Google Big Query : Window Function Row Wise Cumulative Sum Across Columns - sql

I am looking to calculate cumulative sum across columns in Google Big Query.
Assume there are five columns (NAME,A,B,C,D) with two rows of integers, for example:
NAME | A | B | C | D
----------------------
Bob | 1 | 2 | 3 | 4
Carl | 5 | 6 | 7 | 8
I am looking for a windowing function or UDF to calculate the cumulative sum across rows to generate this output:
NAME | A | B | C | D
-------------------------
Bob | 1 | 3 | 6 | 10
Carl | 5 | 11 | 18 | 27
Any thoughts or suggestions greatly appreciated!

I think, there are number of reasonable workarounds for your requirements mostly in the area of designing better your table. All really depends on how you input your data and most importantly how than you consume it
Still, if to stay with presented requirements - Below is not exactly what you expect in your question as an output, but might be usefull as an example:
SELECT name, GROUP_CONCAT(STRING(cum)) AS all FROM (
SELECT name,
SUM(INTEGER(num))
OVER(PARTITION BY name
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cum
FROM (
SELECT name, SPLIT(all) AS num FROM (
SELECT name,
CONCAT(STRING(a),',',STRING(b),',',STRING(c),',',STRING(d)) AS all
FROM yourtable
)
)
)
GROUP BY name
Output is:
name all
Bob 1,3,6,10
Carl 5,11,18,26
Depends on how you than consume this data - it still can work for you
Note, not you avoiding now writing something like col1 + col2 + .. + col89 + col90 - but still need to explicitelly mention each column just ones.
in case if you have "luxury" of implementing your requirements outside of GBQ UI, but rather in some Client- you can use BigQuery API to programatically aquire table schema and build on fly your logic/query and than execute it
Take a look at below APIs to start with:
To get table schema - https://cloud.google.com/bigquery/docs/reference/v2/tables/get
To issue query job - https://cloud.google.com/bigquery/docs/reference/v2/jobs/insert

