
The main point is to get all needed data in one table in ascending order. But this is only working, if every day has an entry. If there is no entry for a day, the results for "start" do not match the datevalue in column A. 3 in column D would be in the first row of the table then.
I need it like this:
A B C D
Date Logins Sessions Starts
05.09.2018 1 2 0
06.09.2018 3 4 3
07.09.2018 4 5 0
Maybe this is easy to fix, but I don't see it.
Thanks in advance!
You can do some pre-processing before the query. Ex: check if column B contains 'start' with regexmatch and use a double unary (--) to force the boolean values into 1's and 0's. The use query to sum.
=Query(Arrayformula({--regexmatch(Test!$B2:B; "start")\ Test!$A2:P}); "select Col17, sum(Col1) where Col17 is not null group by Col17 label sum(Col1) 'Starts'")
Change ranges to suit.
I have some data like this as shown below:
Acc_Id || Row_No
1 1
2 1
2 2
2 3
3 1
3 2
3 3
3 4
and I need a query to get the results as shown below:
Acc_Id || Row_No
1 1
2 3
3 4
Please consider that I'm a beginner in SQL.
I assume you want the Count of the row
SELECT Acc_Id, COUNT(*)
FROM Table
GROUP BY Acc_Id
Try this:
select Acc_Id, MAX(Row_No)
from table
group by Acc_Id
As a beginner then this is your first exposure to aggregation and grouping. You may want to look at the documentation on group by now that this problem has motivated your interest in a solutions. Grouping operates by looking at rows with common column values, that you specify, and collapsing them into a single row which represents the group. In your case values in Acc_Id are the names for your groups.
The other answers are both correct in the the final two columns are going to be equivalent with your data.
select Acc_Id, count(*), max(Row_No)
from T
group by Acc_Id;
If you have gaps in the numbering then they won't be the same. You'll have to decide whether you're actually looking for a count of rows of a maximum of a value within a column. At this point you can also consider a number of other aggregate functions that will be useful to you in the future. (Note that the actual values here are pretty much meaningless in this context.)
select Acc_Id, min(Row_No), sum(Row_No), avg(Row_No)
from T
group by Acc_Id;
I have a requirement wherein i need to find the record number of the records that are returned from the resultset. I know that i can use ROWNUM to get the record number from the resultset but my issue is slightly different. below are the details
Table : ProcessSummary
Columns:
PS_PK ProcessId StepId AsscoiateId ProcessName AssetAmount
145 25 50 Process1 3,500.00
267 26 45 Process2 4,400.00
356 27 70 Process3 2,400.00
456 28 80 90 Process4 780.00
556 29 56 67 Process5 4,500.00
656 45 70 Process6 6,000.00
789 31 75 Process7 8,000.00
Now what i need to do is fetch all the records from the ProcessSummary Table when either of ProcessId OR StepId OR AssociateId is NULL. I wrote the below query
select * from ProcessSummary where ProcessId IS NULL OR StepId IS NULL OR AsscoiateId IS NULL
As expected i got 1st, 2nd, 3rd, 6th and 7th records in the resultset that got returned.
Now what i need is to get the records numbers 1,2,3,6,7. I tried to use the ROWNUM as below but i got the values of 1,2,3,4,5 and not 1,2,3,6,7.
select ROWNUM from ProcessSummary where ProcessId IS NULL OR StepId IS NULL OR AsscoiateId IS NULL
Is it possible to get the ROWNUM values in the sequence that i want and if yes then can you please let me know how can i do this. Also if ROWNUM cannot be used then what would be the other option that i can use to get the result in the form that i want.
Any help would be greately appericiated as i could not find much on the net or SO regarding this sort of requirement.
Thanks
Vikeng21
rownum is an internal numbering that gives you a row number based on the current query results only, so that numbering is not tied to a specific record, and it will change when you change the data or the query.
But the numbering you ask for is already in your table. It looks like you just need to SELECT PS_PK .. instead. PS_PK is the field in your table that contains the actual number you want.
