CQL Overlapping Pattern Matching

Question

Hi,
I need to match the pattern "A" followed by "B" in a series of events but the pattern may overlap so if I have the sequence "A","A","A","B","B","B" the pattern should match 3 times. I've tried this with and without the ALL MATCHES clause but cannot get the desired behaviour. Is this possible in CQL?
My event is
<wlevs:event-type type-name="simpleEvent">
          <wlevs:properties>
               <wlevs:property name="DataField" type="char" />
               <wlevs:property name="Id" type="int" />
          </wlevs:properties>
</wlevs:event-type>
and my pattern matching CQL is:
select T.Id1 as Id1, T.Id2 as Id2
from
EventChannel
MATCH_RECOGNIZE (
MEASURES A.Id as Id1, C.Id as Id2
ALL MATCHES
PATTERN (A B*? C)
DEFINE A as A.DataField = 'A',
B as B.DataField != 'B',
C as C.DataField = 'B'
) as T
Thanks
Mike

Answer 1

As per your requirement, it looks like the requirement is to match Nth occurrence of A with Nth occurrence of B(counted after at least one A has been seen), in a stream of inputs containing A's and B's in any order.
Here the pattern matching is done in two steps as follows:
1. An intermediate view helloWorldView is created. It has an additional field numOccur. Corresponding to every input on helloWorldInputChannel, the view has one output and the value of numOccur field in that output is the number of events seen so far, whose DataField value is the same as the DataField value of the current input. Leading B's (prior to any A) will have numOccur as 0. So effectively, for each occurrence of 'A' and 'B' we have a number representing which occurrence it is i.e. if it is the 1st occurrence or 2nd occurrence etc.
2. One more pattern-matching query is applied on top of the output of helloWorldView. Here we start a match whenever we encounter an 'A' and end the match whenever we get a B, whose numOccur field is equal to that of A.
The input stream helloworldInputChannel has the schema - (eid int, DataField char).
<view id="helloWorldView">
<![CDATA[
select T.eid as eid, T.DataField as DataField, T.numOccur as numOccur
from helloworldInputChannel MATCH_RECOGNIZE (
MEASURES
eid as eid,
DataField as DataField,
(case
when((DataField = 'A') and (firstOccur.DataField = 'A')) then count(A.*) + 1
when((DataField = 'A') and (firstOccur.DataField != 'A')) then count(A.*)
when(DataField = 'B') then count(B.*)
when(firstOccur.DataField = 'A') then 1
end) as numOccur
ALL MATCHES
PATTERN(firstOccur ignoreB* (A | B)*)
DEFINE
firstOccur as prev(firstOccur.DataField) is null,
ignoreB as (ignoreB.DataField = 'B' and (firstOccur.DataField = 'B') and (count(A.*) = 0)),
A as A.DataField = 'A',
B as(B.DataField = 'B' and ((firstOccur.DataField = 'A') or (count(A.*) > 0)))
) as T
]]>
</view>
<query id="helloworldRule">
<![CDATA[
            select
                   T.Id1 as Id1,
                   T.Id2 as Id2
            from
                    helloWorldView MATCH_RECOGNIZE (
                        MEASURES
                                 A.eid as Id1, 
                                 C.eid as Id2
                        ALL MATCHES
                        PATTERN (A  B* C)
                        DEFINE 
                               A as A.DataField = 'A',
                               C as (C.DataField = 'B' and C.numOccur = A.numOccur)
                    ) as T
        ]]>
</query>
Hopefully this will address your requirement.
Regards
Unmesh