There's no need for a UDF:
SELECT name, a, a+b, a+b+c, a+b+c+d
FROM tab

Related

Recursive self join over file data

I know there are many questions about recursive self joins, but they're mostly in a hierarchical data structure as follows:
ID | Value | Parent id
-----------------------------
But I was wondering if there was a way to do this in a specific case that I have where I don't necessarily have a parent id. My data will look like this when I initially load the file.
ID | Line |
-------------------------
1 | 3,Formula,1,2,3,4,...
2 | *,record,abc,efg,hij,...
3 | ,,1,x,y,z,...
4 | ,,2,q,r,s,...
5 | 3,Formula,5,6,7,8,...
6 | *,record,lmn,opq,rst,...
7 | ,,1,t,u,v,...
8 | ,,2,l,m,n,...
Essentially, its a CSV file where each row in the table is a line in the file. Lines 1 and 5 identify an object header and lines 3, 4, 7, and 8 identify the rows belonging to the object. The object header lines can have only 40 attributes which is why the object is broken up across multiple sections in the CSV file.
What I'd like to do is take the table, separate out the record # column, and join it with itself multiple times so it achieves something like this:
ID | Line |
-------------------------
1 | 3,Formula,1,2,3,4,5,6,7,8,...
2 | *,record,abc,efg,hij,lmn,opq,rst
3 | ,,1,x,y,z,t,u,v,...
4 | ,,2,q,r,s,l,m,n,...
I know its probably possible, I'm just not sure where to start. My initial idea was to create a view that separates out the first and second columns in a view, and use the view as a way of joining in a repeated fashion on those two columns. However, I have some problems:
I don't know how many sections will occur in the file for the same
object
The file can contain other objects as well so joining on the first two columns would be problematic if you have something like
ID | Line |
-------------------------
1 | 3,Formula,1,2,3,4,...
2 | *,record,abc,efg,hij,...
3 | ,,1,x,y,z,...
4 | ,,2,q,r,s,...
5 | 3,Formula,5,6,7,8,...
6 | *,record,lmn,opq,rst,...
7 | ,,1,t,u,v,...
8 | ,,2,l,m,n,...
9 | ,4,Data,1,2,3,4,...
10 | *,record,lmn,opq,rst,...
11 | ,,1,t,u,v,...
In the above case, my plan could join rows from the Data object in row 9 with the first rows of the Formula object by matching the record value of 1.
UPDATE
I know this is somewhat confusing. I tried doing this with C# a while back, but I had to basically write a recursive decent parser to parse the specific file format and it simply took to long because I had to get it in the database afterwards and it was too much for entity framework. It was taking hours just to convert one file since these files are excessively large.
Either way, #Nolan Shang has the closest result to what I want. The only difference is this (sorry for the bad formatting):
+----+------------+------------------------------------------+-----------------------+
| ID | header | x | value
|
+----+------------+------------------------------------------+-----------------------+
| 1 | 3,Formula, | ,1,2,3,4,5,6,7,8 |3,Formula,1,2,3,4,5,6,7,8 |
| 2 | ,, | ,1,x,y,z,t,u,v | ,1,x,y,z,t,u,v |
| 3 | ,, | ,2,q,r,s,l,m,n | ,2,q,r,s,l,m,n |
| 4 | *,record, | ,abc,efg,hij,lmn,opq,rst |*,record,abc,efg,hij,lmn,opq,rst |
| 5 | ,4, | ,Data,1,2,3,4 |,4,Data,1,2,3,4 |
| 6 | *,record, | ,lmn,opq,rst | ,lmn,opq,rst |
| 7 | ,, | ,1,t,u,v | ,1,t,u,v |
+----+------------+------------------------------------------+-----------------------------------------------+
