oracle - SQL Grouping by Ranges -

i have data set has timestamped entries on various sets of groups.

timestamp -- group -- value --------------------------- 1         --     -- 10 2         --     -- 20 3         -- b     -- 15 4         -- b     -- 25 5         -- c     -- 5 6         --     -- 5 7         --     -- 10 

i want sum these values group field, parsed appears in data. example, above data result in following output:

group  --  sum      --  30 b      --  40 c      --  5      --  15 

i not want this, i've been able come on own far:

group  --  sum      --  45 b      --  40 c      --  5 

using oracle 11g, i've hobbled togther far. know wrong, i'm hoping i'm @ least on right track rank(). in real data, entries same group 2 timestamps apart, or 100; there 1 entry in group, or 100 consecutive. not matter, need them separated.

with sub_q   (select k_id     , grp     , val     -- rank timestamp separate groups same name     , rank() over(partition k_id order tmstamp) rnk   my_table   k_id = 123) select t1.k_id   , t1.grp   , sum(case     when t1.grp = t2.grp         t1.val     else         0     end) total_value sub_q t1 -- main value inner join sub_q t2 -- timstamp after on t1.k_id = t2.k_id   , t1.rnk = t2.rnk - 1 group t1.k_id   , t1.grp 

is possible group in way? how go doing this?

i approach problem defining group different of 2 row_number():

select group, sum(value) (select t.*,              (row_number() on (order timestamp) -               row_number() on (partition group order timestamp)              ) grp       my_table t      ) t group group, grp order min(timestamp); 

the difference of 2 row numbers constant adjacent values.