assuming  Document.StreamGC is of type %Stream.Object you may
use method FindAt which combines Read() and Contains()
see: method FindAt() 

method FindAt(position As %Integer, target As %RawString,
   ByRef tmpstr As %RawString = "", caseinsensitive As %Boolean = 0) as %Integer

Find the first occurrence of target in the stream, starting the search at position.
The method returns the position of this match, counting from the beginning of 
the stream, and leaves the stream positioned at an indeterminate location. 
If it does not find the target string, it returns -1.

If position=-1 then it starts searching from the location found in the 
previous search and returns the offset from the last search. 
This is useful for searching through the entire file. 
If you are doing this, you should pass in tmpstr by reference in every call. 
This is used to store the last buffer read, so the next call will start 
where the last one left off. 
If caseinsensitive=1 then the search will be case insensitive, 
rather than the default case-sensitive search.

A very dirty hack:  possible not really a recommendation

IF Source table is not mapped to ONS

1. copy class definition including <Storage> from ENS to ONS
2. in <STORAGE> you see

3. <DataLocation>^Sample.EmployeeD</DataLocation>
   <IdLocation>^Sample.EmployeeD</IdLocation>
   <IndexLocation>^Sample.EmployeeI</IndexLocation>
   <StreamLocation>^Sample.EmployeeS</StreamLocation> 

4. Now replace all  ^  by ^|"ENS"|  looks similar 
   For my example, it's ^|"SAMPLES"|

5. <DataLocation>^|"SAMPLES"|Sample.EmployeeD</DataLocation>
   <IdLocation>^"|SAMPLES"|Sample.EmployeeD</IdLocation>
   <IndexLocation>^|"SAMPLES"|Sample.EmployeeI</IndexLocation>
   <StreamLocation>^|"SAMPLES"|Sample.EmployeeS</StreamLocation> 

6. Extended Global references allow access to the other namespace. Nothing new, but risky

YES Sir!

• Process private globals live in IRISTEMP and are strictly bound to the process and its lifetime.
• The package with %ZZ.* makes sure it is visible from ALL namespaces
• At process start it is always empty
  • you fill it implicitely by CREATE at definiftion time
  • or if reused by 
    INSERT INTO %ZZ.resultset SELECT .......
  • this SELECT must be the same as in CREATE ...
    to  fit  column names
• Either CREATE once + INSERT later works
• Or DROP TABLE %ZZ..... and fresh CREATE ...
• It's a matter of taste, I tried both

Warning: 
Test from MGMT Portal (SQL) doesn't work as the process 
(and its ID) in background changes. The process stops and the content is gone.

IF your table must survive several processes, you need some 
other tricky solution based on the same basic idea of %ZZ*
combined with a manually tuned Class definition.