从周日开始按周分组

Question

I have a table with eventdatetime , userid etc. The data is inserted in the table daily.

For the report , I need to give count of userid , projectid grouped by week : Tue-Mon for a month range at a time.

I need help on grouping the data by week for month. I'm using Oracle.

select count(distinct( table1.projectid))as Projects, count(distinct( table2.userid)) as Users,?? 
from table1
join table 2
on table1.a= table2.a
where table1.e='1'
and  table1.eventdatetime between sysdate-30 and sysdate-1  

group by ??

I want the output to be grouped by week like : WeekBegin 2013-04-14

2013-04-21

Answer 1

http://www.techonthenet.com/oracle/functions/to_char.php Use the To_Char function with IW to get the week. Then you can GROUP BY that IW value.

Note that the date the Oracle week starts on is dependent on the language settings of the database. Some countries start on Sunday and some Monday. You'll have to look at your settings to see. If it already starts on Sunday, then you're in luck!

Answer 2

if the example you have posted is your work in progress version - before worrying about getting the days of the week in you should look into getting the basics of the query right

you are selecting e.projectid and u.userid but you haven't got any tables named e or u in your query - it looks like you want to alias them as e and u?

the where clause of your query is also looking for the table e which isn't present

in that case you should change

from table1 
join table2
on table1.a= table2.a

to

from table1 e        -- select from table1 using alias e
  join table2 u      -- join table2 using alias u
    on ( e.a = u.a ) -- joining on column a from table1 (e) = a from table2 (u)

once you have replaced the a's in the on section with the column names you want to join using it might well run after you remove the last column ", ??" from the select - perhaps something along these lines

select
  count (e.projectid) PROJECTS,
  count (u.userid) USERS

from table1 e
  join table2 u
    on ( e.a = u.a )

where e.FILTERING_COLUMN = '1'
  and e.eventdatetime >= sysdate-30

note that as sysdate is the current time on the server (depending on localisation and session settings) you can use greater than sysdate-30 instead of between which may well be give the query optimiser an easier time if the table is suitable indexed

the basic rule for grouping is that to select a column you need to either be grouping by it or using an aggregate function such as COUNT()

so you'll probably want something like

select
  count (e.projectid) PROJECTS,
  count (u.userid) USERS,
  to_char(e.eventdatetime,'MM') MONTH

from table1 e
  join table2 u
    on ( e.a = u.a )

where e.FILTERING_COLUMN = '1'
  and e.eventdatetime >= sysdate-30

group by e.eventdatetime

though this won't be the most optimal way to do this it would be easier if you posted the schemas involved in the issue