I agree that it would be better to export this to a scripting language and do it there. This will be a lot of work in TSQL.
You've intimated that there are other possible scenarios you haven't shown, so I obviously can't give a comprehensive solution. I'm guessing this isn't something you need to do quickly on a repeated basis. More of a one-time transformation, so performance isn't an issue.
One approach would be to do a LEFT JOIN to a hard-coded table of the possible identifying sub-strings like:
3,Formula,
*,record,
,,1,
,,2,
,4,Data,
Looks like it pretty much has to be human-selected and hard-coded because I can't find a reliable pattern that can be used to SELECT only these sub-strings.
Then you SELECT from this artificially-created table (or derived table, or CTE) and LEFT JOIN to your actual table with a LIKE to get all the rows that use each of these values as their starting substring, strip out the starting characters to get the rest of the string, and use the STUFF..FOR XML trick to build the desired Line.
How you get the ID column depends on what you want, for instance in your second example, I don't know what ID you want for the ,4,Data,... line. Do you want 5 because that's the next number in the results, or do you want 9 because that's the ID of the first occurrance of that sub-string? Code accordingly. If you want 5 it's a ROW_NUMBER(). If you want 9, you can add an ID column to the artificial table you created at the start of this approach.
BTW, there's really nothing recursive about what you need done, so if you're still thinking in those terms, now would be a good time to stop. This is more of a "Group Concatenation" problem.
Here is a sample, but has some different with you need.
It is because I use the value the second comma as group header, so the ,,1 and ,,2 will be treated as same group, if you can use a parent id to indicated a group will be better
DECLARE #testdata TABLE(ID int,Line varchar(8000))
INSERT INTO #testdata
SELECT 1,'3,Formula,1,2,3,4,...' UNION ALL
SELECT 2,'*,record,abc,efg,hij,...' UNION ALL
SELECT 3,',,1,x,y,z,...' UNION ALL
SELECT 4,',,2,q,r,s,...' UNION ALL
SELECT 5,'3,Formula,5,6,7,8,...' UNION ALL
SELECT 6,'*,record,lmn,opq,rst,...' UNION ALL
SELECT 7,',,1,t,u,v,...' UNION ALL
SELECT 8,',,2,l,m,n,...' UNION ALL
SELECT 9,',4,Data,1,2,3,4,...' UNION ALL
SELECT 10,'*,record,lmn,opq,rst,...' UNION ALL
SELECT 11,',,1,t,u,v,...'
;WITH t AS(
SELECT *,REPLACE(SUBSTRING(t.Line,LEN(c.header)+1,LEN(t.Line)),',...','') AS data
FROM #testdata AS t
CROSS APPLY(VALUES(LEFT(t.Line,CHARINDEX(',',t.Line, CHARINDEX(',',t.Line)+1 )))) c(header)
)
SELECT MIN(ID) AS ID,t.header,c.x,t.header+STUFF(c.x,1,1,'') AS value
FROM t
OUTER APPLY(SELECT ','+tb.data FROM t AS tb WHERE tb.header=t.header FOR XML PATH('') ) c(x)
GROUP BY t.header,c.x
+----+------------+------------------------------------------+-----------------------------------------------+
| ID | header | x | value |
+----+------------+------------------------------------------+-----------------------------------------------+
| 1 | 3,Formula, | ,1,2,3,4,5,6,7,8 | 3,Formula,1,2,3,4,5,6,7,8 |
| 3 | ,, | ,1,x,y,z,2,q,r,s,1,t,u,v,2,l,m,n,1,t,u,v | ,,1,x,y,z,2,q,r,s,1,t,u,v,2,l,m,n,1,t,u,v |
| 2 | *,record, | ,abc,efg,hij,lmn,opq,rst,lmn,opq,rst | *,record,abc,efg,hij,lmn,opq,rst,lmn,opq,rst |
| 9 | ,4, | ,Data,1,2,3,4 | ,4,Data,1,2,3,4 |
+----+------------+------------------------------------------+-----------------------------------------------+