You can generate a numbering using an analytical function, and then filter that query. You need some fields to order by, though. In this case I've chosen PS_PK, but it can be another field, like ProcessName or a combination of other fields as well.
select
*
from
(select
dense_rank() over (order by PS_PK) as RANKING,
p.*
from
ProcessSummary p)
where
ProcessId IS NULL OR StepId IS NULL OR AsscoiateId IS NULL
So, in this query, first a numbering is calculated for each row that is returned from the inner query. The numbering is returned as the field RANKING. And then the other query filters further, but still will return the field RANKING with the original numbering.
Instead of dense_rank there is also rank and row_number. The differences are subtle, but you can just experiment and read some docs here and here to learn about the differences and see which one fits you best.
Note that this might slow down your query, because the inner query first generates a number for each row in the table (there is no filtering on that level now).
I Have an SQL query giving me X results, I want the query output to have a coulmn called
count making the query somthing like this:
count id section
1 15 7
2 3 2
3 54 1
4 7 4
How can I make this happen?
So in your example, "count" is the derived sequence number? I don't see what pattern is used to determine the count must be 1 for id=15 and 2 for id=3.
count id section
1 15 7
2 3 2
3 54 1
4 7 4
If id contained unique values, and you order by id you could have this:
count id section
1 3 2
2 7 4
3 15 7
4 54 1
Looks to me like mikeY's DSum approach could work. Or you could use a different approach to a ranking query as Allen Browne described at this page
Edit: You could use DCount instead of DSum. I don't know how the speed would compare between the two, but DCount avoids creating a field in the table simply to store a 1 for each row.
DCount("*","YourTableName","id<=" & [id]) AS counter
Whether you go with DCount or DSum, the counter values can include duplicates if the id values are not unique. If id is a primary key, no worries.
I frankly don't understand what it is you want, but if all you want is a sequence number displayed on your form, you can use a control bound to the form's CurrentRecord property. A control with the ControlSource =CurrentRecord will have an always-accurate "record number" that is in sequence, and that will update when the form's Recordsource changes (which may or may not be desirable).
You can then use that number to navigate around the form, if you like.
But this may not be anything like what you're looking for -- I simply can't tell from the question you've posted and the "clarifications" in comments.
The only trick I have seen is if you have a sequential id field, you can create a new field in which the value for each record is 1. Then you do a running sum of that field.
Add to your query
DSum("[New field with 1 in it]","[Table Name]","[ID field]<=" & [ID Field])
as counterthing
That should produce a sequential count in Access which is what I think you want.
HTH.
(Stolen from Rob Mills here:
http://www.access-programmers.co.uk/forums/showthread.php?p=160386)
Alright, I guess this comes close enough to constitute an answer: the following link specifies two approaches: http://www.techrepublic.com/blog/microsoft-office/an-access-query-that-returns-every-nth-record/
The first approach assumes that you have an ID value and uses DCount (similar to #mikeY's solution).
The second approach assumes you're OK creating a VBA function that will run once for EACH record in the recordset, and will need to be manually reset (with some VBA) every time you want to run the count - because it uses a "static" value to run its counter.
As long as you have reasonable numbers (hundreds, not thousands) or records, the second approach looks like the easiest/most powerful to me.
This function can be called from each record if available from a module.
Example: incrementingCounterTimeFlaged(10,[anyField]) should provide your query rows an int incrementing from 0.
'provides incrementing int values 0 to n
'resets to 0 some seconds after first call
Function incrementingCounterTimeFlaged(resetAfterSeconds As Integer,anyfield as variant) As Integer
Static resetAt As Date
Static i As Integer
'if reset date < now() set the flag and return 0
If DateDiff("s", resetAt, Now()) > 0 Then
resetAt = DateAdd("s", resetAfterSeconds, Now())
i = 0
incrementingCounterTimeFlaged = i
'if reset date > now increments and returns
Else
i = i + 1
incrementingCounterTimeFlaged = i
End If
End Function
autoincrement in SQL
SELECT (Select COUNT(*) FROM table A where A.id<=b.id),B.id,B.Section FROM table AS B ORDER BY B.ID Asc
You can use ROW_NUMBER() which is in SQL Server 2008
SELECT ROW_NUMBER() OVER (ORDER By ID DESC) RowNum,
ID,
Section
FROM myTable
Then RowNum displays sequence of row numbers.