Find spectators that have seen the same shows (match multiple rows for each)

For an assignment I have to write several SQL queries for a database stored in a PostgreSQL server running PostgreSQL 9.3.0. However, I find myself blocked with last query. The database models a reservation system for an opera house. The query is about associating the a spectator the other spectators that assist to the same events every time.
The model looks like this:
Reservations table
id_res | create_date | tickets_presented | id_show | id_spectator | price | category
-------+---------------------+---------------------+---------+--------------+-------+----------
1 | 2015-08-05 17:45:03 | | 1 | 1 | 195 | 1
2 | 2014-03-15 14:51:08 | 2014-11-30 14:17:00 | 11 | 1 | 150 | 2
Spectators table
id_spectator | last_name | first_name | email | create_time | age
---------------+------------+------------+----------------------------------------+---------------------+-----
1 | gonzalez | colin | colin.gonzalez#gmail.com | 2014-03-15 14:21:30 | 22
2 | bequet | camille | bequet.camille#gmail.com | 2014-12-10 15:22:31 | 22
Shows table
id_show | name | kind | presentation_date | start_time | end_time | id_season | capacity_cat1 | capacity_cat2 | capacity_cat3 | price_cat1 | price_cat2 | price_cat3
---------+------------------------+--------+-------------------+------------+----------+-----------+---------------+---------------+---------------+------------+------------+------------
1 | madama butterfly | opera | 2015-09-05 | 19:30:00 | 21:30:00 | 2 | 315 | 630 | 945 | 195 | 150 | 100
2 | don giovanni | opera | 2015-09-12 | 19:30:00 | 21:45:00 | 2 | 315 | 630 | 945 | 195 | 150 | 100
So far I've started by writing a query to get the id of the spectator and the date of the show he's attending to, the query looks like this.
SELECT Reservations.id_spectator, Shows.presentation_date
FROM Reservations
LEFT JOIN Shows ON Reservations.id_show = Shows.id_show;
Could someone help me understand better the problem and hint me towards finding a solution. Thanks in advance.
So the result I'm expecting should be something like this
id_spectator | other_id_spectators
-------------+--------------------
1| 2,3
Meaning that every time spectator with id 1 went to a show, spectators 2 and 3 did too.
Note based on comments: Wanted to make clear that this answer may be of limited use as it was answered in the context of SQL-Server (tag was present at the time)
There is probably a better way to do it, but you could do it with the 'stuff 'function. The only drawback here is that, since your ids are ints, placing a comma between values will involve a work around (would need to be a string). Below is the method I can think of using a work around.
SELECT [id_spectator], [id_show]
, STUFF((SELECT ',' + CAST(A.[id_spectator] as NVARCHAR(10))
FROM reservations A
Where A.[id_show]=B.[id_show] AND a.[id_spectator] != b.[id_spectator] FOR XML PATH('')),1,1,'') As [other_id_spectators]
From reservations B
Group By [id_spectator], [id_show]
This will show you all other spectators that attended the same shows.
Meaning that every time spectator with id 1 went to a show, spectators 2 and 3 did too.
In other words, you want a list of ...
all spectators that have seen all the shows that a given spectator has seen (and possibly more than the given one)
This is a special case of relational division. We have assembled an arsenal of basic techniques here:
How to filter SQL results in a has-many-through relation
It is special because the list of shows each spectator has to have attended is dynamically determined by the given prime spectator.
Assuming that (d_spectator, id_show) is unique in reservations, which has not been clarified.
A UNIQUE constraint on those two columns (in that order) also provides the most important index.
For best performance in query 2 and 3 below also create an index with leading id_show.
1. Brute force
The primitive approach would be to form a sorted array of shows the given user has seen and compare the same array of others:
SELECT 1 AS id_spectator, array_agg(sub.id_spectator) AS id_other_spectators
FROM (
SELECT id_spectator
FROM reservations r
WHERE id_spectator <> 1
GROUP BY 1
HAVING array_agg(id_show ORDER BY id_show)
#> (SELECT array_agg(id_show ORDER BY id_show)
FROM reservations
WHERE id_spectator = 1)
) sub;
But this is potentially very expensive for big tables. The whole table hast to be processes, and in a rather expensive way, too.
2. Smarter
Use a CTE to determine relevant shows, then only consider those
WITH shows AS ( -- all shows of id 1; 1 row per show
SELECT id_spectator, id_show
FROM reservations
WHERE id_spectator = 1 -- your prime spectator here
)
SELECT sub.id_spectator, array_agg(sub.other) AS id_other_spectators
FROM (
SELECT s.id_spectator, r.id_spectator AS other
FROM shows s
JOIN reservations r USING (id_show)
WHERE r.id_spectator <> s.id_spectator
GROUP BY 1,2
HAVING count(*) = (SELECT count(*) FROM shows)
) sub
GROUP BY 1;
#> is the "contains2 operator for arrays - so we get all spectators that have at least seen the same shows.
Faster than 1. because only relevant shows are considered.
3. Real smart
To also exclude spectators that are not going to qualify early from the query, use a recursive CTE:
WITH RECURSIVE shows AS ( -- produces exactly 1 row
SELECT id_spectator, array_agg(id_show) AS shows, count(*) AS ct
FROM reservations
WHERE id_spectator = 1 -- your prime spectator here
GROUP BY 1
)
, cte AS (
SELECT r.id_spectator, 1 AS idx
FROM shows s
JOIN reservations r ON r.id_show = s.shows[1]
WHERE r.id_spectator <> s.id_spectator
UNION ALL
SELECT r.id_spectator, idx + 1
FROM cte c
JOIN reservations r USING (id_spectator)
JOIN shows s ON s.shows[c.idx + 1] = r.id_show
)
SELECT s.id_spectator, array_agg(c.id_spectator) AS id_other_spectators
FROM shows s
JOIN cte c ON c.idx = s.ct -- has an entry for every show
GROUP BY 1;
Note that the first CTE is non-recursive. Only the second part is recursive (iterative really).
This should be fastest for small selections from big tables. Row that don't qualify are excluded early. the two indices I mentioned are essential.
SQL Fiddle demonstrating all three.
It sounds like you have one half of the total question--determining which id_shows a particular id_spectator attended.
What you want to ask yourself is how you can determine which id_spectators attended an id_show, given an id_show. Once you have that, combine the two answers to get the full result.
So the final answer I got, looks like this :
SELECT id_spectator, id_show,(
SELECT string_agg(to_char(A.id_spectator, '999'), ',')
FROM Reservations A
WHERE A.id_show=B.id_show
) AS other_id_spectators
FROM Reservations B
GROUP By id_spectator, id_show
ORDER BY id_spectator ASC;
Which prints something like this:
id_spectator | id_show | other_id_spectators
-------------+---------+---------------------
1 | 1 | 1, 2, 9
1 | 14 | 1, 2
Which suits my needs, however if you have any improvements to offer, please share :) Thanks again everybody!

SQL Server 2008 pivot query gone wrong with column name

I have some problems regarding a pivot query. I am new to this. So look for something in the internet so I found dozens of them. So I decided to follow this Link. Been practice but seems like I ran into some obvious error.
My code is:
select
risk, [Quick] AS Quick, [Brown] AS Brown, [Fox] AS Fox
from
(select risk, site
from tst) as ps
PIVOT
(
count(risk)
for site in ([Brown], [Brown], [Fox])
) AS pvt
But it is throwing an error:
Invalid column name 'risk'.
Basically I want to have an output like this:
|Foo | Quick | Brown | Fox |
| 1 | 10 | 3 | 2 |
| 2 | 5 | 4 | 4 |
| 3 | 4 | 1 | 5 |
| 4 | 2 | 3 | 7 |
| 5 | 3 | 2 | 1 |
Something like that. Just counting how many there is in a specific number. Any help would be much appreciated. Thanks
The problem with your existing query is you are using the column risk in your final select list as well as inside of the aggregate function. Once you've counted the risk values for each site this is not available to display.
To get around this you can add a second version of the risk column to your subquery similar to the following. You then count this other column of risk and display one in the final select:
select risk, [ADAB] AS ADAB, [Bahrain] AS Bahrain, [Thumrait] AS Thumrait
from
(
select risk, piv_risk = risk, site
from qcqcif
) as ps
PIVOT
(
count(piv_risk)
for site in ([ADAB], [Bahrain], [Thumrait])
) AS pvt;
See SQL Fiddle with Demo

Access 2007 select first value of query results

I am running into a rather annoying thingy in Access (2007) and I am not sure if this is a feature or if I am asking for the impossible.
Although the actual database structure is more complex, my problem boils down to this:
I have a table with data about Units for specific years. This data comes from different sources and might overlap.
Unit | IYR | X1 | Source |
-----------------------------
A | 2009 | 55 | 1 |
A | 2010 | 80 | 1 |
A | 2010 | 101 | 2 |
A | 2010 | 150 | 3 |
A | 2011 | 90 | 1 |
...
Now I would like the user to select certain sources, order them by priority and then extract one data value for each year.
For example, if the user selects source 1, 2 and 3 and orders them by (3, 1, 2), then I would like the following result:
Unit | IYR | X1 | Source |
-----------------------------
A | 2009 | 55 | 1 |
A | 2010 | 150 | 3 |
A | 2011 | 90 | 1 |
I am able to order the initial table, based on a specific order. I do this with the following query
SELECT Unit, IYR, X1, Source
FROM TestTable
WHERE Source In (1,2,3)
ORDER BY Unit, IYR,
IIf(Source=3,1,IIf(Source=1,2,IIf(Source=2,3,4)))
This gives me the following intermediate result:
Unit | IYR | X1 | Source |
-----------------------------
A | 2009 | 55 | 1 |
A | 2010 | 150 | 3 |
A | 2010 | 80 | 1 |
A | 2010 | 101 | 2 |
A | 2011 | 90 | 1 |
Next step is to only get the first value of each year. I was thinking to use the following query:
SELECT X.Unit, X.IYR, first(X.X1) as FirstX1
FROM (...) AS X
GROUP BY X.Unit, X.IYR
Where (…) is the above query.
Now Access goes bananas. Whatever order I give to the intermediate results, the result of this query is.
Unit | IYR | X1 |
--------------------
A | 2009 | 55 |
A | 2010 | 80 |
A | 2011 | 90 |
In other words, for year 2010 it shows the value of source 1 instead of 3. It seems that Access does not care about the ordering of the nested query when it applies the FIRST() function and sticks to the original ordering of the data.
Is this a feature of Access or is there a different way of achieving the desired results?
Ps: Next step would be to use a self join to add the source column to the results again, but I first need to resolve above problem.
Rather than use first it may be better to determine the MIN Priority and then join back e.g.
SELECT
t.UNIT,
t.IYR,
t.X1,
t.Source ,
t.PrioritySource
FROM
(SELECT
Unit,
IYR,
X1,
Source,
SWITCH ( [Source]=3, 1,
[Source]=1, 2,
[Source]=2, 3) as PrioritySource
FROM
TestTable
WHERE
Source In (1,2,3)
) as t
INNER JOIN
(SELECT
Unit,
IYR,
MIN(SWITCH ( [Source]=3, 1,
[Source]=1, 2,
[Source]=2, 3)) as PrioritySource
FROM
TestTable
WHERE
Source In (1,2,3)
GROUP BY
Unit,
IYR ) as MinPriortiy
ON t.Unit = MinPriortiy.Unit and
t.IYR = MinPriortiy.IYR and
t.PrioritySource = MinPriortiy.PrioritySource
which will produce this result (Note I include Source and priority source for demonstration purposes only)
UNIT | IYR | X1 | Source | PrioritySource
----------------------------------------------
A | 2009 | 55 | 1 | 2
A | 2010 | 150 | 3 | 1
A | 2011 | 90 | 1 | 2
Note the first subquery is to handle the fact that Access won't let you join on a Switch
Yes, FIRST() does use an arbitrary ordering. From the Access Help:
These functions return the value of a specified field in the first or
last record, respectively, of the result set returned by a query. If
the query does not include an ORDER BY clause, the values returned by
these functions will be arbitrary because records are usually returned
in no particular order.
I don't know whether FROM (...) AS X means you are using an ORDER BY inline (assuming that is actually possible) or if you are using a VIEW ('stored Query object') here but either way I assume the ORDER BY is being disregarded (because an ORDER BY should only apply to the final result).
The alternative is to use MIN() (or possibly MAX()).
This is the most concise way I have found to write such queries in Access that require pulling back all columns that correspond to the first row in a group of records that are ordered in a particular way.
First, I added a UniqueID to your table. In this case, it's just an AutoNumber field. You may already have a unique value in your table, in which case you can use that.
This will choose the row with a Source 3 first, then Source 1, then Source 2. If there is a tie, it picks the one with the higher X1 value. If there is a further tie, it is broken by the UniqueID value:
SELECT t.* INTO [Chosen Rows]
FROM TestTable AS t
WHERE t.UniqueID=
(SELECT TOP 1 [UniqueID] FROM [TestTable]
WHERE t.IYR=IYR ORDER BY Choose([Source],2,3,1), X1 DESC, UniqueID)
This yields:
Unit IYR X1 Source UniqueID
A 2009 55 1 1
A 2010 150 3 4
A 2011 90 1 5
I recommend (1) you create an index on the IYR field -- this will dramatically increase your performance for this type of query, and (2) if you have a lot (>~100K) records, this isn't the best choice. I find it works quite well for tables in the 1-70K range. For larger datasets, I like to use my GroupIncrement function to partition each group (similar to SQL Server's ROW_NUMBER() OVER statement).
The Choose() function is a VBA function and may not be clear here. In your case, it sounds like there is some interactivity required. For that, you could create a second table called "Choices", like so:
Rank Choice
1 3
2 1
3 2
Then, you could substitute the following:
SELECT t.* INTO [Chosen Rows]
FROM TestTable AS t
WHERE t.UniqueID=(SELECT TOP 1 [UniqueID] FROM
[TestTable] t2 INNER JOIN [Choices] c
ON t2.Source=c.Choice
WHERE t.IYR=t2.IYR ORDER BY c.[Rank], t2.X1 DESC, t2.UniqueID);
Indexing Source on TestTable and Choice on the Choices table may be helpful here, too, depending on the number of choices required.
Q:
Can you get this to work without the need for surrogate key? For
example what if the unique key is the composite of
{Unit,IYR,X1,Source}
A:
If you have a compound key, you can do it like this-- however I think that if you have a large dataset, it will totally kill the performance of the query. It may help to index all four columns, but I can't say for sure because I don't regularly use this method.
SELECT t.* INTO [Chosen Rows]
FROM TestTable AS t
WHERE t.Unit & t.IYR & t.X1 & t.Source =
(SELECT TOP 1 Unit & IYR & X1 & Source FROM [TestTable]
WHERE t.IYR=IYR ORDER BY Choose([Source],2,3,1), X1 DESC, Unit, IYR)
In certain cases, you may have to coalesce some of the individual parts of the key as follows (though Access generally will coalesce values automatically):
t.Unit & CStr(t.IYR) & CStr(t.X1) & CStr(t.Source)
You could also use a query in your FROM statements instead of the actual table. The query itself would build a composite of the four fields used in the key, and then you'd use the new key name in the WHERE clause of the top SELECT statement, and in the SELECT TOP 1 [key] of the subquery.
In general, though, I will either: (a) create a new table with an AutoNumber field, (b) add an AutoNumber field, (c) add an integer and populate it with a unique number using VBA - this is useful when you get a MaxLocks error when trying to add an AutoNumber, or (d) use an already indexed unique key.

Querying sequences of rows in SQL

Suppose I am storing events associated with users in a table as follows (with dt standing in for the timestamp of the event):
| dt | user | event |
| 1 | 1 | A |
| 2 | 1 | D |
| 3 | 1 | B |
| 4 | 1 | C |
| 5 | 1 | B |
| 6 | 2 | B |
| 7 | 2 | B |
| 8 | 2 | A |
| 9 | 2 | A |
| 10 | 2 | C |
Such that we could say:
user 1 has an event-sequence of ADBCB
user 2 has event-sequence BBAAC
The types of questions I would want to answer about these users are very easy to express as regular expresions on the event-sequences, e.g. "which users have an event-sequence matching A.*B?" or "which users have an event-sequence matching A[^C]*B[^C]*D?" etc.
What would be a good SQL technique or operator I could use to answer similar queries over this table structure?
Is there a way to efficiently/dynamically generate a table of user-to-event-sequence which could then be queried with regex?
I am currently looking at using Postgres, but I am curious to know if any of the bigger DBMS's like SQLServer or Oracle have specialized operators for this as well.
With Postgres 9.x this is actually quite easy:
select userid,
string_agg(event, '' order by dt) as event_sequence
from events
group by userid;
Using that result you can now apply a regular expression on the event_sequence:
select *
from (
select userid,
string_agg(event, '' order by dt) as event_sequence
from events
group by userid
) t
where event_sequence ~ 'A.*B'
With Postgres 8.x you need to find a replacement for the string_agg() function (just google for it, there are a lot of examples out there) and you need a sub-select to ensure the ordering of the aggregate as 8.x does support an order by in an aggregate function.
I'm not at a computer to write code for this answer, but here's how I would go about a RegEx-based solution in SQL Server:
Build a string from the resultset. Something like http://blog.sqlauthority.com/2009/11/25/sql-server-comma-separated-values-csv-from-table-column/ should work if you omit the comma
Run your RegEx match against the resulting string. Unfortunately, SQL Server does not provide this functionality natively, however, you can use a CLR function for this purpose as described at http://www.ideaexcursion.com/2009/08/18/sql-server-regular-expression-clr-udf/
This should ultimately provide you with the functionality in SQL Server that your original question requests, however, if you're analyzing a very large dataset, this could be quite slow and there may be better ways to accomplish what you're looking for.
For Oracle (version 11g R2):
By chance if you are using Oracle DB 11g R2, take look at listagg. The below code should work, but I haven't tested. The point is: you can use listagg.
SQL> select user,
2 listagg( event, '' )
3 within group (order by dt) events
4 from users
5 group by user
6 order by dt
7 /
USER EVENTS
--------- --------------------
1 ADBCB
2 BBAAC
In prior versions you can do with CONNECT BY clause. More details on listagg.